WO2022264640A1 - 状態推定システム及び状態推定方法 - Google Patents
状態推定システム及び状態推定方法 Download PDFInfo
- Publication number
- WO2022264640A1 WO2022264640A1 PCT/JP2022/016221 JP2022016221W WO2022264640A1 WO 2022264640 A1 WO2022264640 A1 WO 2022264640A1 JP 2022016221 W JP2022016221 W JP 2022016221W WO 2022264640 A1 WO2022264640 A1 WO 2022264640A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- culture
- state
- medium
- unit
- target
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 318
- 238000000605 extraction Methods 0.000 claims abstract description 315
- 238000005259 measurement Methods 0.000 claims abstract description 258
- 239000012472 biological sample Substances 0.000 claims abstract description 103
- 238000009826 distribution Methods 0.000 claims abstract description 66
- 230000008569 process Effects 0.000 claims description 289
- 239000013598 vector Substances 0.000 claims description 124
- 239000000654 additive Substances 0.000 claims description 110
- 230000000996 additive effect Effects 0.000 claims description 91
- 238000005457 optimization Methods 0.000 claims description 71
- 230000005856 abnormality Effects 0.000 claims description 59
- 238000004458 analytical method Methods 0.000 claims description 51
- 238000012545 processing Methods 0.000 claims description 43
- 239000000284 extract Substances 0.000 claims description 39
- 238000001819 mass spectrum Methods 0.000 claims description 34
- 239000000126 substance Substances 0.000 claims description 14
- 108090000623 proteins and genes Proteins 0.000 claims description 11
- 102000039446 nucleic acids Human genes 0.000 claims description 10
- 108020004707 nucleic acids Proteins 0.000 claims description 10
- 150000007523 nucleic acids Chemical class 0.000 claims description 10
- 238000012163 sequencing technique Methods 0.000 claims description 10
- 238000005481 NMR spectroscopy Methods 0.000 claims description 9
- 238000011002 quantification Methods 0.000 claims description 9
- 238000001069 Raman spectroscopy Methods 0.000 claims description 8
- 238000004587 chromatography analysis Methods 0.000 claims description 8
- 238000007847 digital PCR Methods 0.000 claims description 8
- 238000011534 incubation Methods 0.000 claims description 7
- 238000001237 Raman spectrum Methods 0.000 claims description 5
- 238000012300 Sequence Analysis Methods 0.000 claims description 5
- 239000000427 antigen Substances 0.000 claims description 4
- 102000036639 antigens Human genes 0.000 claims description 4
- 108091007433 antigens Proteins 0.000 claims description 4
- 238000009395 breeding Methods 0.000 claims description 4
- 230000001488 breeding effect Effects 0.000 claims description 4
- 238000003860 storage Methods 0.000 claims description 2
- 230000000977 initiatory effect Effects 0.000 claims 1
- 239000001963 growth medium Substances 0.000 abstract description 79
- 238000012827 research and development Methods 0.000 abstract description 19
- 239000002609 medium Substances 0.000 description 333
- 210000004027 cell Anatomy 0.000 description 299
- 238000004949 mass spectrometry Methods 0.000 description 174
- 238000012258 culturing Methods 0.000 description 88
- 239000000306 component Substances 0.000 description 87
- 230000004069 differentiation Effects 0.000 description 52
- 238000012937 correction Methods 0.000 description 45
- 241000588724 Escherichia coli Species 0.000 description 32
- 238000011282 treatment Methods 0.000 description 21
- 238000000513 principal component analysis Methods 0.000 description 19
- 238000010801 machine learning Methods 0.000 description 17
- 239000000047 product Substances 0.000 description 17
- 238000010586 diagram Methods 0.000 description 16
- 230000001939 inductive effect Effects 0.000 description 15
- 239000007788 liquid Substances 0.000 description 12
- 239000012138 yeast extract Substances 0.000 description 12
- 238000004113 cell culture Methods 0.000 description 11
- 238000013531 bayesian neural network Methods 0.000 description 10
- 229940041514 candida albicans extract Drugs 0.000 description 10
- 210000004748 cultured cell Anatomy 0.000 description 10
- 150000002500 ions Chemical class 0.000 description 10
- 238000013527 convolutional neural network Methods 0.000 description 9
- 239000000523 sample Substances 0.000 description 9
- 238000003384 imaging method Methods 0.000 description 8
- 238000007637 random forest analysis Methods 0.000 description 8
- 230000009471 action Effects 0.000 description 7
- 230000000694 effects Effects 0.000 description 7
- 238000005194 fractionation Methods 0.000 description 7
- 238000004519 manufacturing process Methods 0.000 description 7
- 238000010186 staining Methods 0.000 description 7
- 238000013019 agitation Methods 0.000 description 6
- 239000012530 fluid Substances 0.000 description 6
- 230000035772 mutation Effects 0.000 description 6
- 210000000130 stem cell Anatomy 0.000 description 6
- 230000002159 abnormal effect Effects 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 5
- 210000004978 chinese hamster ovary cell Anatomy 0.000 description 5
- 210000003743 erythrocyte Anatomy 0.000 description 5
- 238000000491 multivariate analysis Methods 0.000 description 5
- 235000018102 proteins Nutrition 0.000 description 5
- 102000004169 proteins and genes Human genes 0.000 description 5
- 239000002994 raw material Substances 0.000 description 5
- 239000007864 aqueous solution Substances 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 4
- 239000000203 mixture Substances 0.000 description 4
- 150000001413 amino acids Chemical class 0.000 description 3
- 229960000074 biopharmaceutical Drugs 0.000 description 3
- 210000000601 blood cell Anatomy 0.000 description 3
- 235000014633 carbohydrates Nutrition 0.000 description 3
- 150000001720 carbohydrates Chemical class 0.000 description 3
- 230000012010 growth Effects 0.000 description 3
- 150000002632 lipids Chemical class 0.000 description 3
- 239000012533 medium component Substances 0.000 description 3
- 244000005700 microbiome Species 0.000 description 3
- 238000010238 partial least squares regression Methods 0.000 description 3
- 239000000243 solution Substances 0.000 description 3
- 239000002699 waste material Substances 0.000 description 3
- 241000894006 Bacteria Species 0.000 description 2
- 102000004127 Cytokines Human genes 0.000 description 2
- 108090000695 Cytokines Proteins 0.000 description 2
- 102000004190 Enzymes Human genes 0.000 description 2
- 108090000790 Enzymes Proteins 0.000 description 2
- 210000001772 blood platelet Anatomy 0.000 description 2
- 210000004413 cardiac myocyte Anatomy 0.000 description 2
- 230000024245 cell differentiation Effects 0.000 description 2
- 238000000451 chemical ionisation Methods 0.000 description 2
- 238000000132 electrospray ionisation Methods 0.000 description 2
- 238000010265 fast atom bombardment Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 210000000265 leukocyte Anatomy 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000000816 matrix-assisted laser desorption--ionisation Methods 0.000 description 2
- 210000003716 mesoderm Anatomy 0.000 description 2
- 230000004060 metabolic process Effects 0.000 description 2
- 239000002207 metabolite Substances 0.000 description 2
- 238000007481 next generation sequencing Methods 0.000 description 2
- 238000004611 spectroscopical analysis Methods 0.000 description 2
- 239000000758 substrate Substances 0.000 description 2
- 241000203069 Archaea Species 0.000 description 1
- 244000063299 Bacillus subtilis Species 0.000 description 1
- 235000014469 Bacillus subtilis Nutrition 0.000 description 1
- 241000192700 Cyanobacteria Species 0.000 description 1
- 238000001712 DNA sequencing Methods 0.000 description 1
- 241000233866 Fungi Species 0.000 description 1
- 241000238631 Hexapoda Species 0.000 description 1
- 206010020843 Hyperthermia Diseases 0.000 description 1
- 240000004808 Saccharomyces cerevisiae Species 0.000 description 1
- 241000700605 Viruses Species 0.000 description 1
- 235000001014 amino acid Nutrition 0.000 description 1
- 239000012491 analyte Substances 0.000 description 1
- 210000004102 animal cell Anatomy 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 239000007853 buffer solution Substances 0.000 description 1
- 239000006227 byproduct Substances 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000000747 cardiac effect Effects 0.000 description 1
- 230000010261 cell growth Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000003153 chemical reaction reagent Substances 0.000 description 1
- 238000004925 denaturation Methods 0.000 description 1
- 230000036425 denaturation Effects 0.000 description 1
- 235000014113 dietary fatty acids Nutrition 0.000 description 1
- 210000001671 embryonic stem cell Anatomy 0.000 description 1
- 229930195729 fatty acid Natural products 0.000 description 1
- 239000000194 fatty acid Substances 0.000 description 1
- 150000004665 fatty acids Chemical class 0.000 description 1
- 210000003754 fetus Anatomy 0.000 description 1
- 238000000684 flow cytometry Methods 0.000 description 1
- 238000000799 fluorescence microscopy Methods 0.000 description 1
- 238000010353 genetic engineering Methods 0.000 description 1
- 229910001385 heavy metal Inorganic materials 0.000 description 1
- 210000003494 hepatocyte Anatomy 0.000 description 1
- 210000005260 human cell Anatomy 0.000 description 1
- 230000036031 hyperthermia Effects 0.000 description 1
- 238000012744 immunostaining Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000005040 ion trap Methods 0.000 description 1
- 230000001678 irradiating effect Effects 0.000 description 1
- 210000004153 islets of langerhan Anatomy 0.000 description 1
- 238000001698 laser desorption ionisation Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 235000013372 meat Nutrition 0.000 description 1
- 210000001704 mesoblast Anatomy 0.000 description 1
- 238000000386 microscopy Methods 0.000 description 1
- 231100000350 mutagenesis Toxicity 0.000 description 1
- 239000003471 mutagenic agent Substances 0.000 description 1
- 231100000707 mutagenic chemical Toxicity 0.000 description 1
- 230000003505 mutagenic effect Effects 0.000 description 1
- 230000002107 myocardial effect Effects 0.000 description 1
- 210000004165 myocardium Anatomy 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 235000015097 nutrients Nutrition 0.000 description 1
- 238000011369 optimal treatment Methods 0.000 description 1
- 108090000765 processed proteins & peptides Proteins 0.000 description 1
- 102000004196 processed proteins & peptides Human genes 0.000 description 1
- 230000035755 proliferation Effects 0.000 description 1
- 210000001525 retina Anatomy 0.000 description 1
- 230000002207 retinal effect Effects 0.000 description 1
- 150000003839 salts Chemical class 0.000 description 1
- 238000007480 sanger sequencing Methods 0.000 description 1
- 210000002966 serum Anatomy 0.000 description 1
- 235000002639 sodium chloride Nutrition 0.000 description 1
- 239000002904 solvent Substances 0.000 description 1
- 238000003756 stirring Methods 0.000 description 1
- 239000006228 supernatant Substances 0.000 description 1
- 239000013076 target substance Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 210000004881 tumor cell Anatomy 0.000 description 1
- 238000011144 upstream manufacturing Methods 0.000 description 1
- 235000013343 vitamin Nutrition 0.000 description 1
- 239000011782 vitamin Substances 0.000 description 1
- 229940088594 vitamin Drugs 0.000 description 1
- 229930003231 vitamin Natural products 0.000 description 1
Images
Classifications
-
- C—CHEMISTRY; METALLURGY
- C12—BIOCHEMISTRY; BEER; SPIRITS; WINE; VINEGAR; MICROBIOLOGY; ENZYMOLOGY; MUTATION OR GENETIC ENGINEERING
- C12M—APPARATUS FOR ENZYMOLOGY OR MICROBIOLOGY; APPARATUS FOR CULTURING MICROORGANISMS FOR PRODUCING BIOMASS, FOR GROWING CELLS OR FOR OBTAINING FERMENTATION OR METABOLIC PRODUCTS, i.e. BIOREACTORS OR FERMENTERS
- C12M29/00—Means for introduction, extraction or recirculation of materials, e.g. pumps
- C12M29/26—Conditioning fluids entering or exiting the reaction vessel
-
- C—CHEMISTRY; METALLURGY
- C12—BIOCHEMISTRY; BEER; SPIRITS; WINE; VINEGAR; MICROBIOLOGY; ENZYMOLOGY; MUTATION OR GENETIC ENGINEERING
- C12M—APPARATUS FOR ENZYMOLOGY OR MICROBIOLOGY; APPARATUS FOR CULTURING MICROORGANISMS FOR PRODUCING BIOMASS, FOR GROWING CELLS OR FOR OBTAINING FERMENTATION OR METABOLIC PRODUCTS, i.e. BIOREACTORS OR FERMENTERS
- C12M41/00—Means for regulation, monitoring, measurement or control, e.g. flow regulation
- C12M41/48—Automatic or computerized control
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
Definitions
- the present invention relates to a state estimation system and a state estimation method.
- Natural media and other biologically derived raw materials are widely used as media, samples, reagents, etc. used in cell culture, but it is known that there are lot and batch differences in biologically derived raw materials. This is a direct cause of variations in the quality of cell culture, and in the processes of bioproduction, research and development, and inspection, it is often the case that even if the process is carried out under exactly the same conditions, the results of the execution will differ greatly. occur.
- Analyzing the state of the medium before performing cell culture using the medium is important in increasing the efficiency of bioproduction and research and development using the medium, for example, as shown in Patent Document 1.
- an object of the present invention is to provide a state estimation system and a state estimation method that can improve the efficiency of bioproduction and research and development using biological samples.
- the state estimation system includes any one of a medium, a culture cultured in the medium, a medium extract extracted from the medium, and a culture extract extracted from the culture.
- a database that stores a feature quantity extraction model that characterizes the distribution of input data that includes at least characteristic values of the plurality of biological samples acquired via a measurement unit;
- a state estimating unit for estimating a culture state of the culture or estimating a culture time for the culture to reach a predetermined culture state from the input data;
- the state estimation method includes a medium, a culture cultured in the medium, a medium extract extracted from the medium, and a culture extract extracted from the culture, Either of the biological samples and the target biological sample is measured by a measurement unit for the characteristics of the plurality of biological samples at a certain point in time and at least one target biological sample of the same type as the biological samples at the point in time.
- a measuring step a storage step of storing in a database a feature quantity extraction model characterizing the distribution of input data including at least characteristic values of the plurality of biological samples acquired through the measuring unit; based on the input data of the target biological sample, the state estimating unit estimates the culture state of the culture, or the state estimating unit estimates the culture time for the culture to reach a predetermined culture state. and a state estimation step.
- the present invention it is possible to estimate in advance whether or not a desired culture state can be obtained in bioproduction and research and development using a target biological sample.
- the time required for proper judgment can be shortened.
- the efficiency of bio-production and research and development using biological samples can be enhanced.
- FIG. 4 is a schematic diagram for explaining the timing of measuring medium characteristics and the timing of measuring the culture state of a culture.
- 4 is a graph showing an example mass spectrum obtained from a mass spectrometer and an example measurement vector generated from the mass spectrum;
- 2 is a table showing the results of culturing Escherichia coli using each sample.
- FIG. 10 is a table showing the results of culturing Escherichia coli using other samples and the state estimation results predicted by the state estimation system;
- FIG. 7 is a graph showing the relationship between actual culture results and state estimation results;
- FIG. 10 is a graph showing measurement vectors for additives;
- FIG. 10 is a table showing the inner product between the additive measurement vector and the principal component vector.
- FIG. It is a table
- FIG. 10 is a schematic diagram for explaining another embodiment regarding the measurement timing of medium properties.
- FIG. 4 is a schematic diagram for explaining another embodiment regarding the measurement timing of the culture state of the culture.
- FIG. 11 is a block diagram showing configuration (2) of another state estimation system;
- FIG. 2 is a schematic diagram for explaining an example in which a culture A is induced to differentiate into a culture B, a culture C, or a non-target cell by culturing. It is a block diagram which shows the structure of the state estimation system which concerns on 4th Embodiment.
- FIG. 4 is a schematic diagram for explaining a method of sorting culture A in the sorting section; 16A is a schematic diagram for explaining an example in which culture A is differentiated into culture E or non-target cells by culturing, and 16B is a schematic diagram for explaining an example in which culture A is cultured into culture D or non-target cells by culturing.
- FIG. 2 is a schematic diagram for explaining an example in which differentiation is induced and culture D differentiates into culture E or non-target cells by culturing.
- FIG. 1 is a block diagram showing the configuration of a state estimation system 1a according to this embodiment.
- the example of a culture medium is shown as an example of a biological sample.
- the first embodiment is an example in which a mass spectrometer is applied as the measurement unit 2, which will be described later.
- the state estimation system 1a estimates the state of the culture medium and the culture state of the culture used in the cell culture experiment based on the feature amount extraction model that defines the feature amount. , an arithmetic processing unit 3 a , a database 4 , and a notification unit 5 .
- the "culture state” indicates, for example, the state of yield of the culture per unit volume of the medium contained in the medium used in the culture process, or the state of the quality of the culture.
- the yield of the culture includes the yield of the culture extract extracted from the cultivated culture and the yield of components in the medium secreted by the cultivated culture.
- the yield of the culture extract includes the yield of useful proteins (eg, enzymes) extracted from Escherichia coli.
- the yield of a component in the medium secreted by the cultured culture includes the yield of a biopharmaceutical (eg, antibody) extracted from the medium.
- the state of yield of the culture per unit volume of the medium contained in the medium used in the culture process will be mainly described as the culture state.
- the quality of the culture indicates the functionality of the culture. It corresponds to the index indicating the quality such as whether or not the cell has the quality, or the index indicating the differentiation state of the cell as the quality.
- the quality of the culture as the culture state of the biological sample includes the quality of the culture extract extracted from the cultured culture and the quality of the components in the medium secreted by the cultured culture. For example, if the culture is Escherichia coli, the quality of the culture extract is expressed as an indicator of how efficiently the useful proteins (e.g., enzymes) extracted from Escherichia coli have good reaction efficiency (whether they have functionality). There are things, etc.
- the quality of the components in the medium secreted by the culture refers to the extent to which the biopharmaceutical (e.g., antibody) extracted from the medium has the intended function. Those indicated by the index are applicable.
- the measurement unit 2 is a mass spectrometer, and obtains a mass spectrum by mass spectrometry of the culture medium.
- a mass spectrometer is applied as a measurement unit
- the present invention is not limited to this, and for example, an impedance sensor, a microscope device, a Raman spectrometer, chromatography, digital PCR measurement
- An apparatus, a nuclear magnetic resonance (Nuclear Magnetic Resonance) apparatus, an antibody quantification kit, or a nucleic acid sequencing apparatus may be applied as the measurement unit. Examples in which these impedance sensors, microscope devices, and the like are applied as measurement units will be described later.
- the mass spectrometer as the measurement unit 2 generates ions by irradiating a medium mixed with a matrix, a solvent, etc. with a laser beam, detects the ions with an ion detector, and detects the ions from the ion detector.
- a mass spectrum is generated based on the detected signals of
- Mass spectrometers include, for example, EI (Electron Ionization) method, CI (Chemical Ionization) method, FAB (Fast Atom Bombardment) method, MALDI (Matrix Assisted Laser Desorption Ionization, matrix assisted A mass spectrometer using an ion source such as a laser desorption ionization) method or an ESI (Electro Spray Ionization) method may be applied.
- the mass spectrometer may be a magnetic sector type, quadrupole type, ion trap type, or time-of-flight type mass spectrometer using a mass separator (analyzer), an electron multiplier or a microchannel plate. You may apply the mass spectrometer etc. which used ion detectors, such as.
- the term "medium” includes nutrients used for the growth of cells, microorganisms or viruses (hereinafter simply referred to as cultures), and provides a growth environment for the cultures. It is a medium for The medium according to the present embodiment can be any known medium used for culturing various cultures, and is appropriately determined depending on the type of culture. Examples of media include buffer solutions containing inorganic salts, carbohydrates, amino acids, vitamins, proteins, peptides, fatty acids, lipids, serum and the like.
- the "cell” is not particularly limited as long as it can be cultured in a medium, and examples thereof include insect cells, animal cells, human cells, and plant cells.
- Leaf stem cells embryonic stem cells, hepatocytes, corneal stem cells, pancreatic islet cells, tumor cells, iPS cells, ES cells, blood cells and the like.
- a specific example is 293T cells.
- the term “cell” also includes cell tissue, which is an aggregate of cells, such as artificial meat, retinal tissue, myocardial sheet, and multicellular biological tissue (eg, fetus, plant, etc.).
- microorganisms include bacteria, fungi, microalgae, protozoa, yeasts, and the like.
- Bacillus subtilis include cyanobacteria, and the like.
- each medium for each lot or batch which may have differences in components due to lot or batch differences, is referred to as one medium.
- plural of media ie, media prepared from different lots or media prepared from different batches, they are referred to as "plurality of media.”
- the state estimation system 1a performs mass spectrometry on a plurality of culture media using the measurement unit 2, and generates a feature quantity extraction model based on the obtained mass spectrometry data.
- the mass spectrometry data refers to the mass spectrum measured when the medium is subjected to mass spectrometry by the measurement unit 2, or a measurement vector (described later) generated from the mass spectrum, which is the characteristic value of the medium. It is what you get.
- the state estimation system 1a performs mass spectrometry on the medium with unknown components in the measurement unit 2, and based on the resulting mass spectrometry data and the above feature extraction model, the medium with unknown components It is possible to estimate the culture state of the culture when using.
- a biological sample for estimating the culture state of a culture using the feature quantity extraction model thus generated is referred to as a "target biological sample”.
- a biological sample for estimating the culture state of a culture using the feature quantity extraction model thus generated is referred to as a "target biological sample”.
- the measurement unit 2 performs mass spectrometry on each of a plurality of culture media, and outputs the obtained mass spectra to the arithmetic processing unit 3a and the database 4.
- mass spectra mass spectrometry data
- the database 4 associates the mass spectra of each culture medium with the dates and times when the mass spectra were measured and stores them.
- the information of the background of the medium is, for example, the composition of the medium (specification information), the model number, the lot number, the date of manufacture, the date of arrival, the past usage history, etc.
- the database 4 also stores various data such as the feature quantity extraction model generated by the arithmetic processing unit 3a.
- the arithmetic processing unit 3a has a state estimating unit 10a, a determining unit 12, and a measurement target optimization estimating unit 14. Mass spectra of a plurality of culture media are input from the measurement unit 2 or the database 4 to the state estimation unit 10a.
- the state estimating unit 10a the mass spectrometry data of a plurality of culture media obtained via the measurement unit 2 at a certain time, and the culture of the culture when the culture is actually cultured using the medium Using the state results (hereinafter also referred to as culture results), for example, multivariate analysis is performed to generate a feature quantity extraction model.
- a feature quantity extraction model it is possible to analyze the components of the culture medium that affect the culture result when the culture is cultured using the culture medium.
- the feature extraction model is an estimation result of the future culture state that is finally obtained when it is assumed that the mass spectrometry data of the target medium is input and the culture is continued as it is with the target medium. It is desirable to apply a regression model or the like that outputs a state estimation result (hereinafter also referred to as a state estimation result).
- a state estimation result hereinafter also referred to as a state estimation result.
- the correspondence between the mass spectrometry data of multiple culture media and the culture results when using these media is learned, and the mass spectrometry data is obtained from the culture results. It is desirable to characterize the distribution of
- each culture result when using these culture media and whether the culture result is the desired result (for example, the correct answer rate indicating whether or not the desired culture result is obtained), and the correct label is attached when learning the mass spectrometry data of the medium and the culture result, and the mass A distribution of analytical data may be characterized.
- Feature extraction models include, for example, principal component analysis (PCA), partial least squares (PLS), polynomial regression, Gaussian process regression, random forest regression, and various other features. Extraction models can be applied. Also, as the feature quantity extraction model, for example, a machine learning model generated by machine learning the characteristic values (mass spectrometry data) of the culture medium and the culture results may be applied.
- PCA principal component analysis
- PLS partial least squares
- polynomial regression Gaussian process regression
- random forest regression random forest regression
- various other features Extraction models can be applied.
- the feature quantity extraction model for example, a machine learning model generated by machine learning the characteristic values (mass spectrometry data) of the culture medium and the culture results may be applied.
- a neural network eg, CNN (Convolutional Neural Network) or BNN (Bayesian Neural Network)
- unsupervised learning may also be used for the feature quantity extraction model. The details of generating a feature quantity extraction model by unsupervised learning will be described later, but when generating a feature quantity extraction model by unsupervised learning, mass spectrometry data of a plurality of culture media are used. It is desirable to allow learning to characterize the distribution of data.
- the determination unit 12 determines the state of the target culture medium based on the estimation result (state estimation result) of the culture state of the culture obtained by the state estimation unit 10a.
- the determination unit 12 adds, in a state estimation space defined by a principal component vector (feature amount) in the feature amount extraction model, an identification value capable of distinguishing between a good medium and a bad medium in stages. Then, the mass spectrometry data of the target culture medium is projected onto the state estimation space, and the state of the target culture medium is estimated using the identification value as a guide (state determination processing).
- the mass of the culture per unit volume of the medium finally obtained when culturing the culture using the medium is a predetermined optimum set value (hereinafter referred to as a good judgment value)
- a good judgment value A medium having the above-mentioned quality
- a medium having a mass equal to or less than a predetermined set value lower than the good judgment value (hereinafter referred to as a bad judgment value) is called a "poor medium”.
- the determination unit 12 may set a threshold for abnormality determination in the state estimation space in which the principal component vector is defined in the feature quantity extraction model (abnormality determination processing). As a result, if the state estimation vector obtained by projecting the mass spectrometry data of the target medium into the state estimation space does not exceed the threshold, the determination unit 12 determines that the target medium provides good culture results. On the other hand, if it is less than the threshold value, it can be determined that the target medium gives poor culture results.
- a medium whose characteristics do not meet the required standards due to lot differences, batch differences, etc. of the medium is referred to as an abnormal medium, and the characteristics of the medium do not meet the required standards.
- the medium that fills is called normal medium.
- the measurement target optimization estimating unit 14 estimates additives to be added to the target medium in order to obtain a target medium that provides a desired culture state (measurement target optimization estimation process). As a result, even if the target medium is estimated that the desired culture state cannot be obtained in the state estimation unit 10a, the operator or the like can add an appropriate additive to the target medium to achieve the desired culture state. It can be corrected to the target medium obtained.
- the measurement target optimization estimating unit 14 estimates what additives should be added to the target medium to obtain the desired culture result
- the obtained additive estimation result is sent to the notification unit 5 described later.
- output to The additive estimation result obtained by the measurement target optimization estimation unit 14 is presented to the operator or the like via the notification unit 5 .
- the operator or the like can perform the culture process in the target medium with improved performance by adding the additive (estimated additive) presented by the state estimation system 1a to the target medium. can be attempted to obtain a culture result of
- the notification unit 5 displays, for example, the state estimation result, the determination result, etc. of the state estimation unit 10a on the display unit so that the worker or the like can visually recognize it, and also transmits an e-mail to the worker or the like. It is possible to apply a transmitting device that notifies the state estimation result, the determination result, etc., and a sound emitting device, etc. that notifies the state estimation result, the determination result, etc. to the worker or the like by voice.
- the operator or the like is an operator of the state estimation system 1a, an experimenter of the culture experiment, or an arbitrarily determined manager or the like.
- the notification unit 5 also receives the feature quantity extraction model generated by the arithmetic processing unit 3a, the measurement result (mass spectrum and measurement vector) in the measurement unit 2, and the estimated additive obtained by the measurement target optimization estimation unit 14.
- a notification device, a transmission device, or the like can be used to let the worker or the like grasp the content.
- the culture state of the culture estimated by the state estimating unit 10a may be the quality (functionality) of the culture obtained in the culture process as described above. The case where the yield of the culture obtained is used will be mainly described.
- FIG. 2 is a schematic diagram showing the timing of measuring the mass spectrometry data of a plurality of culture media and the culture state of the culture, which are acquired when the feature quantity extraction model is generated by the state estimation unit 10a.
- S indicates the start time of the culture process
- F indicates the end time after a predetermined culture period has passed in the culture process.
- starting time S of the culture process indicates when the culture starts to be cultured with a medium under culture conditions that define predetermined culture parameters
- the "end time F of the culture process” refers to the time according to the medium. Shows when the culture is finished culturing.
- the state estimation system 1a according to the present embodiment, as shown in FIG.
- mass spectrometry data of a plurality of culture media are acquired through the measurement unit 2 before the start of the culture process (before disclosure time S) (Sa1 ).
- the state estimation system 1a after the culture process is completed (after the termination period F), the yield of the culture after actually culturing the culture with each medium is examined, and the culture state of the culture is measured. (Fa1).
- the thus obtained mass spectrometry data of the medium before the start of the culture process (before use for culture) and the mass spectrometry data of the culture contained in the medium after the end of the culture process (after use for culture) Characterize the relationship between yield (i.e., culture results) and mass spectrometry data measured on pre-cultivation subject media of unknown composition to determine future potential when the subject media are used in the culture process. This is to estimate the yield of the culture (culturing state of the culture) obtained in This makes it possible to estimate whether or not the target medium provides the desired culture conditions before starting the culture process.
- E. coli for example, an example of culturing E. coli as a culture using a medium supplemented with yeast extract will be described.
- a culture of E. coli even if the same amount of yeast extract is added to the same medium, the growth performance of E. coli may differ depending on the lot of the yeast extract.
- the state estimation system 1a the mass spectrometry data of each medium to which different yeast extracts are added, and the culture results when E. coli is cultured using the media to which these different yeast extracts are added.
- a feature extraction model is constructed to discover relationships and clearly detect functional differences between these media.
- the state estimation system 1a can estimate the state of the yeast extract in the medium, which is related to the growth performance of E. coli, for a target medium whose components are unknown. .
- mass spectrometry is performed by the measurement unit 2 on multiple types of yeast extract aqueous solutions (medium) from multiple different lots to obtain a mass spectrometry mass spectrum for each yeast extract aqueous solution.
- 3A, 3B and 3C of FIG. 3 show examples of mass spectra obtained by mass spectrometry of different types of media (yeast extract aqueous solution).
- the horizontal axis is m/z and the vertical axis is signal intensity.
- the peak intensity in the mass spectrum corresponds to the amount of ions with each m/z produced by the ionization of the analyte (medium).
- the state estimating unit 10a When the state estimating unit 10a according to the present embodiment receives the mass spectrum obtained by mass spectrometry from the measuring unit 2 or the database 4, the mass spectrum is standardized and coarse-grained to reduce the amount of signal intensity information. Generate a vector. Specifically, each peak intensity is corrected so that the maximum peak intensity of the signal intensity of the mass spectrum becomes 100 as standardization. Further, as coarse-graining, the horizontal axis m / z of the mass spectrum is divided into predetermined dimensions (here, 30 dimensions), the signal intensity is averaged for each dimension, and the peak intensity for the maximum peak intensity of 100 is calculated for each dimension. Obtain the measurement vector shown. In this embodiment, the measurement vector obtained from the mass spectrum is the mass spectrometry data, and is used as the characteristic value of the target medium.
- 3D, 3E and 3F in FIG. 3 show examples of measurement vectors, 3D showing the measurement vector generated from the mass spectrum of 3A and 3E showing the measurement vector generated from the mass spectrum of 3B.
- 3F denotes the measurement vector generated from the mass spectrum of 3C.
- the horizontal axis m / z of the mass spectrum by mass spectrometry is divided into a predetermined number of dimensions, the maximum peak intensity of the signal intensity is defined as 100 for each dimension, and within each dimension An example is shown in which the peak intensity with respect to the maximum peak intensity of 100 is obtained for each dimension by subtracting the maximum value and the minimum value.
- the database 4 also stores measurement vectors generated by the state estimation unit 10a. Further, in the database 4, for example, as the culture result, as shown in FIG. 4, the final culture result when actually culturing Escherichia coli, which is a culture product, using the medium is stored. In FIG. 4, the media from which the culture results were obtained are denoted as samples 0, 1, 2, . . . and the actual culture results are shown for each medium. Also, "train" shown in FIG. 4 indicates that the data attribute is sample data for generating a feature quantity extraction model.
- the same E. coli is actually cultured, and E. coli is recovered from the culture solution after the E. coli is cultured, and the dry mass of E. coli per unit volume of the medium (hereinafter simply referred to as mass or yield Also called) (g/L) is measured to obtain the culture result.
- the culture results (actual results) obtained in this way when E. coli was actually cultured using each medium are stored in the database 4 in association with the measurement vector of each medium.
- the state estimating unit 10a uses measurement vectors of a plurality of media (that is, mass spectrometry data and characteristic values of the media) as explanatory variables, and uses each medium to culture Escherichia coli (culture). PLS regression is performed using the resulting mass (yield) of E. coli per unit volume of medium (g/L) as the objective variable. PLS regression is a technique that selects the principal component vectors (basis) of the measurement vector to maximize the covariance between the material content in the measurement vector and the E. coli yield.
- the state estimating unit 10a uses highly relevant multidimensional (for example, 30-dimensional) principal component vectors (teacher ) is extracted to obtain a feature quantity extraction model.
- the feature quantity extraction model it is possible to identify the component of the medium that contributed most to the difference in the culture result based on the obtained principal component vector.
- mass spectrometry is performed by the measurement unit 2 before the start of the culture process for a target medium whose state is to be estimated and whose components are unknown, to obtain a mass spectrum.
- the state estimator 10a generates a measurement vector as mass spectrometry data from the mass spectrum of the target culture medium whose state is to be estimated.
- the state estimating unit 10a normalizes and projects the obtained measurement vector onto the principal component vector defined in the state estimation space of the feature extraction model, thereby obtaining a state estimation vector related to the target culture medium (hereinafter referred to as state estimation result ) is obtained.
- the value of the state estimation vector which is the state estimation result, is the estimated mass of E. coli per unit volume of the culture solution (i.e., the culture after the culture process is finished) when culture is performed using each yeast extract. culture state), and indicates that the more positive the state estimation vector goes, the greater the yield of E. coli expected as a result of the culture.
- the state estimating unit 10a roughly estimates the yield of E. coli (hereinafter referred to as state estimation (also referred to as results) can be output.
- the state estimation result obtained by the state estimation unit 10a is stored in the database 4 and presented to the worker or the like via the notification unit 5.
- the state estimation result output by the state estimation unit 10a is directly transmitted to the operator or the like via the notification unit 5 without performing the state determination of the target culture medium by the determination unit 12 described later.
- the state estimation system 1a based on the state estimation result of estimating the culture state of the culture, whether or not the target medium provides good culture results is determined by the operator or the like when the target medium is actually cultured. can be estimated before being used for
- FIG. 5 shows that using five types of target media (yeast extract aqueous solution) with different components, the same E. coli was actually cultured, respectively, and E. coli was recovered from the culture solution after culturing E. coli.
- target media yeast extract aqueous solution
- FIG. 10 is a table showing state estimation results (results of estimated yield of E. coli per unit volume of medium contained in target medium after use for culture) obtained from each measurement vector of the target medium.
- target media with unknown components are indicated as samples 0, 1, 2, 3, and 4.
- Test shown in FIG. 5 indicates that the attribute of the data is sample data for analyzing the target culture medium of unknown components using the generated feature quantity extraction model.
- the state estimation results obtained by the state estimation system 1a are close to the actual culture results. On the other hand, it is possible to predict the culture result that will be obtained when using the target medium without actually performing the culture process using the target medium.
- FIG. 6 is a graph showing the relationship between the state estimation results shown in FIG. 5 and the actual culture results (actual results) shown for reference.
- the horizontal axis of FIG. 6 indicates the actual culture result (actual state result), and the vertical axis indicates the state estimation result.
- FIG. 6 also shows, as a rough trend, that the state estimation results obtained by the state estimation system 1a can predict the actual culture results.
- the determination unit 12 determines the state of the target culture medium based on the state estimation result obtained by the state estimation unit 10a.
- the target medium before being used for culture is not subjected to the abnormality determination described later, and it is simply determined whether the state of the target medium is suitable for culturing the culture in a stepwise manner.
- the determination unit 12 When performing the state determination process, prior to the state determination process, the determination unit 12 stores a good medium and a bad medium in a state estimation space in which principal component vectors (supervised bases) are defined as feature amounts in the feature amount extraction model. Set an identification value that can be identified step by step between and. In addition, this identification value may be associated with an identification display that expresses the degree of goodness of the state of the medium in stages, for example, based on the culture results obtained when the medium is used for culturing the culture. .
- the identification value for the state estimation space may be arbitrarily set by an operator or the like, and the principal component vectors defined in the state estimation space are automatically divided at predetermined intervals, can be set automatically.
- the determining unit 12 determines the state of the target culture medium from the value of the state estimation vector obtained by projecting the mass spectrometry data (measurement vector) of the target culture medium onto the state estimation space and the identification value.
- the determination unit 12 may set a threshold value for abnormality determination in the state estimation space of the feature extraction model based on a rule for abnormality determination.
- the threshold may be set in the state estimation space by an operator or the like in consideration of the desired culture result. For example, the threshold is set so that if the threshold is not exceeded, a good culture result equal to or greater than the reference value is obtained, and if the threshold is exceeded, a poor culture result below the reference value is obtained.
- the determination unit 12 determines the presence or absence of abnormality in the target culture medium from the value of the state estimation vector obtained by projecting the mass spectrometry data of the target culture medium onto the state estimation space defining the principal component vector and the threshold value. Further, the degree of abnormality in the target culture medium may be determined by setting the above thresholds so as to be divided into a plurality of stages.
- the notification unit 5 notifies the operator or the like of the determination result obtained by determining the state of the target medium before use for culture based on the identification value by inputting the determination result of the target medium by the state determination process. In addition, the notification unit 5 notifies the operator or the like of the presence or absence of abnormality or the degree of abnormality by receiving the determination result of the presence or absence of abnormality or the degree of abnormality by the abnormality determination process.
- the notification unit 5 may notify the operator or the like of information indicating that there is an abnormality, including the characteristic value of the target culture medium and an error code.
- the notification unit 5 may notify the worker or the like that there is no abnormality.
- the measurement target optimization estimating unit 14 estimates the culture state of the culture (state estimation result) based on the feature extraction model from the mass spectrometry data of the target medium before the start of the culturing process. When obtained, the state estimation result is received from the state estimation unit 10a.
- the measurement target optimization estimating unit 14 analyzes the state estimation result obtained by the state estimating unit 10a, and adds it to the target medium before use for culture, so that the medium unit volume contained in the target medium after use for culture Predict additives that can maximize the yield of the culture.
- the measurement target optimization estimation unit 14 corrects the components of the target medium used in the culture process by having the operator or the like add the estimated additive (also referred to as an estimated additive) to the target medium. be.
- an additive that maximizes the yield of the culture per unit volume of the medium contained in the target medium after culture use will be described as an example.
- the invention is not limited to this.
- it may be an additive that improves the quality of the culture contained in the target medium after use for culture (for example, brings the culture to a desired state of differentiation).
- the composition of the target medium is changed by adding the appropriate type of additive to the target medium, and before the target medium is actually used in the culture process, the target medium contained in the target medium after the culture is used. It is possible to improve the yield (culture state) of the culture per unit volume of medium.
- the state estimation system 1a uses the measurement unit 2 to perform mass spectrometry on three types of additives to obtain mass spectra, and the state estimation unit 10a acquires additive measurement vectors obtained by reducing the amount of information in the mass spectra. .
- FIG. 7 is a graph showing examples of additive measurement vectors for three types of additives, FIG. 7A showing the additive measurement vector for the first additive and FIG. and FIG. 7C shows the additive measurement vector for the third additive.
- the three types of additive measurement vectors obtained by the state estimating section 10a are stored in the database 4.
- FIG. 7A showing the additive measurement vector for the first additive
- FIG. 7C shows the additive measurement vector for the third additive.
- the measurement target optimization estimation unit 14 determines the relationship (similarity) between the additive measurement vector, which is mass spectrometry data on multiple types of additives, and the principal component vector defined in the state estimation space in the feature extraction model. calculate.
- the measurement target optimization estimating unit 14 estimates the additive of the additive measurement vector that has the greatest relationship with the base of the feature extraction model from among the three types of additives as the optimal additive to be added to the target medium. do. In this way, the culture result can be improved by adding the additive of the additive measurement vector having the greatest relevance to the basis of the feature extraction model to the target medium before use for culture.
- the additive of the additive measurement vector highly related to the basis of the feature quantity extraction model has an additive measurement vector highly similar to the principal component vector defined in the state estimation space in the feature quantity extraction model. It is an additive.
- the degree of similarity between the principal component vector of the feature quantity extraction model and the additive measurement vector is obtained by calculating the inner product between the principal component vector of the feature quantity extraction model and the additive measurement vector.
- the principal component vector with the largest eigenvalue is selected, and the inner product between the selected principal component vector and the additive measurement vector is calculated.
- the inner product with the additive measurement vector may be calculated for each principal component vector, and the total value obtained by adding the calculated inner products may be obtained. In this case, among the total values obtained for each additive measurement vector, the additive with the largest total value is selected as the correction candidate additive.
- FIG. 8 is a table summarizing an example of the inner product between the principal component vector of the feature quantity extraction model and the three types of additive measurement vectors.
- the first additive is indicated as “additive0”
- the second additive is indicated as “additive1”
- the third additive is indicated as “additive2”.
- the third additive "additive2" has the largest inner product with the principal component vector of the feature quantity extraction model, and has the additive measurement vector that is most closely related to the principal component vector of the feature quantity extraction model. .
- the measurement target optimization estimation unit 14 calculates the inner product between the principal component vector of the feature quantity extraction model and the additive measurement vector as the degree of similarity between the principal component vector of the feature quantity extraction model and the additive measurement vector. is calculated, and the additive with the largest inner product is selected.
- the measurement target optimization estimating unit 14 outputs additive information such as the name of the selected additive as a correction candidate additive to the notification unit 5 .
- the notification unit 5 notifies the content of the additive information received from the measurement target optimization estimation unit 14 to the operator or the like by displaying the information.
- FIG. 9 three types of additives having additive measurement vectors shown in FIG. 7 were added to the five types of target media (samples 0 to 4) shown in FIG. E. coli is actually cultured using the target medium corrected for , and an example of the result of the yield of E. coli per unit volume of medium contained in each target medium after use for culture is shown.
- the attribute in FIG. 9 indicates that the additive was added to each target medium prepared as a sample to correct the component.
- corrected (additive2) is the third additive added to the sample
- corrected (additive1) indicates that the second additive was added to the sample
- corrected (additive0) indicates that the first additive was added to the sample.
- a third additive which has an additive measurement vector that has the highest correlation with the principal component vector of the feature extraction model, is added to each target medium after use for culture.
- the yield of E. coli per unit volume of the culture medium increases, and the range of improvement is greater than that of other additives.
- the culture state can be improved by adding the optimum additive to the target medium.
- a correction technique may be presented to an operator or the like.
- the operator can add the optimum additive estimated by the measurement target optimization estimation unit 14 to the target medium, thereby performing the culture process in the target medium with improved performance, and performing the desired culture. status can be obtained.
- the similarity between the principal component vector defined in the state estimation space in the feature quantity extraction model and the multiple types of additive measurement vectors is calculated, and the similarity with the principal component vector is calculated.
- an additive with a high additive measurement vector is specified by the measurement target optimization estimation unit 14
- the present invention is not limited to this.
- one or more additive information is associated in advance with the culture state estimation result that will be obtained by the state estimating unit 10a, and based on the feature amount extraction model, from the mass spectrometry data of the target medium
- one or more additives previously associated with the estimation result may be estimated as the optimum additive to be added to the target medium.
- the state estimation system 1a includes a measurement unit 2 that measures mass spectrometry data of a plurality of culture media and at least one target culture medium as characteristic values (input data), and a plurality of A state estimating unit 10a for estimating the culture state of the culture from the characteristic values of the target medium based on the characteristic value extraction model that characterizes the distribution of the characteristic values (input data) of the medium is provided.
- a measurement unit 2 that measures mass spectrometry data of a plurality of culture media and at least one target culture medium as characteristic values (input data)
- a state estimating unit 10a for estimating the culture state of the culture from the characteristic values of the target medium based on the characteristic value extraction model that characterizes the distribution of the characteristic values (input data) of the medium is provided.
- PCA principal component analysis
- mass spectrometry data of a plurality of culture media and multidimensional principal component vectors (unsupervised bases) that characterize the distribution of mass spectrometry data are extracted to obtain feature quantities.
- a state estimation system 1a for obtaining an extraction model will be described. Note that the measurement target optimization estimation processing performed by the measurement target optimization estimation unit 14 using the state estimation results obtained based on the feature quantity extraction model is the same as in the above-described embodiment. The description is omitted here.
- the measurement unit 2 performs mass spectrometry on a plurality of culture media, and using the obtained plurality of mass spectrometry data, performs, for example, multivariate analysis, and extracts feature quantities that can clearly detect differences between culture media. build a model; Specifically, when the state estimating unit 10a receives the mass spectrum obtained by mass spectrometry from the measuring unit 2 or the database 4, the mass spectrum is standardized and coarse-grained to reduce the amount of signal intensity information. Vectors are generated as mass spectrometry data.
- the state estimating unit 10a performs PCA processing on the measurement vectors of a plurality of culture media, extracts principal component vectors (unsupervised bases) that characterize the distribution of mass spectrometry data as feature quantities, and creates a feature quantity extraction model. Generate.
- the target medium whose components are unknown is the past medium It is possible to estimate which medium has high similarity compared with .
- the determination unit 12 can set a threshold for abnormality determination in the state estimation space of the feature quantity extraction model based on a certain rule for abnormality determination.
- the threshold is simply defined as a quantity associated with the distribution of each mass spectrometry data of a plurality of culture media projected onto the state estimation space defined by the principal component vector.
- the threshold value for abnormality is It can be given as a quantity related to the Mahalanobis distance from ⁇ .
- a point group obtained by projecting mass spectrometry data is classified into a normal group and an abnormal group, and the mass spectrometry data of the normal group is projected onto a certain one-dimensional space to approximate the distribution to a normal distribution.
- ⁇ is the average value of the distribution of time
- ⁇ is the absolute value of the standard deviation of the normal distribution
- ⁇ 3 ⁇ can be set as the threshold value. That is, a threshold can be set so that when the mass spectrometry data of a certain culture medium is projected onto the one-dimensional space, and the coordinates are separated from ⁇ by 3 ⁇ or more, the culture medium is judged to be abnormal.
- the determination unit 12 uses the set threshold value and the coordinate x when the mass spectrometry data of the target culture medium obtained from the state estimation unit 10a is projected into the state estimation space to determine the presence or absence of an abnormality. conduct. As an example, if the distance between the coordinate x and the average value ⁇ of the distribution of the normal group is 3 or more in terms of the Mahalanobis distance of the distribution of the normal group, it is determined that there is an abnormality. Otherwise, it is determined that there is no abnormality.
- the degree of abnormality in the target culture medium may be determined by setting the above thresholds so as to be divided into a plurality of stages. As an example, if the distance between the coordinate x and the average value ⁇ of the distribution of the normal group is 2 or more and less than 3 in terms of the Mahalanobis distance of the distribution of the normal group, the degree of abnormality is small. , and if the number is 3 or more, it may be determined that the degree of abnormality is large.
- the judging unit 12 outputs the result of judging the presence or absence of abnormality or the degree of abnormality in the target culture medium.
- the state estimation system 1a in the state estimation system 1a, as shown in 10A of FIG. 10, at a point of time Sa2 during the culture process (S in FIG. 10 indicates the start time of the culture process, as in FIG. 2). , mass spectrometry is performed on a plurality of culture media by the measurement unit 2 to acquire mass spectrometry data.
- the mass spectrometry data at a certain time point Sa2 after a predetermined period of time has elapsed from the start of the culture is accumulated with respect to the plurality of media in which the culture is currently being cultured through the culture process.
- the culture is cultured until the end of the originally planned culture period, and when the culture in each medium is completed, the culture after the culture contained in the medium is collected. and measure the yield of the culture that has been cultured by the culture process (Fa1). In this way, the state of each culture after culturing is measured, and actual culture results are obtained for each medium.
- the state estimation system 1a when the yield (actual culture result) of the culture obtained for each medium is input by the worker or the like, the result of the culture state is stored in the database 4. At this time, the database 4 stores mass spectrometry data of each medium measured at a certain time point Sa2 during the culturing process in association with the final culture result obtained from the medium.
- the state estimating unit 10a reads from the database 4 the mass spectrometry data of each medium measured at a certain point Sa2 during the culture process and the final culture state obtained from the medium as sample data. to generate a feature extraction model.
- This feature quantity extraction model is input with mass spectrometry data of a target medium at a certain time point Sa2 (for example, 10 days after the start of culture) after starting the culture of the culture by a predetermined culture process, and cultured with the target medium.
- Sa2 for example, 10 days after the start of culture
- the output is an estimated value of the future culture state that is finally obtained when it is assumed that the culture of the product is continued as it is.
- the feature amount extraction model here, as in the above-described embodiment, for example, various feature amount extraction models such as PCA, PLS, polynomial regression, Gaussian process regression, Random Forest Regression, medium during culture
- PCA Principal Component Analysis
- PLS Principal Component Analysis
- polynomial regression Gaussian process regression
- Random Forest Regression medium during culture
- a feature value extraction model generated by machine learning of the characteristic values (mass spectrometry data) and the culture results thereof can be applied.
- the correspondence between the mass spectrometry data of multiple media and the culture results when using these media is learned, and the distribution of the mass spectrometry data is obtained from the culture results. should be characterized.
- the state estimating unit 10a uses the mass spectrometry data of the medium in the process of culturing the culture in the culture process as an explanatory variable, and the actual final culture state obtained in the culture process. Using (the yield of the culture contained in the medium at the end of the culture) as the objective variable, a feature extraction model is generated in advance that shows a characteristic relationship such as a correlation or an inverse correlation between the mass spectrometry data and the culture state.
- unsupervised learning may be applied.
- the state estimation system 1a After that, during the culture process using the unknown target medium, the culture result will be finally obtained from the target medium using the feature extraction model. Future culturing results can be estimated during the culturing process before ending the culturing process.
- a series of culture processes that induce cell culture differentiation and obtain the desired cells require a long time to obtain the desired cells.
- a series of culture processes for inducing differentiation of undifferentiated human stem cells to obtain desired cells usually takes about three months or longer.
- a conceivable method is to analyze the obtained actual culture results and determine the next parameters (components of the culture medium to be used, culture temperature, etc.) each time.
- the estimation result of the culture state that will be finally obtained by the target medium during the culture process is obtained based on the feature extraction model.
- various parameters of the culture process such as the components of the medium used and the culture temperature
- various parameters of the culture process can be obtained based on the obtained estimated results without waiting for the end of the culture process. can be changed as needed to allow a new culture process to take place.
- a new culturing process can be started in the middle of the current culturing process by trial and error with the parameters of the culturing process so as to obtain the desired culturing result. Parameter optimization can proceed efficiently.
- the above-mentioned "(1-2) state estimation unit” uses the mass spectrometry data of the medium before the start of the culture process. It can also be applied in this embodiment by replacing it with analysis data. Similarly, for the “(1-3) determination unit” and “(1-4) measurement object optimization estimation unit”, the mass spectrometry data of the medium before the start of the culture process is obtained during the culture process. It can also be applied in place of the mass spectrometry data of the culture medium obtained.
- the determination unit 12 determines the state of the target medium based on the state estimation result obtained from the mass spectrometry data of the medium during the culture process in the state estimation unit 10a.
- the measurement target optimization estimating unit 14 uses the state estimating unit 10a to determine the target medium that can obtain the desired culture state based on the state estimation result obtained from the mass spectrometry data of the medium during the culturing process. , to estimate additives to be added to the target medium (measurement target optimization estimation processing).
- mass spectrometry is performed on the medium by the measurement unit 2 only at certain points Sa2 and Sa3 during the culture process, and a plurality of mass spectrometry data are obtained. You may make it so that you can get it. That is, the mass spectrometry data of the medium on the 10th day from the start of the culture process and the mass spectrometry data of the medium on the 20th day from the start of the culture process were collected multiple times at intervals in time series from the start of the culture. Analytical data may be measured.
- the measuring unit 2 Perform mass spectrometric analysis of a plurality of culture media to obtain mass spectrometric data.
- the state estimation system 1a measures the culture state (yield and quality of the culture) after actually culturing in each medium (Fa1).
- the state estimating unit 10a uses the mass spectrometry data in the medium before the start of the culture process as an explanatory variable, and the mass spectrometry data in the medium after the culture process ends (after the culture is used). Using the culture state as an objective variable, a feature extraction model is generated that indicates the relationship between the mass spectrometry data and the culture state.
- the state estimating unit 10a obtains the mass spectrometry data measured before the start of the culture process for the target medium whose components are unknown
- the target medium is cultured from the mass spectrometry data based on the feature extraction model. Estimate the culture state (yield and quality of the culture) of the culture that will be obtained in the future when used in the process.
- the state estimating unit 10a also calculates each mass spectrometry data in the medium at a certain point Sa2 after the start of the culture process, and the culture state of the culture contained in the medium after the culture process is finished (after the culture is used). Similarly, each mass spectrometry data in the medium at Sa3 at a certain point after the start of the culture process, and the medium after the end of the culture process (after use for culture) Generate a feature extraction model that shows the relationship between the culture state of the culture contained in .
- the state estimating unit 10a uses the mass spectrometry data measured at each point in time (Sa2, Sa3) for a target medium in use for culture whose components are unknown, based on the feature amount extraction model at the corresponding point in time Sa2, Sa3. , respectively estimate the culture state (yield and quality of the culture) of the culture to be obtained in the future when the target medium is used in the culture process.
- the mass spectrometry is performed on the target medium by the measuring unit 2 at each time point Sa1 before starting the culture process and at each time point Sa2 and Sa3 during the culture process, and the mass of the target medium that changes over time is measured.
- the culture process can be executed while estimating the culture state of the culture after the culture process is completed, based on the feature quantity extraction models corresponding to certain time points Sa1, Sa2, and Sa3.
- each feature quantity extraction model is generated by supervised learning
- the present invention is not limited to this, and the above “(1-6) Generation of feature quantity extraction model without supervision” Accordingly, each feature quantity extraction model may be generated by unsupervised learning.
- the measurement unit 2 performs mass spectrometry on a plurality of culture media at a certain point Sa1 before starting the culture process to acquire mass spectrometry data.
- the state estimation system 1a measures the culture state (yield and quality of the culture) of the culture actually being cultured in each medium at a certain point Fa2 before the culture process ends.
- the state estimating unit 10a generates a feature quantity extraction model by supervised learning based on the thus obtained mass spectrometry data of the medium before starting the culture process and the culture state of the culture during the culture process. do. That is, the state estimating unit 10a uses the mass spectrometry data in the medium before the start of the culture process as an explanatory variable, and the culture state of the culture contained in the medium at a certain time point Fa2 before the end of the culture process as an objective variable.
- a feature extraction model is generated that indicates the relationship between data and culture conditions.
- the state estimating unit 10a obtains mass spectrometry data measured before the start of the culture process for a target medium whose components are unknown
- the target medium is cultured from the mass spectrometry data based on the feature extraction model. It is possible to estimate the culture state (yield and quality of the culture) during the culture that will be obtained in the future when used in the process.
- the above-mentioned "(1-2) state estimating unit” uses the measurement result of the culture state of the culture after the culture process is completed, but the measurement result of the culture state is obtained during the culture process. By replacing with the measurement result of the culture state of the culture during the culture, it can be applied to this "(1-8) another embodiment for estimating the culture state of the culture during the culture before the end of the culture process". can.
- the measurement result of the culture state of the culture after the end of the culture process is It can be applied in place of the measurement result of the culture state of the culture obtained along the way.
- the measurement unit 2 performs mass spectrometry on a plurality of culture media at a certain point Sa1 before starting the culture process to acquire mass spectrometry data.
- the present invention is not limited to this.
- the measurement unit 2 performs mass spectrometry on a plurality of culture media to acquire mass spectrometry data. You may do so.
- the state estimating unit 10a uses the mass spectrometry data in the medium at a point Sa2 during the culture process as an explanatory variable, and determines the culture state of the culture contained in the medium at a point Fa2 before the end of the culture process. is the objective variable, a feature extraction model that characterizes the relationship between the mass spectrometry data and the culture state is generated.
- the state estimating unit 10a obtains mass spectrometry data measured at a certain point Sa2 during the culture process of the target medium whose components are unknown, based on the feature amount extraction model from the mass spectrometry data , it is possible to estimate the culture state (yield and quality of the culture) at a certain point Fa2 during the culture that will be obtained in the future when the target medium is used in the culture process.
- mass spectrometry data obtained multiple times.
- mass spectrometry is performed on the medium by the measurement unit 2 only at certain points Sa2 and Sa3 during the culture process, and a plurality of mass spectrometry data are obtained. You may make it so that you can get it.
- the state estimating unit 10a uses the mass spectrometry data in the medium before the start of the culture process as an explanatory variable, and the culture state of the culture contained in the medium at a certain point Fa2 before the end of the culture process as an objective variable, and performs mass spectrometry.
- a feature extraction model is generated that characterizes the association between data and culture conditions. After that, when the state estimating unit 10a obtains the mass spectrometry data measured before the start of the culture process for the target medium whose components are unknown, the target medium is cultured from the mass spectrometry data based on the feature extraction model. It is possible to estimate the cultivation state (yield and quality of the culture) of the culture at some point Fa2 during the cultivation, which will be obtained in the future when used in the process.
- the state estimating unit 10a calculates the mass spectrometry data in the medium at a point Sa2 after the start of the culture process and the culture state of the culture contained in the medium at a point Fa2 before the end of the culture process. Similarly, each mass spectrometry data in the medium at a time point Sa3 after the start of the culture process and the medium at a time point Fa2 before the end of the culture process.
- a feature extraction model is generated that characterizes the relationship between the culture state of the involved cultures.
- the state estimating unit 10a acquires mass spectrometry data measured at certain points in time (Sa2, Sa3) for a target medium in use for culture whose components are unknown, and obtains the corresponding points in time Sa2, Sa3 from the mass spectrometry data. Based on the feature value extraction model of Sa3, the culture state (yield and quality of the culture) at a certain point Fa2 during the culture, which will be obtained in the future when the target medium is used in the culture process. can be estimated.
- FIG. 12 is a block diagram showing the configuration of a state estimation system 1c according to another embodiment.
- the state estimation system 1c differs from the state estimation system 1a of FIG. 1 in the configuration of the arithmetic processing unit 3c. 4. Since the description of the notification unit 5 is duplicated, it is omitted.
- a situation may arise in which culture conditions are examined to bring the cells to a specific culture state.
- culture conditions For example, in the continuous culture of CHO cells, after the cells are added to the initial medium, the cells are grown to a certain number. It is necessary to examine the optimal culture conditions for proliferation. At this time, trial and error are performed to determine the parameters (culture parameters) related to the culture conditions of the culture process so that the desired culture state can be obtained. need. Therefore, it is desirable to optimize the culture parameters efficiently in terms of time.
- the state estimation system 1c of FIG. 12 is configured such that the desired culture state is determined with respect to parameters (culture parameters) related to culture conditions such as culture temperature, culture time, medium pH, and medium agitation speed of the culture process.
- the expected culture parameters can be estimated by the arithmetic processing unit 3c, and the estimation result can be presented to the operator or the like via the notification unit 5.
- FIG. Thereby, the culturing process can be performed with the optimum culturing parameters for obtaining the desired culturing conditions.
- the arithmetic processing unit 3c has a state estimating unit 10c, a determining unit 12, and a culture parameter optimization estimating unit 15.
- the state estimation unit 10c needs to generate a feature quantity extraction model for estimating the culture parameters, which will be described below.
- a set of culture parameters that serve as standards such as culture temperature, culture time, medium pH, and medium agitation speed when culturing a culture. decide.
- a plurality of correction candidate values ⁇ a of the culture parameter are determined from the culture parameter that serves as the reference.
- the candidate correction value ⁇ a for the culture temperature is, for example, +2° C. or ⁇ 1° C.
- the candidate correction value ⁇ a for the culture time is, for example, +30 minutes or ⁇ 30 minutes.
- the correction candidate value ⁇ a for the medium pH is +0.1, ⁇ 0.2, etc.
- the correction candidate value ⁇ a for the stirring speed of the medium is, for example, +1s ⁇ 1 , ⁇ 1s ⁇ 1 , etc. is.
- each of the 20 medium cultures is divided into four and transferred to separate culture vessels.
- the culture is continued with the culture parameters (a+ ⁇ a0), (a+ ⁇ a1), (a+ ⁇ a2), and (a+ ⁇ a3).
- ⁇ a1, ⁇ a2, and ⁇ a3 indicate different correction candidate values for the culture parameter. show.
- the mass spectrometric data of the medium, the reference culture parameter a, the correction candidate values ⁇ a0, ⁇ a1, ⁇ a2, and ⁇ a3 of the culture parameters, and the culture parameters (a+ ⁇ a0), (a+ ⁇ a1), (a+.DELTA.a2) and (a+.DELTA.a3) are associated with actual results indicating the culture state of the culture when the culture was actually cultured, and these are stored in the database 4 as data sets, respectively.
- the state estimating unit 10c reads out these data sets stored in the database 4, mass spectrometry data, the reference culture parameter a before correction, and correction candidate values ⁇ a0 and ⁇ a1 of the culture parameter a. , ⁇ a2, ⁇ a3 as input data, a feature extraction model for estimating culture parameters that characterizes the distribution of the input data is generated. Specifically, for the culture 3 hours after the start of the culture, feature extraction for estimating the culture parameters to estimate what kind of correction should be made to the standard culture parameters to obtain the desired culture state. Generate a model.
- mass spectrometry data, the reference culture parameter a before correction, and the correction candidate values ( ⁇ a0, ⁇ a1, ⁇ a2, ⁇ a3) of the culture parameter a are used as explanatory variables, and the culture A feature quantity extraction model for parameter estimation is generated, but the present invention is not limited to this, mass spectrometry data and culture parameters (a + ⁇ a0), (a + ⁇ a1), (a + ⁇ a2), (a + ⁇ a3) may be used as explanatory variables to generate a feature quantity extraction model for estimating culture parameters.
- the feature quantity extraction model for estimating the culture parameters is, as described in the above "(1-1) Configuration of the state estimation system", for example, PCA, PLS, polynomial regression, Gaussian process regression, convolutional neural It is the same that various feature amount extraction models such as networks and Bayesian neural networks can be applied.
- the state estimating unit 10c uses the mass spectrometric data of the culture medium obtained by the measuring unit 2, the reference culture parameter a before correction, and the correction candidate values ⁇ a0, ⁇ a1, ⁇ a1, ⁇ a1 is input (explanatory variable), and the culture state (here, yield) of the culture contained in the medium after culture use is output (objective variable). Generate.
- a feature extraction model for estimating culture parameters is generated, which characterizes the distribution of the input data from the culture state of the culture cultured in the culture process.
- the state estimating unit 10c uses the mass spectrometry data of the target medium obtained through the measuring unit 2, the culture parameter a, and the plurality of correction candidate values ⁇ a of the culture parameter as input data for estimating the culture parameter described above.
- the culture state (yield) of the culture contained in the medium used for culture is estimated by inputting to the feature quantity extraction model and analyzing.
- the state estimating unit 10 c outputs a plurality of estimation results respectively obtained by changing the correction candidate value ⁇ a to the culture parameter optimization estimating unit 15 .
- the culture parameter optimization estimation unit 15 selects an estimation result that is closest to the preset optimal culture state result from among the plurality of estimation results obtained by the state estimation unit 10c, and sets with the selected estimation result.
- a correction candidate value ⁇ a for the culture parameter a that is set is specified.
- the culture parameter optimization estimation unit 15 can output the identified correction candidate value ⁇ a to the notification unit 5 and present the correction candidate value ⁇ a to the worker or the like via the notification unit 5 . In this way, by presenting the correction candidate value ⁇ a obtained by the culture parameter optimization estimation unit 15 to the worker or the like via the notification unit 5, the worker or the like can perform cultivation based on the correction candidate value ⁇ a. Parameters can be corrected and the culture can be cultivated under optimal culture conditions (culture parameters).
- the state estimation system 1c uses as input data a plurality of medium characteristic values and a plurality of culture parameters measured by the measurement unit 2, and the distribution of this input data is a feature.
- the state estimating unit 10c estimates the culture state of the culture from the characteristic values and the culture parameters of the target medium based on the feature quantity extraction model for estimating the culture parameters. In this way, the state estimation system 1c can estimate in advance whether or not the desired culture state can be obtained in bioproduction and research and development using the target medium, so the time required to determine the appropriateness of the target medium can be shortened. As a result, the efficiency of bioproduction and research and development using the medium can be enhanced.
- the state estimation system 1c specifies the correction candidate value ⁇ a of the culture parameter a that provides the optimum culture state based on the estimation result of the culture state obtained by the state estimation unit 10c. do.
- the operator or the like can correct the culture parameter based on the obtained correction candidate value ⁇ a of the culture parameter, and the optimum culture state can be obtained.
- the estimation result obtained by the state estimation unit 10c may be output to the determination unit 12, and the determination unit 12 may determine the culture parameter based on the estimation result.
- the culture parameter is determined based on the estimation result by the state determination process or the abnormality determination process described in "(1-3) Determination section" above.
- the determination unit 12 when performing the state determination process, stores a good estimation result and A discriminant value that can be discriminated step by step from a bad estimation result is set.
- the determination unit 12 selects an identification value close to the value of the state estimation vector obtained by projecting the mass spectrometry data and the culture parameters of the target medium into the state estimation space, identifies the estimation result from the selected identification value, and cultures it. Judge the appropriateness of the parameters.
- the determination unit 12 may set a threshold for the abnormality determination in the state estimation space of the feature quantity extraction model for estimating the culture parameter based on a rule for the abnormality determination. good.
- the determination unit 12 identifies the estimation result from the value of the state estimation vector obtained by projecting the mass spectrometry data and the culture parameter of the target medium onto the state estimation space defining the principal component vector, and the threshold, and cultures. Determine whether there is an abnormality in the parameters. Further, the degree of abnormalities in the culture parameters may be determined by setting the above thresholds so as to be divided into a plurality of stages.
- the feature extraction model uses the mass spectrometry data of the target medium as input, and assumes that the culture is continued as it is with the target medium. It is desirable to apply a regression model or the like that outputs a time estimation result (hereinafter also referred to as a time estimation result).
- a time estimation result hereinafter also referred to as a time estimation result.
- the correspondence relationship between the mass spectrometry data of multiple media and the culture time at which the culture reaches the desired culture state when using these media is determined. It is desirable to learn and characterize the distribution of mass spectrometry data from incubation times.
- a correct label for example, a correct answer rate indicating whether or not it is the desired culture time
- the correct label is attached when learning the mass spectrometry data of the medium and the culture time. , may characterize the distribution of the mass spectrometry data.
- Various feature extraction models such as PCA, PLS, polynomial regression, Gaussian process regression, and Random Forest Regression can be applied as feature extraction models.
- a machine learning model generated by machine learning the result of the culture medium characteristic value (mass spectrometry data) and culture time may be applied.
- a neural network eg, CNN (convolutional neural network) or BNN (Bayesian neural network)
- unsupervised learning may also be used for the feature quantity extraction model. The details of generating a feature quantity extraction model by unsupervised learning will be described later, but when generating a feature quantity extraction model by unsupervised learning, mass spectrometry data of a plurality of culture media are used. It is desirable to allow learning to characterize the distribution of data.
- the determination unit 12 determines the state of the target medium based on the estimation result (time estimation result) of the culture time required for the culture to reach the desired culture state obtained by the state estimation unit 10a.
- the determination unit 12 sets, in a state estimation space in which a principal component vector (feature amount) is defined in the feature amount extraction model, an identification value capable of distinguishing between a good medium and a bad medium in stages, and sets a target medium is projected onto the state estimation space, and using the identification value as a guide, it is estimated whether the target culture medium is in a state where the desired culture time can be achieved (state determination processing).
- the culture time required until the mass of the culture per unit volume of the medium obtained when culturing the culture using the medium reaches a predetermined optimum set value (good judgment value) or more is set in advance.
- a medium in which the culturing time is within an arbitrary period of time is called a "good medium”
- a medium in which the culture time exceeds the arbitrary period of time is called a “poor medium”.
- the determination unit 12 may set a threshold value for abnormality determination in the state estimation space of the feature extraction model based on a rule for abnormality determination.
- the threshold may be set in the state estimation space by an operator or the like in consideration of the desired incubation time. For example, if the threshold is not exceeded, it can be determined that the culture time is short and a good culture time can be obtained. A threshold setting is made to obtain
- the determination unit 12 determines the presence or absence of an abnormality in the target medium from the viewpoint of incubation time from the value of the state estimation vector obtained by projecting the mass spectrometry data of the target medium onto the state estimation space defining the principal component vector and the threshold value. judge. Further, the degree of abnormality in the target culture medium may be determined by setting the above thresholds so as to be divided into a plurality of stages.
- the state estimating unit 10a uses the mass spectrometry data of the target medium before the start of the culturing process to estimate the culture time of the culture based on the feature extraction model (time estimation result ) is obtained, the time estimation result is received from the state estimation unit 10a.
- the measurement target optimization estimating unit 14 analyzes the time estimation result obtained by the state estimating unit 10a, and by adding it to the target medium before culture use, for example, an additive that can shorten the culture time of the culture to estimate
- the measurement target optimization estimation unit 14 corrects the components of the target medium used in the culture process by having the operator or the like add the estimated additive (also referred to as an estimated additive) to the target medium. be.
- the measurement target optimization estimating unit 14 in accordance with the above-described “(1-4) measurement target optimization estimating unit”, the additive measurement vector, which is mass spectrometry data regarding a plurality of types of additives, and the feature extraction model Calculate the relevance (similarity) with the principal component vector defined in the state estimation space.
- the measurement target optimization estimating unit 14 estimates the additive of the additive measurement vector that has the greatest relationship with the base of the feature extraction model from among the three types of additives as the optimal additive to be added to the target medium. do. In this way, the culture time can be improved by adding the additive of the additive measurement vector having the greatest relevance to the base of the feature extraction model to the target medium before being used for culture.
- the additive of the additive measurement vector highly related to the basis of the feature quantity extraction model has an additive measurement vector highly similar to the principal component vector defined in the state estimation space in the feature quantity extraction model. It is an additive.
- the degree of similarity between the principal component vector of the feature quantity extraction model and the additive measurement vector may be obtained by calculating the inner product between the principal component vector of the feature quantity extraction model and the additive measurement vector.
- the principal component vector with the largest eigenvalue is selected, and the inner product between the selected principal component vector and the additive measurement vector is calculated.
- the similarity between the principal component vector defined in the state estimation space in the feature quantity extraction model and the multiple types of additive measurement vectors is calculated, and the similarity with the principal component vector is calculated.
- an additive with a high additive measurement vector is specified by the measurement target optimization estimation unit 14
- the present invention is not limited to this.
- information on one or more additives is associated in advance with the estimation result of the culture time that will be obtained by the state estimation unit 10a, and based on the feature amount extraction model, from the mass spectrometry data of the target medium
- one or more additives previously associated with the result of estimation may be estimated as the optimum additive to be added to the target medium.
- Optimal culture parameters may be estimated based on the result of estimating the culture time according to "(1-9) Culture parameter optimization estimation process".
- the mass spectrometry data of the medium, the reference culture parameter a, and the correction candidate value ⁇ a0 of the culture parameter , ⁇ a1, ⁇ a2, ⁇ a3, and the culture parameters (a + ⁇ a0), (a + ⁇ a1), (a + ⁇ a2), (a + ⁇ a3), the desired culture is are associated with actual results indicating the culture time required to reach the culture state of , and stored in the database 4 as a data set.
- the state estimating unit 10c reads out these data sets stored in the database 4, mass spectrometry data, the reference culture parameter a before correction, and correction candidate values ⁇ a0, ⁇ a1, and ⁇ a2 for the culture parameter a. , ⁇ a3 as input data, a feature extraction model for estimating culture parameters that characterizes the distribution of the input data is generated. Specifically, a feature quantity extraction model for estimating culture parameters for estimating what kind of correction should be added to the standard culture parameters for the culture 3 hours after the start of culture to achieve the desired culture time. to generate
- the culture parameter optimization estimation unit 15 selects, for example, the estimation result with the shortest culture time from among the plurality of estimation results obtained by the state estimation unit 10c, and the culture parameter a set by the selected estimation result.
- a correction candidate value ⁇ a is specified. In this way, by presenting the correction candidate value ⁇ a obtained by the culture parameter optimization estimation unit 15 to the worker or the like via the notification unit 5, the worker or the like can perform cultivation based on the correction candidate value ⁇ a. Parameters can be corrected and incubation times can be shortened.
- the time estimation result obtained by the state estimation unit 10c may be output to the determination unit 12, and the determination unit 12 may determine the culture parameter based on the time estimation result.
- the culture parameter is determined based on the time estimation result by the state determination process or the abnormality determination process described in "(1-3) Determination section" above. Since the state determination process or the abnormality determination process has been described in the above "(1-3) Determination section", description thereof will be omitted to avoid duplication of description.
- the measurement unit 2 performs mass spectrometry on a plurality of cultures and at least one target culture at a certain point in time, Measure analytical data.
- the state estimating unit 10a estimates the future culture state of the target culture from the characteristic values of the target culture based on a feature extraction model that characterizes the distribution of mass spectrometry data, which are the characteristic values of a plurality of cultures. do.
- the characteristic values of the culture, the medium extract, or the culture extract may be used instead of the characteristic values of the culture medium.
- the medium extract extracted from the medium includes, for example, the substances contained in the original medium raw materials, as well as the target substances and by-products generated by the metabolism of the culture, including the metabolism of microorganisms and cells. substances (e.g., antibodies, proteins, carbohydrates, amino acids, lipids), and substances obtained by denaturation (e.g., Maillard reaction) of substances contained in the original medium raw materials over time.
- the measurement unit 2 performs mass spectrometry on a plurality of culture medium extracts and at least one target culture medium extract at a certain point in time to obtain mass spectrometry data. to measure.
- the state estimating unit 10a estimates the future culture state of the culture from the characteristic values of the target medium extract based on the feature amount extraction model that characterizes the distribution of the mass spectrometry data, which are the characteristic values of the plurality of medium extracts. presume.
- the culture extract extracted from the culture is, for example, a substance that constitutes the inside of the culture or itself, and more specifically, the metabolites ( For example, antibodies, proteins, carbohydrates, amino acids, lipids, etc.) and substances (eg, heavy metals) that cultures take in and accumulate in the body originally contained in medium raw materials.
- the measurement unit 2 performs mass spectrometry on a plurality of culture extracts and at least one target culture extract at a certain point in time to obtain mass spectrometry data. to measure.
- the state estimating unit 10a estimates the future culture state of the culture from the characteristic values of the target culture extract based on a feature extraction model that characterizes the distribution of the mass spectrometry data, which are the characteristic values of the plurality of culture extracts. presume.
- the measurement unit that measures the properties of a plurality of biological samples and a target biological sample measures, for example, mass spectrometry data of a plurality of biological samples and a target biological sample.
- mass spectrometry data measures mass spectrometry data of a plurality of biological samples and a target biological sample.
- a case of applying a mass spectrometer has been described, but in the second embodiment, a case of applying a microscope apparatus for measuring the appearance of a plurality of biological samples and a target biological sample will be described below.
- the state estimating system 1a shown in FIG. Apply a microscopy device (e.g., fluorescence microscopy device).
- the measurement unit 2 which is a microscope device, captures an image of the culture in order to measure the appearance of the culture in the medium as a characteristic of the culture, and obtains image data of the culture.
- the biological sample is a culture, and as the culture, for example, cells used in cell culture experiments will be described as an example.
- undifferentiated cells are cultured using a medium by a predetermined culture process to obtain cultured cells.
- the differentiation state of the cells can be identified by staining the cultured cells, for example, by immunostaining and observing the state of expression of light emitted from the cells with a microscope.
- the microscope device it is desirable to use a fluorescence microscope device capable of imaging the fluorescence emitted from the culture, which is the object of measurement.
- the differentiation state of actually cultured cells can be identified by, for example, analyzing metabolites from the cells using a mass spectrometer. By measuring and analyzing the expression state of genes in cultured cells, the differentiation state of the cells can be estimated.
- a state estimation system 1a according to the second embodiment image data is obtained by imaging an undifferentiated target cell with the measurement unit 2, and this image data is used as an input for the feature amount extraction model, and the image data of the target cell Based on the feature extraction model, the future culture state (differentiation state) of the undifferentiated target cells is estimated before the differentiation state.
- a state estimation system 1a according to the second embodiment includes a measurement unit 2, which is a microscope device, an arithmetic processing unit 3a, a database 4, and a notification unit 5, and mainly includes the above-described first embodiment. The difference is that the “image data” of the culture acquired by the measurement unit 2 is used instead of the “mass spectrometry data” of the medium described in .
- a plurality of undifferentiated cells are imaged by the measurement unit 2, and a feature quantity extraction model is generated based on the obtained plurality of image data.
- the state estimation system 1a captures an image of an undifferentiated target cell whose state is unknown by the measurement unit 2, and based on the resulting image data and the feature amount extraction model, the target cell in the future. It is possible to estimate the differentiation state (culture state) when cultured systematically. In this way, a biological sample whose culture state is to be estimated using a feature quantity extraction model generated in advance is referred to as a “target biological sample”.
- a biological sample whose culture state is to be estimated using a feature quantity extraction model generated in advance is referred to as a “target biological sample”.
- a case where a cell (culture) is applied as an example of a biological sample will be described below. or "subject cells”.
- the measurement unit 2 captures an image of each undifferentiated cell contained in each of the plurality of culture media, and outputs the obtained plurality of image data to the arithmetic processing unit 3a and the database 4.
- the database 4 associates each image data with the date and time when the image data was obtained and stores them.
- information on the features of the captured cells includes, for example, cell type (specification information), model number, lot number, date of manufacture, date of arrival, past usage history, etc.
- parameters related to culture conditions i.e., culture parameters
- culture parameters such as culture temperature, culture time, medium pH, medium agitation speed, etc. in the culture process is also stored in association with the image data. .
- the cells imaged by the measurement unit 2 are actually cultured using a medium according to a predetermined culture process. Then, for example, the differentiated state of the actually cultured cells is confirmed based on the image data obtained by staining the cultured cells with a staining substance and imaging them with a microscope apparatus.
- image data of a plurality of undifferentiated cells obtained by the measuring unit 2 at a certain point in time, and image data of the cells when the cells are actually cultured using a culture medium are actually cultured using a culture medium.
- the differentiation state that is, the result of the culture state (culture result)
- multivariate analysis is performed to generate a feature extraction model.
- a feature quantity extraction model it is possible to analyze the appearance features of cells that may affect the culture results when the cells are cultured using a medium.
- the feature quantity extraction model uses image data of target cells as input, and estimates the future culture state (state It is desirable to apply a regression model or the like that outputs the estimation result).
- the correspondence between the image data of multiple cells and the culture state (differentiation state) of these cells is learned, and the cells of the image data are learned from the culture state. It is desirable to characterize the distribution of the appearance features of
- supervised learning of other feature quantity extraction models for example, image data of a plurality of cells, the culture state of each cell, and a correct label indicating whether or not the culture state is the desired result (for example, , Accuracy rate indicating whether or not the culture is in the desired state), and when learning the image data of the cell and its culture state, the correct label is attached and learned, and the appearance characteristics of the cell in the image data may be used to characterize the distribution of
- the feature quantity extraction model according to the second embodiment is similar to the first embodiment, for example, principal component analysis (PCA), partial least squares (PLS), polynomial regression, Various feature extraction models such as Gaussian process regression and Random Forest Regression can be applied. Also, as the feature quantity extraction model, for example, a machine learning model generated by performing machine learning on the characteristic value (image data) of the cell and the culture result of the cell may be applied.
- PCA principal component analysis
- PLS partial least squares
- Various feature extraction models such as Gaussian process regression and Random Forest Regression
- a machine learning model generated by performing machine learning on the characteristic value (image data) of the cell and the culture result of the cell may be applied.
- a neural network eg, CNN (convolutional neural network) or BNN (Bayesian neural network)
- CNN convolutional neural network
- BNN Bayesian neural network
- the differentiation state of each actually cultured cell is used as an index as an objective variable.
- the characteristic state itself of each cell in the image data obtained by staining each actually cultured cell with a staining substance and imaging it with the measurement unit 2, which is a microscope device i.e., the actually cultured cell
- the differentiation state of cells may not be indicated by an "index", but the image data itself obtained by imaging the cells) may be applied as an objective variable.
- the feature extraction model may be unsupervised learning in addition to supervised learning as described above.
- the determining unit 12 determines the state of the target cell based on the state estimation result obtained by the state estimating unit 10a.
- the determination processing in the determination unit 12 there is a state determination processing that simply determines whether the cultured target cells are in a suitable culture state in a stepwise manner, and an abnormal culture state such that the cultured target cells are not in a desired differentiation state. and an abnormality determination process for determining whether or not the
- the determination unit 12 defines, for example, a principal component vector (supervised basis) as a feature amount in the feature amount extraction model prior to the state determination process. Discrimination values capable of discriminating between good cells and bad cells in stages are set in the state estimation space. In addition, this identification value may be associated with an identification display that expresses the degree of goodness of the differentiation state of the cells in stages, based on the results of culturing the cells using the medium, for example.
- the setting of the identification value for the state estimation space may be arbitrarily set by the operator or the like in the same manner as in the first embodiment described above.
- the identification values may be set automatically by classifying them according to the order.
- the determination unit 12 determines the state of the target cell from the value of the state estimation vector obtained by projecting the image data of the target cell onto the state estimation space and the identification value.
- the determination unit 12 may set a threshold value for abnormality determination in the state estimation space of the feature extraction model based on a rule for abnormality determination.
- the threshold may be set in the state estimation space by an operator or the like in consideration of the desired culture result. For example, the threshold is set such that if the threshold is not exceeded, a good culture result equal to or greater than the reference value is obtained, and if the threshold is exceeded, a poor culture result below the reference value is obtained.
- the determination unit 12 incorporates the image data of the target cell into the feature amount extraction model, and the value of the state estimation vector obtained by projecting the feature amount extracted from the image data onto the state estimation space defining the principal component vector; Based on the threshold value, the presence or absence of abnormality in the target cell is determined. Further, the degree of abnormality of the target cell may be determined by setting the above thresholds so as to be divided into a plurality of stages.
- the notification unit 5 receives the determination result of the target cell by the state determination process, and determines the state of the target cell before use in culture based on the identification value. Notify workers, etc. In addition, the notification unit 5 notifies the operator or the like of the presence or absence of abnormality or the degree of abnormality by receiving the determination result of the presence or absence of abnormality or the degree of abnormality by the abnormality determination process.
- the state estimation system 1a includes a measurement unit 2 that measures image data of a plurality of cultures and at least one target culture as characteristic values (input data), and a plurality of cultures
- a state estimating unit 10a for estimating the culture state of the target culture from the characteristic values of the target culture based on a feature extraction model that characterizes the distribution of the property values (input data) of the object is provided.
- the measurement unit 2 obtains image data of a plurality of cells, and using the obtained plurality of image data, for example, multivariate analysis is performed, and a feature extraction model that can clearly detect differences between cells to build.
- the state estimation unit 10a upon receiving image data from the measurement unit 2 or the database 4, the state estimation unit 10a performs PCA processing on the image data to obtain principal component vectors (unsupervised bases) that characterize the distribution of the image data. ) is extracted as a feature quantity to generate a feature quantity extraction model.
- the state estimation system 1a when the image data of the unknown target cell is subsequently obtained, based on the image data of the target cell and the feature extraction model, which cell is the unknown target cell compared with the past cells? It can be estimated whether the similarity is high with
- the determination unit 12 can set a threshold value for abnormality determination in the state estimation space of the feature quantity extraction model based on a certain rule for abnormality determination.
- the threshold is simply defined as a quantity related to the distribution when each image data of a plurality of cells is projected onto the state estimation space defining the principal component vector.
- the determination unit 12 can determine the presence or absence of an abnormality using the threshold set in the state estimation space and the coordinate x when the image data of the target cell is projected into the state estimation space. Further, the degree of abnormality of the target cell may be determined by setting the above thresholds so as to be divided into a plurality of stages. The determination unit 12 outputs the result of determining the presence or absence of an abnormality in the analysis or the degree of the abnormality to the notification unit 5 . Thereby, the notification unit 5 can present the determination result of the determination unit 12 to the worker or the like.
- the present invention is not limited to this. Also in the second embodiment, for example, as shown in 10A of FIG. 10, after starting the culturing process, the cells being cultured in the culturing process are imaged by the measurement unit 2 to obtain image data. good too.
- the measurement unit 2 images a plurality of cells to acquire image data of the cells. .
- image data at a certain time point Sa2 after a predetermined period of time has elapsed from the start of culturing is accumulated for a plurality of cells currently being cultured through the culturing process.
- the cells in the culture are cultured until the originally scheduled culture termination time, and when the cell culture is completed, the cells contained in the medium after the culture termination are stained, and the cells emit Light is measured to confirm the differentiation state of the cell from the expression state of light emitted from the cell (Fa1). In this way, the state of each cell after culture is measured, and actual culture results are obtained for each cell.
- the state estimation system 1a when an index (actual culture result) indicating the cell differentiation state obtained for each cell is input by an operator or the like, the result of the culture state is stored in the database 4. At this time, the image data of each cell measured at a certain point Sa2 during the culturing process is stored in the database 4 in association with the final culture result obtained when culturing the cell.
- the image data of each cell measured at a certain point Sa2 during the culture process and the final culture state of the cell are read from the database 4 as sample data, and feature values are Generate an extraction model.
- This feature quantity extraction model is input with image data of the target cell at a certain time point Sa2 (for example, 10 days after the start of culture) after the start of culturing the target cell by a predetermined culture process, and culturing the target cell.
- Sa2 for example, 10 days after the start of culture
- the estimated value of the future target cell culture state that is finally obtained assuming that the process continues as it is is output.
- the feature amount extraction model here, as in the above-described embodiment, for example, various feature amount extraction models such as PCA, PLS, polynomial regression, Gaussian process regression, Random Forest Regression, etc., cells during culture
- a feature value extraction model or the like generated by machine learning of the characteristic values (image data) and the culture results thereof can be applied.
- generating a feature quantity extraction model for supervised learning it is possible to learn the correspondence relationship between the image data of multiple cells and the culture results of these cells, and to characterize the distribution of the image data from the culture results. desirable.
- unsupervised learning may be applied.
- the state estimating unit 10a uses the image data of the cells being cultured in the culture process as explanatory variables, and the actual final culture state obtained in the culture process (The differentiation state of the cell at the end of culture) is used as an objective variable, and a feature quantity extraction model is generated in advance that shows a characteristic relationship such as a correlation or an inverse correlation between the image data and the culture state.
- the feature extraction model is used to determine the future of the target cells that will be finally obtained by the culture. It becomes possible to estimate a realistic culture result during the culture before finishing the culture process.
- a long time is required to obtain the desired cells.
- a series of culture processes for inducing differentiation of undifferentiated human stem cells to obtain desired cells usually takes about three months or longer.
- a conceivable method is to analyze the obtained actual culture results and determine the next parameters (components of the culture medium to be used, culture temperature, etc.) each time.
- the estimation result of the culture state that will be finally obtained during the culture process can be obtained.
- various parameters of the culture process such as medium components and culture temperature
- the culture (target cells) can be changed, or a new culture process can be performed. In this way, without waiting for the end of the culture process, a new culture process is started in the middle of the current culture process by trial and error with the parameters and the culture so as to obtain the desired culture result. Therefore, it is possible to proceed with parameter optimization in a time efficient manner.
- the determination unit 12 determines the state of the target cell based on the state estimation result obtained from the image data of the cell during the culture process in the state estimation unit 10a.
- the culture process is performed.
- the cells may be imaged by the measurement unit 2 at a certain point Sa1 before the start and at certain points Sa2 and Sa3 during the culture process, and image data may be obtained a plurality of times.
- the cells are not imaged by the measurement unit 2 before the culture process is started, and the cells are imaged by the measurement unit 2 only at certain points Sa2 and Sa3 during the culture process, and the image data of the cells are obtained a plurality of times. You may do so. That is, the image data of the cells on the 10th day after the start of the culture process and the image data of the cells on the 20th day after the start of the culture process are collected multiple times at intervals in time series from the start of the culture process. You can get it.
- the measuring unit 2 to image a plurality of cells and acquire image data. Further, after the culture process is completed, the state estimation system 1a measures the culture state (differentiation state of the culture) of each target cell actually cultured in the medium (Fa1).
- the image data of the cell at time point Sa1 before the start of the culture process and the image data of the cells at time points Sa2 and Sa3 after the start of the culture process thus obtained are combined into Based on this, each feature extraction model is generated separately. That is, for example, in the case of supervised learning, the state estimating unit 10a uses the image data of the cells before the start of the culture process as an explanatory variable, and the differentiation state of the cells after the end of the culture process as the objective variable. Generate a feature extraction model that shows the relationship between states.
- the state estimation unit 10a obtains image data of the target cell in the undifferentiated state before starting the culture process
- the target cell is cultured by the culture process based on the feature amount extraction model from this image. Estimate the culture state of the target cell (differentiation state of the target cell) to be obtained in the future.
- the state estimating unit 10a extracts a feature quantity indicating the relationship between the image data of the cell at a certain point Sa2 after the start of the culture process and the culture state of the cell after the end of the culture process. Generate a model, and similarly generate a feature extraction model showing the relationship between the image data of the cell at Sa3 at a certain point after the start of the culture process and the culture state of the cell after the end of the culture process. .
- the state estimating unit 10a extracts the image data obtained at certain points in time (Sa2, Sa3) from the target cells in the undifferentiated state in culture, based on the feature amount extraction model at the corresponding points in time Sa2, Sa3, The culture state of the target cell (differentiation state of the target cell) that will be obtained in the future when the target cell is cultured by the culture process is estimated.
- the target cell is imaged by the measurement unit 2 at a certain time point Sa1 before starting the culture process and at certain time points Sa2 and Sa3 during the culture process, and the image data of the target cell changes over time.
- the culturing process can be performed while estimating the culturing state of the target cell after the culturing process.
- each feature quantity extraction model is generated by supervised learning
- the present invention is not limited to this, and the above “(2-2) Generation of feature quantity extraction model without supervision” Accordingly, each feature quantity extraction model may be generated by unsupervised learning.
- the measurement unit 2 captures images of a plurality of cells at a certain point Sa1 before starting the culture process, and acquires image data.
- the state estimation system 1a measures the culture state of the cells actually being cultured (differentiation state of the cells at that point) at a certain point Fa2 before the end of the culture process.
- the state estimating unit 10a generates a feature extraction model by supervised learning based on the image data of the undifferentiated cells before the start of the culture process and the culture state of the cells during the culture thus obtained. Generate. That is, the state estimating unit 10a uses the image data of the cells before the start of the culture process as an explanatory variable and the culture state of the cells at a certain time point Fa2 before the end of the culture process as the objective variable, and compares the image data and the culture state. Generate a feature extraction model that characterizes the relationships between After that, when image data is obtained for the undifferentiated target cell before the start of the culture process, the state estimation unit 10a performs culturing by the culture process based on the feature amount extraction model from the image data. It is possible to estimate the culture state of the target cell (differentiation state of the target cell) in the middle of the culture, which is obtained in .
- the measurement unit 2 images a plurality of undifferentiated cells at a certain time point Sa1 before starting the culture process and acquires image data.
- the present invention is not limited to this.
- the measurement unit 2 images a plurality of cells and acquires image data.
- the state estimating unit 10a the image data of the cell at a certain time point Sa2 during the culture process is used as an explanatory variable, and the culture state of the cell at a certain time point Fa2 before the end of the culture process is used as an objective variable, A feature extraction model is generated that characterizes the relationship between image data and culture conditions. After that, when the state estimating unit 10a obtains image data obtained by imaging the target cell at a certain time point Sa2 during the culture process, the state estimating unit 10a obtains in the future from the image data based on the feature amount extraction model , the culture state of the target cell (differentiation state of the target cell) at a certain point Fa2 during the culture can be estimated.
- image data may be obtained a plurality of times.
- the cells are not imaged by the measurement unit 2 before the start of the culture process, but the cells are imaged by the measurement unit 2 only at certain times Sa2 and Sa3 during the culture process, and image data are obtained multiple times. good too.
- the state estimating unit 10a uses the image data of the cells before the start of the culture process as an explanatory variable, and the culture state of the cells contained in the medium at a certain point Fa2 before the end of the culture process as the objective variable, and uses the image data A feature extraction model is generated that characterizes the relationship between and the culture state. After that, when the state estimating unit 10a obtains the image data of the undifferentiated target cell before the start of the culture process, the state estimating unit 10a, based on the feature amount extraction model, from this image data, The culture state of the target cell (differentiation state of the target cell) at the time point Fa2 can be estimated.
- the state estimating unit 10a also determines the relationship between the image data of the cells at a point in time Sa2 after the start of the culture process and the culture state of the cells contained in the medium at a point in time Fa2 before the end of the culture process. Similarly, the image data of the cells at Sa3 after the start of the culture process and the culture of the cells contained in the medium at Fa2 before the end of the culture process. Generate a feature extraction model that characterizes the relationships between states.
- the state estimating unit 10a obtains image data at certain time points (Sa2, Sa3) for the target cells in culture, and based on the feature amount extraction model at the corresponding certain time points Sa2, Sa3 from the image data , the culture state of the target cell (differentiation state of the target cell) at a certain time point Fa2 during the culture, which will be obtained in the future, can be estimated.
- the state estimation unit 10c inputs image data of a plurality of cultures and culture parameters related to the culture conditions of the culture process using the cultures.
- a feature extraction model for estimating culture parameters is generated by using the image data and the distribution of culture parameters as features.
- the state estimating unit 10c uses the cell image data obtained by the measuring unit 2, the reference culture parameter a before correction, and the correction candidate values ⁇ a0, ⁇ a1, ⁇ a1, ⁇ a1 is input (explanatory variable), and the culture state (differentiation state in this case) of cells after culture is output (objective variable).
- the state estimating unit 10c uses the image data of the target cell obtained through the measuring unit 2, the culture parameter a, and the plurality of correction candidate values ⁇ a of the culture parameter as input data for estimating the culture parameter described above.
- the culture state (differentiation state) of the target cell is estimated by inputting it into the feature quantity extraction model and analyzing it.
- the state estimating unit 10 c outputs a plurality of estimation results respectively obtained by changing the correction candidate value ⁇ a to the culture parameter optimization estimating unit 15 .
- the culture parameter optimization estimation unit 15 selects an estimation result that is closest to the preset optimal culture state result from among the plurality of estimation results obtained by the state estimation unit 10c, and sets with the selected estimation result.
- a correction candidate value ⁇ a for the culture parameter a that is set is specified.
- the state estimation unit 10a acquires image data of a plurality of cells obtained by the measurement unit 2 at a certain point in time.
- a feature quantity extraction model is generated using, as an explanatory variable, the culture time at which the culture reaches a desired culture state when the cells are actually cultured in the culture process, as an objective variable.
- culture time when the culture is cultured using a medium, the culture has a desired culture state (hereinafter simply referred to as culture time).
- supervised learning of other feature quantity extraction models includes, for example, image data of a plurality of cells, each culture time of these cells, and whether the culture time is the desired result.
- a correct label indicating whether or not (for example, a correct answer rate indicating whether or not the culture time is the desired)
- the correct label is attached when learning the image data of the cell and its culture time, and learning is performed.
- Various feature extraction models such as PCA, PLS, polynomial regression, Gaussian process regression, and Random Forest Regression can be applied as feature extraction models.
- a machine learning model generated by machine learning the result of the culture medium characteristic value (mass spectrometry data) and culture time may be applied.
- a neural network eg, CNN (convolutional neural network) or BNN (Bayesian neural network)
- unsupervised learning may also be used for the feature quantity extraction model. The details of generating a feature quantity extraction model by unsupervised learning will be described later, but when generating a feature quantity extraction model by unsupervised learning, image data of a plurality of cells are used, It is desirable to allow learning to characterize the distribution.
- mass spectrometry data is used to estimate optimal culture parameters based on the results of estimating the culture time. It goes without saying that it is possible to apply "mass spectrometry data" as "image data”.
- the state estimating unit 10c uses the image data of the target cell measured via the measuring unit 2, the culture parameter a, and the plurality of correction candidate values ⁇ a of the culture parameter as input data to estimate the culture parameter. It is possible to estimate the culture time required for the culture to reach the desired culture state (differentiation state) by inputting it into the feature quantity extraction model for analysis.
- the measurement unit that measures the properties of a plurality of biological samples and a target biological sample measures, for example, mass spectrometry data of a plurality of biological samples and a target biological sample.
- mass spectrometry data of a plurality of biological samples and a target biological sample.
- a medium, a culture target cultured in the medium, a medium extract extracted from the medium, a culture target extract extracted from the culture target, one of which is defined as a biological sample and a target biological sample, and the impedance of a plurality of biological samples at a certain point in time and at least one target biological sample of the same type as the biological sample at that point in time is used as a characteristic value for the measurement unit Measure at 2.
- the state estimation system 1a the impedances of a plurality of biological samples obtained through the measurement unit 2 are used as input data, a feature extraction regression model that characterizes the distribution of the input data is generated, and stored in the database 4. . Then, in the state estimating system 1a, the state estimating unit 10a estimates the culture state of the culture from the impedance of the target biological sample based on the feature quantity extraction model described above.
- the state estimation unit 10a estimates the culture state of the culture based on the feature amount extraction model from the impedance of the target medium before the start of the culture process.
- the estimation result state estimation result
- the state estimation result is output from the state estimation unit 10 a to the measurement object optimization estimation unit 14 .
- the measurement target optimization estimating unit 14 analyzes the state estimation result obtained by the state estimating unit 10a, and adds it to the target medium before use for culture, so that the medium unit volume contained in the target medium after use for culture Predict additives that can maximize the yield of the culture.
- the measurement target optimization estimation unit 14 corrects the components of the target medium used in the culture process by having the operator or the like add the estimated additive (also referred to as an estimated additive) to the target medium. be.
- the impedance of a plurality of biological samples and the culture parameters related to the culture conditions of the culture process using the biological samples are used as input data, and the characteristic A culture state is estimated from the input data of the target biological sample based on a feature extraction model that characterizes impedance as a value and distribution of culture parameters.
- the state estimation system 1a the impedances of a plurality of biological samples obtained via the measurement unit 2 are used as input data, and the distribution of this input data is A characterized feature extraction regression model is generated and stored in the database 4 . Then, in the state estimating system 1a, the state estimating unit 10a estimates the culturing time for the culture to reach a predetermined culturing state from the impedance of the target biological sample based on the above-described feature quantity extraction model.
- nucleic acid sequencing devices include, but are not limited to, nucleic acid sequencing devices using known DNA sequencing methods such as Sanger sequencing and Next Generation Sequencing (NGS). Also, the flow cytometer described in the fourth embodiment may be applied as the measurement unit 2 .
- the Raman spectrum obtained from the Raman spectrometer, the chromatogram obtained from the chromatography, the amount of the specific substance obtained from the digital PCR measurement device, the analysis data obtained from the nuclear magnetic resonance device, the specific antigen obtained from the antibody quantification kit and the gene sequence analysis data obtained from the nucleic acid sequencing device either of which is described in the above "(1) First embodiment", “(1-1) state estimation system configuration”, “(1-2) state estimation unit”, “(1-3) determination unit”, “(1-4) measurement object optimization estimation unit”, “(1-5) action and effect”, “ (1-6) Unsupervised generation of feature quantity extraction model”, “(1-7) Other embodiments regarding timing of measuring characteristics of biological sample”, “(1-8) In the middle of culturing before the end of the culturing process Other embodiments for estimating culture state of culture”, “(1-9) Optimization estimation process of culture parameters”, “(1-10) State estimation system for estimating culture time” and “(1-11 ) Other”, the “mass spectrometry data” (or measurement vector) acquired via
- the state estimating unit estimates the yield of the culture based on the feature extraction model from the input data of the medium at a certain point before the end of the culture, or
- the state estimating unit that estimates the culture time for a specified yield has been mainly described, as described above, the present invention is not limited to the "culture yield", but the "culture quality”.
- ⁇ yield (quality) of culture extract extracted from culture'', and ⁇ yield (quality) of components in medium secreted by cultured culture'' can be applied as ⁇ culture state of culture''.
- the quality of the culture is estimated based on the feature extraction model, or the culture time at which the culture reaches a predetermined quality It may be a state estimating unit that estimates .
- the yield (quality) of the culture extract extracted from the culture is estimated based on the feature extraction model, or the culture extraction It may be a state estimating unit that estimates the culture time at which the product reaches a predetermined yield (quality). Furthermore, from the input data obtained from the medium and culture at a certain point before the end of the culture, based on the feature extraction model, the yield (quality) of the components in the medium secreted by the cultivated culture can be estimated, or and a state estimating unit for estimating the culture time at which the component in the medium secreted by the culture reaches a predetermined yield (quality).
- the state estimation system uses as input data the measurement result obtained by measuring the characteristic value of the culture at a certain point in time by the measurement unit, and the culture is processed based on the feature amount extraction model from the input data. From among a plurality of types of culture states obtained when culturing according to , it is estimated which culture state will be in the future. Then, the state estimation system determines the next processing step to be performed on the culture at a certain point in time according to the type of culture state of the culture that has been estimated.
- FIG. 13 is a schematic diagram for explaining the types of the culture state of the culture A estimated in the state estimation system according to the fourth embodiment.
- An example of inducing differentiation into C or non-target cells is shown.
- the culture A shown in step S10 for example, human iPS cells (hereinafter simply referred to as iPS cells)
- iPS cells for example, human iPS cells (hereinafter simply referred to as iPS cells)
- step S11 the culture A is induced to differentiate into the paraxial mesoderm cell culture B (step S11) or the culture A is induced to differentiate into the lateral plate mesoderm cell culture C (step S12) by the influence of Alternatively, the case where the culture A becomes a non-target cell (step S13) such as another cell, a dead cell, or a cell in poor condition will be explained as an example. do.
- the optimal culture conditions for inducing the differentiation of the culture A into the culture B and the optimal culture conditions for inducing the differentiation of the culture A into the culture C are different cultures
- the condition In the state estimation system according to the fourth embodiment based on the estimation result of the culture A, a processing step of culturing the culture A under optimal culture conditions for inducing differentiation into the culture B, and a process of inducing differentiation into the culture C
- the culture conditions include, for example, various parameters (culture parameters) related to culture conditions such as culture temperature, culture time, medium pH, medium agitation speed, etc., in addition to the type and quality of medium.
- FIG. 14 in which the same components as in FIG. 1 are assigned the same reference numerals, is a block diagram showing the configuration of a state estimation system 1d according to the fourth embodiment.
- a state estimation system 1d is shown as an example of a biological sample.
- the fourth embodiment shows an example in which a flow cytometer using a flow cytometry method is applied as the measurement unit 2.
- FIG. 14 is a block diagram showing the configuration of a state estimation system 1d according to the fourth embodiment.
- an example of a culture is shown as an example of a biological sample.
- the fourth embodiment shows an example in which a flow cytometer using a flow cytometry method is applied as the measurement unit 2.
- the measurement unit 2 suspends the culture A at a certain point in time, such as before the start of the culture or immediately after the start of the culture, in a liquid, and the liquid is placed in the flow path so that the cultures A are aligned in a line. , a laser beam is applied to each of the cultures A, and reflected light such as scattered light and fluorescence is measured.
- the measurement unit 2 obtains characteristic values for each culture A by analyzing the measurement results of the reflected light, and generates analysis data indicating the characteristic values measured for each culture A as input data for each culture A. do.
- the state estimation system 1d measures a plurality of cultures A at a certain point in time using the measurement unit 2, and generates a feature quantity extraction model based on the obtained analysis data.
- the analysis data refers to the intensity of reflected light, the intensity of scattered light, the intensity of fluorescence, etc., which are measured when the culture A is analyzed by the flow cytometer, which is the measurement unit 2. It can be a characteristic value of the object A.
- the measurement unit 2 measures the culture A, which is undetermined to what kind of culture state it will actually be differentiated, and obtains analysis data, Based on the feature quantity extraction model and the above, it is possible to estimate what kind of culture state will be in the future when the culture A whose future culture state is undetermined at present is continued to be cultured. can.
- the culture A whose culture state is estimated using the feature quantity extraction model thus generated is referred to as a “target culture”.
- the measurement unit 2 which is a flow cytometer, measures each characteristic value of a plurality of cultures A and outputs the obtained measurement results to the arithmetic processing unit 3d and the database 4.
- the arithmetic processing unit 3 d has a state estimating unit 10 a , a measurement target optimization estimating unit 14 and a next process determining unit 22 .
- the measurement target optimization estimating unit 14 is the same as that of the above-described embodiment, so the description is omitted here, and the following description focuses on the state estimating unit 10a and the next process determining unit 22.
- FIG. In this case, when the feature quantity extraction model is generated, the measurement results of the plurality of cultures A by the flow cytometer (measurement unit 2) are input to the state estimation unit 10a as analysis data from the measurement unit 2 or the database 4. be done.
- the analysis data of the plurality of cultures A measured by the measuring unit 2 at a certain point in time, and the data obtained when each culture A was actually cultured using a predetermined medium Using the culture state results (hereinafter also referred to as culture results), for example, multivariate analysis is performed to generate a feature quantity extraction model.
- culture results hereinafter also referred to as culture results
- multivariate analysis is performed to generate a feature quantity extraction model.
- there are a plurality of states when the culture A is cultured namely, a culture B, a culture C, and a non-target cell.
- the state estimating unit 10a has a feature extraction model for the culture B that estimates whether the culture A is cultured to become the culture B, and an estimation of whether the culture A is cultured to become the culture C.
- a feature quantity extraction model for the culture C to be processed and a feature quantity extraction model for the non-target cell for estimating whether the culture A is cultured to become a non-target cell are generated.
- Such a feature quantity extraction model can analyze the state, quality, etc. of the culture A that affect the culture results when the culture A is cultured using a medium.
- the feature quantity extraction model generated for each culture result obtained by culturing culture A may be generated by supervised learning or unsupervised learning.
- a feature extraction model generated by supervised learning it is assumed that analysis data obtained as a result of measuring the target culture A at a certain point in time by the measurement unit 2 is input, and the culture of the target culture A is continued as it is. It is sometimes desirable to apply a regression model or the like that outputs the state estimation result for estimating the future culture state finally obtained.
- the state estimation unit 10a when the state estimation unit 10a generates a feature quantity extraction model for the culture B by supervised learning, a plurality of cultures A at a certain point in time are measured by the measurement unit 2 and obtained. The correspondence relationship between the obtained plurality of analysis data and the culture result of actually culturing each culture A and becoming culture B is learned, and the analysis data is obtained from the culture result when culture B is obtained. characterize the distribution of
- the measurement unit 2 measures a plurality of cultures A at a certain point in time.
- the correspondence relationship between the obtained analysis data and the culture result of actually culturing each of these cultures A to become the culture C is learned, and the culture result when becoming the culture C is analyzed. Characterize the distribution of data.
- the measurement unit 2 measures a plurality of cultures A at a certain point in time.
- the correspondence relationship between the plurality of analysis data obtained by the above and the culture result of actually culturing each of these cultures A and becoming non-target cells is learned, and the culture when becoming non-target cells Characterize the distribution of the analytical data from the results.
- supervised learning of the feature quantity extraction model for culture B for example, analysis data of a plurality of cultures A and each culture result obtained by actually culturing each of these cultures A and , and a correct label indicating whether or not the culture result is the desired culture B (for example, the correct answer rate indicating whether or not the desired culture B is obtained), and the analysis data of the culture A.
- the correct label that the culture A becomes the culture B may be added to the learning to characterize the distribution of the analysis data.
- the supervised learning of the feature quantity extraction model for the culture C (or for non-target cells) is also performed in the same way, and the analysis data of the plurality of cultures A and the cultures A are actually cultured.
- the desired culture C for example, desired culture B (or non-target cells) Accuracy rate indicating whether or not it is
- the correct label that culture A became culture C (or non-target cells) to characterize the distribution of the analytical data.
- Various feature extraction models such as principal component analysis (PCA), partial least squares (PLS), polynomial regression, Gaussian process regression, and random forest regression can be applied as feature extraction models.
- PCA principal component analysis
- PLS partial least squares
- polynomial regression Gaussian process regression
- random forest regression random forest regression
- the feature quantity extraction model for example, a machine learning model generated by performing machine learning on the characteristic values of the culture A (analysis data by a flow cytometer) and the culture results may be applied.
- CNN and BNN neural networks
- analysis data of a plurality of cultures A are used in the same manner as in the above-described embodiment. It is desirable to perform learning that characterizes the distribution of these multiple pieces of analysis data for each culture result of culture B, culture C, or non-target cells.
- the state estimating unit 10a determines the distribution of the analysis data group of a plurality of cultures A that have become cultures B by culturing, and the distribution of the analysis data group of cultures C and the distribution of the analysis data group of the plurality of cultures A obtained is approximated by, for example, a multidimensional Gaussian distribution.
- the non-target cells and others (here, culture B and culture C) It is possible to build a feature extraction model that identifies
- the distribution in the state estimation space of the analysis data groups of the plurality of cultures A that have become the culture B by culturing, and the state estimation space of the analysis data groups of the plurality of cultures A that have become the culture C by culturing A cluster corresponding to culture B and a cluster corresponding to culture C can be obtained by clustering the distribution in, for example, by the k-means method, which distinguishes between culture B and culture C. It is a feature extraction model.
- the state estimating unit 10a When estimating the future culture state of the target culture A, the state estimating unit 10a inputs the analysis data of the target culture A into the feature extraction model for the culture B, and cultures the target culture A as it is. An estimation result obtained by estimating whether or not the culture state of the target culture A will change to the culture B in the future based on the feature quantity extraction model for the culture B is output. Similarly, the state estimating unit 10a also inputs the analysis data of the target culture A to the feature amount extraction model for the culture C, and if the target culture A is continued to be cultured as it is, in the future Then, an estimation result obtained by estimating whether or not the culture state of the target culture A will become the culture C based on the feature amount extraction model for the culture C is output.
- the state estimating unit 10a inputs the analysis data of the target culture A into the feature amount extraction model for non-target cells, and continues to culture the target culture A as it is. An estimation result obtained by estimating whether or not the culture state of A is a non-target cell based on the feature quantity extraction model for the non-target cell is output.
- the state estimation unit 10a combines the estimation result obtained from the feature amount extraction model for culture B, the estimation result obtained from the feature amount extraction model for culture C, and the feature amount extraction model for non-target cells. Among the estimation results obtained from the above, the estimation result with the highest accuracy rate obtained in the estimation result is selected as the estimation result that it will be the future culture state when the target culture A is continued to be cultured. .
- feature quantity extraction models for culture B, culture C, and non-target cells are generated, and input data is input to each feature quantity extraction model, and the feature quantity
- the present invention is not limited to this.
- one feature quantity extraction model is generated, input data is input to the feature quantity extraction model, and an estimation result is obtained by estimating the future culture state of culture A based on one feature quantity extraction model. is desirable.
- the state estimating unit 10a when the state estimating unit 10a generates one feature quantity extraction model by supervised learning, a plurality of analysis data obtained by measuring a plurality of cultures A at a certain point in time with the measuring unit 2, By actually culturing each of these cultures A, learning the correspondence relationship between the culture results of culture B, culture C, or non-target cells, and one characterizing the distribution of analysis data from these culture results Generate a feature extraction model.
- the state estimating unit 10a when generating one feature quantity extraction model by unsupervised learning, the state estimating unit 10a performs the analysis data of the culture A that becomes the culture B and the analysis data of the culture A that becomes the culture C.
- the data and the analysis data of culture A which is the non-target cell, are clustered by the k-means method, and the classification optimization is performed for these multiple analysis data without labels in the state estimation space defined by the feature amount.
- the obtained clusters are labeled with culture B, culture C and target exovellus (adding identification values such as straight lines and curves that can be distinguished step by step) by workers etc.
- Generate a feature extraction model when generating one feature quantity extraction model by unsupervised learning.
- the state estimating unit 10a projects the analysis data of the target culture A onto the state estimation space of the feature quantity extraction model, and when the target culture A is continued to be cultivated, the target culture A will be Whether the culture state is culture B, culture C, or non-target cells can be estimated based on the discrimination value (label or the like).
- the next step determination unit 22 performs a treatment step (hereinafter also referred to as a treatment step for culture B) for culturing the target culture A under optimal culture conditions for inducing the differentiation of the target culture A into the culture B;
- a treatment step of culturing the target culture A under optimal culture conditions for inducing the differentiation of the culture A into the culture C (hereinafter also referred to as a treatment step for culture C), and a treatment step of discarding the target culture A (hereinafter also referred to as a waste treatment step)
- one of the treatment steps is determined based on the estimation result of the target culture A, and is output to the sorting section 25 as the next step information.
- next step determination unit 22 receives the estimation result that the future culture state of the target culture A is culture B from the state estimation unit 10a, it determines the processing step for culture B. On the other hand, when receiving from the state estimating unit 10a the estimated result that the future culture state of the target culture A is culture C, the culture C processing step is determined. Further, when receiving from the state estimating unit 10a the estimation result that the future culture state of the target culture A is non-target cells, the next step determining unit 22 determines the disposal process.
- the sorting unit 25 Based on the next-process information received from the next-process determination unit 22, the sorting unit 25 automatically sorts the target culture A and selects one of the first culture unit 26a, the second culture unit 26b, and the waste unit 27. It has a configuration for delivering the target culture A to a crab. As shown in FIG. 15, the sorting section 25 is provided with a substrate 30 on which a main channel 35, subchannels 38a and 38b, and branch channels 37a, 37b and 37c are formed.
- the main channel 35 formed in the substrate 30 communicates with the channel of the flow cytometer whose upstream side is the measurement unit 2, and the liquid 40 containing the plurality of cultures A arranged in a row flows through the flow cytometer. It has a configuration that flows in from the channel.
- the flow velocity of the liquid 40 flowing into the main channel 35 from the channel of the flow cytometer is controlled by a not-shown syringe pump, rotary pump, centrifugal pump, pneumatic pump, or the like.
- Branch channels 37a, 37b, and 37c communicate downstream with the main channel 35, and one ends of the subchannels 38a and 38b communicate between the channel of the flow cytometer and the branch channels 37a, 37b, and 37c. It is The secondary flow paths 38a and 38b are arranged so as to face each other so as to be orthogonal to the flow direction of the main flow path 35. As shown in FIG. The sub-channels 38a and 38b control the inflow of the control liquid into the main channel 35 to adjust the direction of flow of the liquid 40 in the main channel 35, and the cultures A contained in the liquid 40 are They are made to flow out to the optimum branch channels 37a, 37b, and 37c, respectively.
- the inflow of the control liquid from the sub-flow paths 38a and 38b into the main flow path 35 is controlled by a fluid control unit such as a syringe pump, rotary pump, centrifugal pump, or pneumatic pump provided in the sub-flow paths 38a and 38b. 38.
- a fluid control unit such as a syringe pump, rotary pump, centrifugal pump, or pneumatic pump provided in the sub-flow paths 38a and 38b. 38.
- downstream of the branch channel 37a communicates with the first culture unit 26a that executes the culture B treatment process
- downstream of the branch channel 37b communicates with the second culture unit that executes the culture C treatment process
- 26b and the downstream of the branch flow path 37c communicates with the disposal section 27 that performs the disposal process.
- the fluid control section 38 receives the next process information from the next process determining section 22, and the fluid control section 38 transfers the control liquid from the sub-flow paths 38a and 38b to the main flow path 35 according to the next-process information.
- the inflow is adjusted, and the target culture A for which the next step information is obtained is made to flow out to the branch channels 37a, 37b, and 37c corresponding to the next step information.
- the fractionating unit 25 is controlled by the fluid control unit 38 from the secondary channels 38a and 38b.
- the target culture A is guided to the branch channel 37a and guided to the first culture section 26a via the branch channel 37a.
- the target culture A for which the culture B treatment step has been determined is cultured under optimal culture conditions for inducing differentiation into the culture B by the first culturing unit 26a.
- the fractionating unit 25 is controlled by the fluid control unit 38 from the secondary channels 38a and 38b.
- the target culture A is guided to the branch channel 37b and guided to the second culture section 26b via the branch channel 37b.
- the target culture A for which the treatment process for the culture C has been determined is cultured under optimal culture conditions for inducing differentiation into the culture C by the second culture unit 26b.
- the fractionation unit 25 is configured such that the fluid control unit 38 causes the sub-flow paths 38a and 38b to flow into the main flow path 35.
- the target culture A is guided to the branch channel 37c and guided to the waste section 27 via the branch channel 37c.
- the target culture A for which the disposal process has been determined is discarded by the discarding section 27 .
- the state estimation system 1d uses the characteristic values of the plurality of cultures A measured by the measurement unit 2 as input data, and the distribution of this input data is Based on the feature quantity extraction model that characterizes, from the characteristic value of the target culture A, when the culture A is cultured in the future, any culture state among a plurality of types of culture states (culture B, culture C, or The state estimating unit 10a estimates whether the cell becomes a non-target cell).
- the state estimation system 1d as in the above-described embodiment, it is possible to preliminarily estimate whether or not the desired culture state can be obtained from the target culture A. can shorten the time required for As a result, the efficiency of bioproduction and research and development using the culture A can be enhanced.
- the state estimation system 1d when culturing the culture A, only the culture A that will be in a specific culture state in the future is sorted out at an early stage. , culturing can be carried out under optimum culturing conditions for bringing only the fractionated culture A into a specific culturing state.
- iPS cells are used as culture A, and the above-described state estimation system 1d
- target cells e.g., retina, nerves, red blood cells, myocardium, etc.
- iPS cells are used as culture A, and the above-described state estimation system 1d
- iPS cells are cultured by using the state estimation system 1d described above. Therefore, it is possible to estimate at an early stage before or during the culture of iPS cells whether or not they will become mesoderm cells, so the time required to determine the suitability of iPS cells can be greatly shortened.
- iPS cells when iPS cells are cultured to produce iPS cell-derived cardiomyocytes, which are target cells, iPS cells that can produce target cells with high probability in the future and target can be separated from iPS cells that may not be able to produce cells of interest, and only iPS cells that can produce the desired cells in the future with a high probability can be separated and cultured.
- the culture work can be efficiently performed by omitting the culture work.
- various blood cells such as red blood cells, white blood cells, and platelets obtained by culturing culture A, or substances and components other than cells obtained by culturing culture A are cultured based on a feature extraction model from the measurement results of culture A. You may make it estimate as a future culture state of the thing A.
- the present invention is not limited to this.
- the feature quantity extraction model may be generated using only the culture results obtained when the culture B and the culture C are obtained, without using the culture results of .
- the measurement result of the target culture A by the measurement unit 2 is used as input data, and the culture state when the target culture A is cultured based on the input data based on the feature amount extraction model.
- B, culture C, or non-target cells are estimated, but the present invention is not limited to this.
- an identification value (straight line, curve, etc.) that defines the culture B and the culture C in the state estimation space of the feature extraction model is set.
- a discrimination value that further subdivides the cells is defined, and the possibility that the target culture A induces differentiation into dead cells (cells in poor condition) and the like among non-target cells and other cells is discriminated.
- a separately estimated estimation result may be obtained based on the feature quantity extraction model.
- a feature quantity extraction model obtained from the measurement result (input data) of is generated by supervised learning or unsupervised learning in the same manner as the cultures B and C described above.
- the future culture state of the target culture A (culture B, culture C and target target outer cells) may be estimated.
- culture A when culture A is cultured, differentiation is induced into one of three of culture B, culture C, and non-target cells
- the invention is not so limited.
- culture E e.g., red blood cells
- non-target cells red blood cells, white blood cells, It may be applied when differentiation is induced into various blood cells such as platelets), or when differentiation is induced into four or more cells.
- the culture A when the culture A is cultured, the culture E is multistagely differentiated via the culture D. You may apply the state estimation system 1d which concerns.
- the culture A for example, human iPS cells (iPS cells)
- the culture A may be affected by the state of the culture A.
- Product A is induced to differentiate into culture D of blood progenitor cells (step S15), or other non-target cells such as other cells, dead cells, cells in poor condition, etc. (step S13) An example of two types of culture conditions is shown.
- culture E red blood cells
- dead cells cells in poor condition, etc. 2
- FIG. 2 shows an example of two types of culture states, namely, whether the target cells (step S13) are the non-target cells.
- the first feature amount extraction for estimating whether the culture A is cultured to become the culture D or the culture A is cultured to become the non-target cell A model is generated by the state estimator 10a.
- a state estimation unit generates a second feature quantity extraction model for estimating whether the culture D is cultured to become the culture E or the culture D is cultured to become non-target cells. 10a.
- the generation of the feature quantity extraction model here is generated in the same manner as described above, so the description thereof will be omitted here.
- the feature amount extraction model as described above, the feature amount extraction model for the culture D that estimates whether the culture A is cultured to become the culture D, and the culture A is cultured to obtain the target object
- a feature extraction model for estimating whether or not to become an outer cell a feature extraction model for culture E that estimates whether culture D will become culture E, and a target object by culturing culture D
- a feature quantity extraction model for estimating whether the cell will be an outer cell may be generated separately.
- the measurement unit 2 which is a flow cytometer, measures each characteristic value of a plurality of target cultures A, and obtains the obtained measurement results as analysis data.
- the state estimating unit 10a inputs the analysis data of the target culture A into the first feature quantity extraction model, and when the target culture A is continued to be cultured as it is, the culture state of the target culture A will be a culture in the future. An estimation result is obtained based on the first feature quantity extraction model as to whether or not the cells are D or non-target cells.
- the state estimating unit 10 a When the state estimating unit 10 a obtains an estimation result of estimating the future culture state for each target culture A, it outputs this to the next step determining unit 22 .
- the next step determination unit 22 performs a treatment step (hereinafter also referred to as a treatment step for culture D) for culturing the target culture A under optimal culture conditions for inducing the differentiation of the target culture A into the culture D, and Among the processing steps for discarding the culture A (hereinafter also referred to as the first disposal step), the corresponding processing step is determined based on the estimation result of the target culture A, and is sent to the fractionation unit 25 as the next step information. Output.
- a treatment step for culture D for culturing the target culture A under optimal culture conditions for inducing the differentiation of the target culture A into the culture D
- the processing steps for discarding the culture A hereinafter also referred to as the first disposal step
- the corresponding processing step is determined based on the estimation result of the target culture A, and
- the sorting unit 25 automatically sorts the target culture A based on the next step information received from the next step determination unit 22, and obtains the next step information of the treatment step for the culture D. are selected and delivered to the first culture unit 26a. In addition, the sorting unit 25 sends the target culture A for which the next step information of the first disposal step is obtained based on the estimation result that the cells will be non-target cells to the disposal unit 27, and the target culture A is sent to the disposal unit 27. Culture A is discarded.
- the measurement unit 2 which is a flow cytometer, sets the culture D obtained by culturing the target culture A in the first culture unit 26a as the target culture D, and measures the characteristic values of each target culture D. Then, the obtained measurement results are obtained as analysis data.
- the state estimating unit 10a inputs the analysis data of the target culture D to the second feature quantity extraction model, respectively. Based on the second feature quantity extraction model, it is estimated whether or not it becomes the entity E or the non-target cell.
- the state estimating unit 10 a When the state estimating unit 10 a obtains an estimation result of estimating the future culture state for each target culture D, it outputs this to the next process determining unit 22 .
- the next step determination unit 22 performs a treatment step (hereinafter also referred to as a treatment step for culture E) for culturing the target culture D under optimal culture conditions for inducing the differentiation of the target culture D into the culture E;
- a treatment step for culture E for culturing the target culture D under optimal culture conditions for inducing the differentiation of the target culture D into the culture E
- the processing steps for discarding the culture D hereinafter also referred to as a second disposal step
- the corresponding processing step is determined based on the estimation result of the target culture D, and is sent to the fractionation unit 25 as the next step information. Output.
- the sorting unit 25 automatically sorts the target culture D based on the next step information received from the next step determination unit 22, and obtains the next step information of the treatment step for the culture E. are selected and delivered to the second culture unit 26b. In addition, the sorting unit 25 sends the target culture D, which has obtained the next step information of the second discarding process based on the estimation result that the cells will be non-target cells, to the discarding unit 27, and in the discarding unit 27 Culture D is discarded.
- the measurement unit based on the feature extraction model The culture state of the culture may be estimated from the obtained input data.
- the target culture A for which the estimation result that it will be the culture D is obtained in the state estimation step is fractionated (fractionation step), and cultured in the first culture unit. to obtain a culture D (culture step), and a culture D′ (not shown) is obtained by adding a mutagen that induces mutation to the obtained culture D or performing genetic manipulation, etc. (mutation step). Then, a feature amount extraction model for estimating the future culture state of the culture D' is generated, and the feature amount extraction model is used to estimate the culture state of the culture D' (state estimation step).
- the culture state of the culture is selected and mutated, and the selected culture is (For example, a culture group consisting only of mutated cultures, or a culture group containing both mutated cultures and non-mutated cultures) is cultivated to modify new functions, etc. It is also possible to obtain a cultured product, and it is also possible to breed the culture.
- the above mutation step does not necessarily have to be performed.
- the culture A may be fractionated and the above culture step, state estimation step, fractionation step, etc. may be repeated.
- the target culture A whose culture state was estimated based on the estimation result of the feature quantity extraction model was cultured, and the culture D'' etc. in which the mutation occurred was fractionated and cultured as described above.
- the step, the state estimation step, and the fractionation step may be repeated.
- a target culture A that may become a cultured state of mutation in the future may be sorted based on the estimation result of the feature extraction model, and the above-described culturing step, state estimation step, sorting step, etc. may be repeated. .
- a flow cytometer is applied as a measurement unit and analysis data obtained from the flow cytometer is used as input data was described, but the present invention is not limited to this, for example , a mass spectrometer, a microscope, a Raman spectrometer, a chromatography, a digital PCR measurement device, a nuclear magnetic resonance device, an antibody quantification kit, and a nucleic acid sequencing device may be applied as the measuring unit.
- the input data include mass spectra obtained from a mass spectrometer, image data obtained from a microscope, Raman spectra obtained from a Raman spectrometer, chromatograms obtained from chromatography, and specific data obtained from a digital PCR measurement device.
- the amount of substance, analysis data obtained from a nuclear magnetic resonance apparatus, the amount of a specific antigen obtained from an antibody quantification kit, and gene sequence analysis data obtained from a nucleic acid sequencing apparatus can be applied as input data.
- the input data including at least the characteristic values of the plurality of cultures obtained through the measurement unit 2 is the analysis data of the plurality of cultures obtained by the flow cytometer.
- generation of a feature quantity extraction model using the input data and obtaining an estimation result from the feature quantity extraction model have been described, but the present invention is not limited to this.
- parameters related to culture conditions such as culture temperature, culture time, medium pH, medium agitation speed, etc. (culture parameters ) may be used to generate a feature quantity extraction model, or an estimation result may be obtained from the feature quantity extraction model.
- the mass spectrum obtained from the mass spectrometer described above the image data obtained from the microscope device, the Raman spectrum obtained from the Raman spectrometer, the chromatogram obtained from chromatography, the amount of the specific substance obtained from the digital PCR measurement device, Similarly, when using analysis data obtained from a nuclear magnetic resonance apparatus, the amount of a specific antigen obtained from an antibody quantification kit, and gene sequence analysis data obtained from a nucleic acid sequencing apparatus as input data, in addition to these, culture Input data containing parameters (cultivation parameters) relating to conditions may be used, and the input data may be used to generate a feature amount extraction model or to obtain an estimation result from the feature amount extraction model.
- culture Input data containing parameters (cultivation parameters) relating to conditions may be used, and the input data may be used to generate a feature amount extraction model or to obtain an estimation result from the feature amount extraction model.
Landscapes
- Engineering & Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Organic Chemistry (AREA)
- Zoology (AREA)
- Wood Science & Technology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Health & Medical Sciences (AREA)
- Biochemistry (AREA)
- General Engineering & Computer Science (AREA)
- Analytical Chemistry (AREA)
- Genetics & Genomics (AREA)
- Business, Economics & Management (AREA)
- Sustainable Development (AREA)
- Microbiology (AREA)
- Biotechnology (AREA)
- Biomedical Technology (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Primary Health Care (AREA)
- Manufacturing & Machinery (AREA)
- Physics & Mathematics (AREA)
- Computer Hardware Design (AREA)
- Measuring Or Testing Involving Enzymes Or Micro-Organisms (AREA)
- Other Investigation Or Analysis Of Materials By Electrical Means (AREA)
- Apparatus Associated With Microorganisms And Enzymes (AREA)
Abstract
Description
(1―1)状態推定システムの構成
図1は、本実施形態に係る状態推定システム1aの構成を示すブロック図である。なお、第1実施形態では、生体試料の一例として培地の例を示す。また、第1実施形態は、後述する測定部2として、質量分析装置を適用した例である。本実施形態に係る状態推定システム1aは、細胞培養実験に使用される培地の状態や培養物の培養状態を、特徴量を規定した特徴量抽出モデルに基づいて推定するものであり、測定部2と演算処理部3aとデータベース4と通知部5とを有する。
次に、状態推定部10aにおける状態推定処理について説明する。ここでは、一例として、PLS回帰により特徴量抽出モデルを生成する例について説明する。なお、状態推定部10aにより推定する、培養物の培養状態としては、上述したように、培養プロセスで得られる培養物の品質(機能性)でもよいが、ここでは、一例として、培養プロセスで得られる培養物の収量とした場合について主に説明する。
次に、判定部12における判定処理について説明する。判定部12は、状態推定部10aにより得られた状態推定結果に基づいて、対象培地の状態を判定するものである。
次に、測定対象最適化推定部14における測定対象最適化推定処理について説明する。測定対象最適化推定部14は、状態推定部10aにおいて、培養プロセスの開始前の対象培地の質量分析データから、特徴量抽出モデルに基づいて培養物の培養状態の推定結果(状態推定結果)が得られると、状態推定部10aから当該状態推定結果を受け取る。測定対象最適化推定部14は、状態推定部10aで得られた状態推定結果を解析し、培養使用前の対象培地に添加することで、培養使用後の対象培地に含まれる培地単位体積あたりの培養物の収量を最大化できる、添加物を推定する。測定対象最適化推定部14は、作業者等に対して、推定した当該添加物(推定添加物とも称する)を対象培地に添加させることによって、培養プロセスに用いる対象培地の成分を補正させるものである。
以上の構成において、状態推定システム1aは、複数の培地及び少なくとも1個の対象培地の質量分析データを特性値(入力データ)として測定する測定部2と、複数の培地の特性値(入力データ)の分布を特徴づける特徴量抽出モデルに基づいて、対象培地の特性値から、培養物の培養状態を推定する状態推定部10aと、を設けるようにした。このように、状態推定システム1aでは、対象培地を用いたバイオ生産及び研究開発において所望する培養状態を得ることができるか否かを、事前に推定できるので、対象培地の適正判断に要する時間を短縮化し得る。これにより、培地を用いたバイオ生産や研究開発の効率を高めることができる。
上述した実施形態においては、複数の生体試料の特性値(質量分析データ)の分布を特徴づける特徴量抽出モデルとして、培地単位体積あたりの培養物の収量が大きく良好な培養状態を得る、という目的に対して、関連性の高い多次元(例えば、30次元)の主成分ベクトル(教師あり基底)を抽出して特徴量抽出モデルを得る場合について説明した。
(1-7-1)培養プロセス開始後の培養プロセス中における、培地の特性を測定する場合
上述した実施形態においては、培養プロセスを開始する前に、測定部2によって培地の質量分析を行い、質量分析データを得る場合について説明したが、本発明はこれに限らない。上述した実施形態において、例えば、図10の10Aに示すように、培養プロセスを開始した後、培養プロセスで使用している途中の培地について測定部2により質量分析を行い、質量分析データを得るようにしてもよい。
上述した実施形態においては、培養プロセスを開始する前、及び、培養プロセス中のいずれかに、測定部2によって培地の質量分析を行い、質量分析データを得る場合について説明したが、本発明はこれに限らない。上述した実施形態では、例えば、図10の10Bに示すように、培養プロセスを開始する前のある時点Sa1と、培養プロセス中のある時点Sa2,Sa3とにおいて、それぞれ培地について測定部2により質量分析を行い、質量分析データを複数回得るようにしてもよい。また、培養プロセスを開始する前に測定部2によって培地の質量分析を行わずに、培養プロセス中のある時点Sa2,Sa3でのみそれぞれ培地について測定部2により質量分析を行い、質量分析データを複数回得るようにしてもよい。すなわち、培養プロセスの開始から10日目の培地の質量分析データと、培養プロセスの開始から20日目の培地の質量分析データ等、培養開始から、時系列にある間隔で複数回、培地の質量分析データを測定するようにしてもよい。
上述した実施形態においては、培養プロセスを終了した後(終了時期F後)に、培地に含まれる培養物の培養状態(例えば、対象培地に含まれる培地単位体積あたりの培養物の収量)を推定する場合について説明したが、本発明はこれに限らない。上述した実施形態において、例えば、図11の11Aに示すように、培養プロセスが終了する前の培養途中のある時点Fa2における培養物の培養状態を推定するようにしてもよい。
次に、培養パラメータの最適化推定処理を行う、他の実施形態に係る状態推定システムについて説明する。図12は、他の実施形態に係る状態推定システム1cの構成を示すブロック図である。状態推定システム1cは、上述した図1の状態推定システム1aとは演算処理部3cの構成が異なるものであり、ここでは、この演算処理部3cに着目して以下説明し、測定部2やデータベース4、通知部5の説明については重複するため省略する。
例えば、培養物を培養する際の培養温度、培養時間、培地のpH、培地の撹拌速度等の基準となる培養パラメータのセットを決める。また、この基準となる培養パラメータからの培養パラメータの補正候補値△aを複数決める。具体的には、培養温度の補正候補値△aとは、例えば、+2℃や-1℃等であり、培養時間の補正候補値△aとは、例えば、+30分や-30分等であり、培地のpHの補正候補値△aとは、+0.1や-0.2等であり、培地の撹拌速度の補正候補値△aとは、例えば、+1s-1や-1s-1等である。
状態推定システム1cは、測定部2を介して、成分が未知の対象培地の質量分析データを得ると、培養パラメータ最適化推定部15により、予め設定した範囲内で培養パラメータの補正候補値△aをランダムに選定してゆき、複数の補正候補値△aを状態推定部10cに出力する。なお、ここでは、培養パラメータの補正候補値△aについて、培養パラメータ最適化推定部15でランダムに生成した場合について述べたが、本発明はこれに限らず、例えば、データベース4に予め保存しておいた複数の補正候補値△aを適用してもよく、また、作業者等が選択した補正候補値△aであってよい。
以上の構成において、状態推定システム1cは、測定部2で測定した複数の培地の特性値及び複数の培養パラメータを入力データとし、この入力データの分布を特徴づける、培養パラメータ推定用の特徴量抽出モデルに基づいて、対象培地の特性値及び培養パラメータから、培養物の培養状態を状態推定部10cで推定するようにした。このように、状態推定システム1cでは、対象培地を用いたバイオ生産及び研究開発において所望する培養状態を得ることができるか否かを、事前に推定できるので、対象培地の適正判断に要する時間を短縮化し得る。これにより、培地を用いたバイオ生産や研究開発の効率を高めることができる。
上述した実施形態においては、特徴量抽出モデルに基づいて培養物の培養状態を推定する状態推定システム1a,1cについてそれぞれ説明したが、本発明はこれに限らず、培養プロセスで培養している培養物が予め設定した培養状態になるまでの培養時間を、特徴量抽出モデルに基づいて推定する状態推定システムとしてもよい。
例えば、図1に示す状態推定システム1aを、培養時間を推定する状態推定システム1aとした場合、状態推定部10aでは、教師あり学習の場合、ある時点において測定部2により得られた複数の培地の質量分析データを説明変数とし、当該培地を用いて培養物を実際に培養したときに培養物が所望の培養状態となる培養時間を目的変数として用いて特徴量抽出モデルを生成する。このような、特徴量抽出モデルでは、培地を用いて培養物を培養したときに培養物が所望の培養状態となる培養時間(以下、単に培養時間とも称する)に影響を与える、培地の成分を解析することができる。
また、図12に示す状態推定システム1cは培養パラメータの最適化推定処理を行うものであるが、この状態推定システム1cにおいても、上述した「(1-9)培養パラメータの最適化推定処理」に従い、培養時間の推定結果に基づいて、最適な培養パラメータを推定するようにしてもよい。この場合、上述した「(1-9-1)培養パラメータ推定用の特徴量抽出モデルの生成」では、培地の質量分析データと、基準となる培養パラメータaと、培養パラメータの補正候補値△a0,△a1,△a2,△a3と、培養パラメータ(a+△a0),(a+△a1),(a+△a2),(a+△a3)で実際に培養物を培養したときに培養物が所望の培養状態となるまでにかかる培養時間を示す実態結果と、を対応付け、これをそれぞれデータセットとしてデータベース4に保存する。
なお、上述した実施形態においては、生体試料及び対象生体試料として、培地を適用した場合について述べたが、本発明はこれに限らず、例えば、培地と、培地により培養される培養物と、培地から抽出される培地抽出物と、培養物から抽出される培養抽出物とのうち、いずれかを生体試料及び対象生体試料として適用してもよい。
上述した第1実施形態においては、複数の生体試料及び対象生体試料の特性を測定する測定部として、例えば、複数の生体試料及び対象生体試料の質量分析データを測定する質量分析装置を適用した場合について述べたが、第2実施形態では、複数の生体試料及び対象生体試料の外見を測定する顕微鏡装置を適用する場合について以下説明する。
この場合、図1に示す状態推定システム1aでは、測定部2として、CCD(Charge Coupled Device)カメラ等の撮像装置を備えた顕微鏡装置(例えば、蛍光顕微鏡装置)を適用する。顕微鏡装置である測定部2は、培地内における培養物の外見状態を当該培養物の特性として測定するために当該培養物を撮像し、培養物の画像データを取得する。なお、第2実施形態では、生体試料が培養物であり、培養物として、例えば、細胞培養実験に使用される細胞を一例に説明する。
なお、実際に培養した細胞の分化状態は、例えば、質量分析装置によって細胞からの代謝物を分析することで当該細胞の分化状態を特定することができ、また、RNAシーケンシング(RNA-Seq)によって、培養後の細胞における遺伝子の発現状態を計測、解析することで当該細胞の分化状態を推定することができる。
例えば、複数の細胞の画像データを用いてPCA(主成分分析)処理を行い、画像データの細胞の外見的状態の分布を特徴づける多次元の主成分ベクトル(教師なし基底)を抽出して特徴量抽出モデルを得る状態推定システム1aについて説明する。
(2-3-1)培養プロセス開始後の培養プロセス中における、細胞の特性を測定する場合
第2実施形態でも、上述した第1実施形態と同様に、例えば、図10の10Bに示すように、培養プロセスを開始する前のある時点Sa1と、培養プロセス中のある時点Sa2,Sa3とにおいて、それぞれ細胞を測定部2により撮像し、画像データを複数回得るようにしてもよい。また、培養プロセスを開始する前に測定部2によって細胞を撮像せずに、培養プロセス中のある時点Sa2,Sa3でのみそれぞれ細胞について測定部2により撮像し、当該細胞の画像データを複数回得るようにしてもよい。すなわち、培養プロセスの開始から10日目の細胞の画像データと、培養プロセスの開始から20日目の細胞の画像データ等、培養開始から、時系列にある間隔で複数回、細胞の画像データを得るようにしてもよい。
上述した第2実施形態においては、培養プロセスを終了した後(終了時期F後)に、培地に含まれる対象細胞の培養状態(例えば、対象細胞の分化状態)を推定する場合について説明したが、本発明はこれに限らない。上述した第2実施形態でも、例えば、図11の11Aに示すように、培養プロセスが終了する前の培養途中のある時点Fa2における対象細胞の培養状態を推定するようにしてもよい。
上述した第1実施形態においては、複数の生体試料及び対象生体試料の特性を測定する測定部として、例えば、複数の生体試料及び対象生体試料の質量分析データを測定する質量分析装置を適用した場合について述べたが、第3実施形態では、複数の生体試料及び対象生体試料のインピーダンスを測定する測定装置を適用する場合について、以下概略を説明する。
また、第4実施形態にて説明するフローサイトメータを測定部2として適用してもよい。
(4-1)第4実施形態に係る状態推定システム
次に、第4実施形態について説明する。第4実施形態に係る状態推定システムは、ある時点の培養物の特性値を測定部で測定した測定結果を入力データとし、当該入力データから特徴量抽出モデルに基づいて、当該培養物を培養プロセスに従って培養したときに得られる複数種類の培養状態のうちから、将来いずれの培養状態になるかを推定する。そして、状態推定システムは、推定した培養物の培養状態の種類に応じて、ある時点の培養物に対して次に行う処理工程を決定する。
以上の構成において、第4実施形態に係る状態推定システム1dでは、測定部2で測定した複数の培養物Aの特性値を入力データとし、この入力データの分布を特徴づける特徴量抽出モデルに基づいて、対象培養物Aの特性値から、将来的に培養物Aを培養したときに複数種類の培養状態のうちいずれの培養状態(培養物B、培養物C又は目的対象外細胞)になるかを状態推定部10aで推定するようにした。このように、状態推定システム1dでも、上述した実施形態と同様に、対象培養物Aから所望する培養状態を得ることができるか否かを、事前に推定できるので、対象培養物Aの適正判断に要する時間を短縮化し得る。これにより、培養物Aを用いたバイオ生産や研究開発の効率を高めることができる。
(4-3)他の実施形態
なお、上述した第4実施形態においては、培養物Aの測定結果から特徴量抽出モデルに基づいて、培養物Aを培養して得られる細胞(培養物B、培養物C)を、培養物Aの将来的な培養状態として推定するようにした場合について述べたが、本発明はこれに限らない。例えば、培養物Aの測定結果から特徴量抽出モデルに基づいて、培養物Aを培養して得られる赤血球、白血球、血小板等の様々な血液細胞、又は、細胞以外の物質や成分等を、培養物Aの将来的な培養状態として推定するようにしてもよい。
なお、上述した第1実施形態から第4実施形態に係る状態推定システム1a,1c,1dでは、培養物の育種のために、特徴量抽出モデルに基づいて、測定部で得られた入力データから培養物の培養状態を推定するようにしてもよい。
2 測定部
10a、10c 状態推定部
12 判定部
14 測定対象最適化推定部
15 培養パラメータ最適化推定部
Claims (19)
- 培地と、前記培地により培養される培養物と、前記培地から抽出される培地抽出物と、前記培養物から抽出される培養抽出物と、のうち、いずれかを生体試料及び対象生体試料とし、ある時点における複数の前記生体試料と、前記時点における、前記生体試料と同じ種類の少なくとも1個の前記対象生体試料と、の特性を測定する測定部と、
前記測定部を介して取得した前記複数の生体試料の特性値を少なくとも含む入力データについて分布を特徴づけた特徴量抽出モデルを記憶するデータベースと、
前記特徴量抽出モデルに基づいて、前記対象生体試料の前記入力データから、前記培養物の培養状態を推定するか、又は、前記培養物が予め定めた培養状態となる培養時間を推定する、状態推定部と、
を備える、状態推定システム。 - 前記測定部は、
培養プロセスを開始する前における、前記複数の生体試料及び前記対象生体試料の特性を測定する、
請求項1に記載の状態推定システム。 - 前記測定部は、
培養プロセスで使用している途中の前記複数の生体試料及び前記対象生体試料の特性を測定する、
請求項1に記載の状態推定システム。 - 前記状態推定部は、
培養プロセスで定められた培養期間を経過し、培養を終了した前記培養物の培養状態を推定する、
請求項1~3のいずれか1項に記載の状態推定システム。 - 前記状態推定部は、
培養プロセスにより培養している培養途中の前記培養物の培養状態を推定する、
請求項1~3のいずれか1項に記載の状態推定システム。 - 前記測定部は、
培養プロセスで使用している途中の前記複数の生体試料及び前記対象生体試料の特性を時系列に測定し、
前記状態推定部は、
前記特性値が前記入力データであり、前記特性値の測定時期ごとに前記複数の生体試料の前記特性値の分布を特徴づけた前記特徴量抽出モデルに基づいて、前記特性値の測定時期ごとに前記対象生体試料の前記特性値から、前記培養物の培養状態を推定するか、又は、前記培養時間を推定する、
請求項1~5のいずれか1項に記載の状態推定システム。 - 前記状態推定部により得られた前記培養状態の推定結果、又は、前記培養時間の推定結果に基づいて、前記対象生体試料の状態を判定する判定部を有する、
請求項1~6のいずれか1項に記載の状態推定システム。 - 前記判定部は、
前記状態推定部により得られた前記培養状態の推定結果、又は、前記培養時間の推定結果に基づいて、前記対象生体試料の異常の有無を判定する、
請求項7に記載の状態推定システム。 - 前記複数の生体試料及び前記対象生体試料は前記培地であって、
前記状態推定部により得られた前記培養状態の推定結果、又は、前記培養時間の推定結果に基づいて前記培地に添加する添加物を推定する測定対象最適化推定部を有する、
請求項1~5のいずれか1項に記載の状態推定システム。 - 前記測定対象最適化推定部は、
前記特徴量抽出モデルにおいて状態推定空間に定義された主成分ベクトルと、複数種類の前記添加物をそれぞれ前記測定部により測定することで得られる各添加物測定ベクトルと、の類似度をそれぞれ計算し、前記主成分ベクトルと前記類似度が高い前記添加物測定ベクトルの前記添加物を推定する、
請求項9に記載の状態推定システム。 - 前記状態推定部は、
前記複数の生体試料の前記特性値と、前記生体試料を用いた培養プロセスの培養条件に関する培養パラメータと、を前記入力データとして用い、前記特性値及び前記培養パラメータの分布を特徴づけた前記特徴量抽出モデルに基づいて、前記対象生体試料の前記入力データから、前記培養状態を推定するか、又は、前記培養時間を推定する、
請求項1~5のいずれか1項に記載の状態推定システム。 - 前記状態推定部により得られた前記培養状態の推定結果、又は、前記培養時間の推定結果に基づいて、前記培養パラメータの適正を判定する判定部を有する、
請求項11に記載の状態推定システム。 - 前記測定部は、
質量分析装置、顕微鏡装置、ラマン分光装置、クロマトグラフィー、デジタルPCR測定装置、核磁気共鳴装置、抗体定量キット、核酸配列決定装置、及び、フローサイトメータのうち、いずれか1種以上であり、
前記測定部が測定する前記特性値は、
前記質量分析装置から得られるマススペクトル、前記顕微鏡装置から得られる画像データ、前記ラマン分光装置から得られるラマンスペクトル、前記クロマトグラフィーから得られるクロマトグラム、前記デジタルPCR測定装置から得られる特定物質の量、前記核磁気共鳴装置から得られる分析データ、前記抗体定量キットから得られる特定抗原の量、前記核酸配列決定装置から得られる遺伝子配列分析データ、及び、フローサイトメータから得られる分析データである、
請求項1~12のいずれか1項に記載の状態推定システム。 - 前記測定部は、インピーダンス測定装置であり、
前記測定部が測定する前記特性値は、前記複数の生体試料及び前記対象生体試料のインピーダンスである、
請求項1~12のいずれか1項に記載の状態推定システム。 - 前記培養物の前記培養状態とは、前記培養物の収量、前記培養物の品質、前記培地抽出物の収量、又は、前記培養抽出物の品質、である、
請求項1~14のいずれか1項に記載の状態推定システム。 - 前記測定部は、
前記時点における前記培養物の特性を測定し、
前記状態推定部は、
前記測定部から得られた測定結果を前記入力データとし、前記入力データから、前記特徴量抽出モデルに基づいて、前記培養物が培養により、複数種類の培養状態のうち、いずれの培養状態になるかを推定する、
請求項1~15のいずれか1項に記載の状態推定システム。 - 前記状態推定部により推定した前記培養物の培養状態の種類により、前記培養物に対する次の処理工程を決定する次工程決定部を備える、
請求項16に記載の状態推定システム。 - 前記状態推定部は、
前記培養物の育種のために、前記特徴量抽出モデルに基づいて前記入力データから前記培養物の培養状態を推定する、
請求項1~17のいずれか1項に記載の状態推定システム。 - 培地と、前記培地により培養される培養物と、前記培地から抽出される培地抽出物と、前記培養物から抽出される培養抽出物と、のうち、いずれかを生体試料及び対象生体試料とし、ある時点における複数の前記生体試料と、前記時点における、前記生体試料と同じ種類の少なくとも1個の前記対象生体試料と、の特性を測定部で測定する測定ステップと、
前記測定部を介して取得した前記複数の生体試料の特性値を少なくとも含む入力データについて分布を特徴づけた特徴量抽出モデルをデータベースに記憶する記憶ステップと、
前記特徴量抽出モデルに基づいて、前記対象生体試料の前記入力データから、前記培養物の培養状態を状態推定部により推定するか、又は、前記培養物が予め定めた培養状態となる培養時間を状態推定部により推定する、状態推定ステップと、
を含む、状態推定方法。
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP22824636.9A EP4357446A1 (en) | 2021-06-18 | 2022-03-30 | State inference system and state inference method |
CN202280042792.1A CN117561326A (zh) | 2021-06-18 | 2022-03-30 | 状态推定系统及状态推定方法 |
JP2023529621A JPWO2022264640A1 (ja) | 2021-06-18 | 2022-03-30 |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2021102070 | 2021-06-18 | ||
JP2021-102070 | 2021-06-18 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2022264640A1 true WO2022264640A1 (ja) | 2022-12-22 |
Family
ID=84527069
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2022/016221 WO2022264640A1 (ja) | 2021-06-18 | 2022-03-30 | 状態推定システム及び状態推定方法 |
Country Status (4)
Country | Link |
---|---|
EP (1) | EP4357446A1 (ja) |
JP (1) | JPWO2022264640A1 (ja) |
CN (1) | CN117561326A (ja) |
WO (1) | WO2022264640A1 (ja) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2018117567A (ja) * | 2017-01-25 | 2018-08-02 | 株式会社日立製作所 | 細胞培養装置 |
WO2019031545A1 (ja) * | 2017-08-08 | 2019-02-14 | 東京エレクトロン株式会社 | 多能性幹細胞の未分化状態を判定する方法、多能性幹細胞の継代培養方法およびそれら方法に使用される装置 |
WO2019069378A1 (ja) * | 2017-10-03 | 2019-04-11 | オリンパス株式会社 | 培養情報処理装置 |
WO2020039683A1 (ja) * | 2018-08-22 | 2020-02-27 | 富士フイルム株式会社 | 細胞培養支援装置の作動プログラム、細胞培養支援装置、細胞培養支援装置の作動方法 |
WO2021049044A1 (ja) * | 2019-09-13 | 2021-03-18 | エピストラ株式会社 | 培地製造方法、培地製造パラメータ決定方法、培地、およびプログラム |
JP2021043600A (ja) * | 2019-09-09 | 2021-03-18 | 株式会社カネカ | 学習モデルの生成方法、評価推定装置、学習モデル、及びコンピュータプログラム |
-
2022
- 2022-03-30 JP JP2023529621A patent/JPWO2022264640A1/ja active Pending
- 2022-03-30 WO PCT/JP2022/016221 patent/WO2022264640A1/ja active Application Filing
- 2022-03-30 EP EP22824636.9A patent/EP4357446A1/en active Pending
- 2022-03-30 CN CN202280042792.1A patent/CN117561326A/zh active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2018117567A (ja) * | 2017-01-25 | 2018-08-02 | 株式会社日立製作所 | 細胞培養装置 |
WO2019031545A1 (ja) * | 2017-08-08 | 2019-02-14 | 東京エレクトロン株式会社 | 多能性幹細胞の未分化状態を判定する方法、多能性幹細胞の継代培養方法およびそれら方法に使用される装置 |
WO2019069378A1 (ja) * | 2017-10-03 | 2019-04-11 | オリンパス株式会社 | 培養情報処理装置 |
WO2020039683A1 (ja) * | 2018-08-22 | 2020-02-27 | 富士フイルム株式会社 | 細胞培養支援装置の作動プログラム、細胞培養支援装置、細胞培養支援装置の作動方法 |
JP2021043600A (ja) * | 2019-09-09 | 2021-03-18 | 株式会社カネカ | 学習モデルの生成方法、評価推定装置、学習モデル、及びコンピュータプログラム |
WO2021049044A1 (ja) * | 2019-09-13 | 2021-03-18 | エピストラ株式会社 | 培地製造方法、培地製造パラメータ決定方法、培地、およびプログラム |
Also Published As
Publication number | Publication date |
---|---|
JPWO2022264640A1 (ja) | 2022-12-22 |
EP4357446A1 (en) | 2024-04-24 |
CN117561326A (zh) | 2024-02-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Fan et al. | Characterizing transcriptional heterogeneity through pathway and gene set overdispersion analysis | |
JP2021532350A (ja) | 機械学習を適用して高スループットシステムにおけるマイクロコピー画像を分析するためのシステムおよび方法 | |
DeHaven et al. | Software techniques for enabling high-throughput analysis of metabolomic datasets | |
Chen et al. | Automated flow cytometric analysis across large numbers of samples and cell types | |
JPWO2014080447A1 (ja) | データ解析装置、データ解析プログラム | |
Emerson et al. | Multivariate data analysis in cell gene therapy manufacturing | |
US20220146987A1 (en) | Multivariate process chart to control a process to produce a chemical, pharmaceutical, biopharmaceutical and/or biological product | |
Huffman et al. | Prioritized single-cell proteomics reveals molecular and functional polarization across primary macrophages | |
JP2024071385A (ja) | レーザー力細胞学を利用した高度な生物物理的および生化学的細胞モニタリングおよび定量化 | |
US11500352B2 (en) | System and method for monitoring a production process | |
WO2022264640A1 (ja) | 状態推定システム及び状態推定方法 | |
Pargett et al. | Live‐Cell Imaging and Analysis with Multiple Genetically Encoded Reporters | |
CN110277139B (zh) | 一种基于互联网的微生物限度检查系统及方法 | |
Kamimura et al. | Mining of biological data I: identifying discriminating features via mean hypothesis testing | |
Goodman et al. | High-throughput, automated image processing for large-scale fluorescence microscopy experiments | |
US20220137579A1 (en) | Optimized process control | |
JP2023521757A (ja) | ラマンスペクトルに基づいて試料の特質を識別するためのモデルを決定するための遺伝的アルゴリズムの使用 | |
JP2022551117A (ja) | 撮像システムおよびその使用方法 | |
JP6977977B1 (ja) | 培地の異常検知装置及び異常検知方法 | |
JP7155281B2 (ja) | 細胞情報処理方法 | |
Steier et al. | Joint analysis of transcriptome and proteome measurements in single cells with totalVI: a practical guide | |
Ramm et al. | Systematic Preprocessing of Dielectric Spectroscopy Data and Estimating Viable Cell Densities | |
WO2023199234A1 (en) | System and method for impedance-based analysis of biological entities | |
Niederle et al. | VADA: a Data-Driven Simulator for Nanopore Sequencing | |
Genuth et al. | Ergodic patterns of cell state transitions underlie the reproducibility of embryonic development |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 22824636 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2023529621 Country of ref document: JP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 202280042792.1 Country of ref document: CN |
|
WWE | Wipo information: entry into national phase |
Ref document number: 18571492 Country of ref document: US |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2022824636 Country of ref document: EP |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
ENP | Entry into the national phase |
Ref document number: 2022824636 Country of ref document: EP Effective date: 20240118 |