CN116312741A - Culture medium formula optimization method, device, computer equipment and storage medium - Google Patents
Culture medium formula optimization method, device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN116312741A CN116312741A CN202310102448.3A CN202310102448A CN116312741A CN 116312741 A CN116312741 A CN 116312741A CN 202310102448 A CN202310102448 A CN 202310102448A CN 116312741 A CN116312741 A CN 116312741A
- Authority
- CN
- China
- Prior art keywords
- fitness
- candidate
- culture medium
- prediction
- index
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000002609 medium Substances 0.000 title claims abstract description 161
- 239000001963 growth medium Substances 0.000 title claims abstract description 151
- 238000000034 method Methods 0.000 title claims abstract description 49
- 238000005457 optimization Methods 0.000 title abstract description 18
- 238000000605 extraction Methods 0.000 claims abstract description 63
- 238000004590 computer program Methods 0.000 claims abstract description 29
- 238000009472 formulation Methods 0.000 claims description 80
- 239000000203 mixture Substances 0.000 claims description 80
- 230000004927 fusion Effects 0.000 claims description 41
- 238000012512 characterization method Methods 0.000 claims description 38
- 102000004169 proteins and genes Human genes 0.000 claims description 30
- 108090000623 proteins and genes Proteins 0.000 claims description 30
- 230000006978 adaptation Effects 0.000 abstract description 4
- 239000000306 component Substances 0.000 description 20
- 238000004422 calculation algorithm Methods 0.000 description 18
- 230000000694 effects Effects 0.000 description 15
- 230000006870 function Effects 0.000 description 10
- 238000004891 communication Methods 0.000 description 8
- 238000002474 experimental method Methods 0.000 description 8
- 230000003044 adaptive effect Effects 0.000 description 7
- 238000012545 processing Methods 0.000 description 7
- 238000012549 training Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 235000015097 nutrients Nutrition 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 239000013598 vector Substances 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 238000004113 cell culture Methods 0.000 description 3
- 238000012258 culturing Methods 0.000 description 3
- 238000010606 normalization Methods 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 230000003833 cell viability Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000013500 data storage Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000002068 genetic effect Effects 0.000 description 2
- 230000035772 mutation Effects 0.000 description 2
- 239000000243 solution Substances 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 241000196324 Embryophyta Species 0.000 description 1
- 241001465754 Metazoa Species 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 150000001413 amino acids Chemical class 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 239000007853 buffer solution Substances 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 239000000470 constituent Substances 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000013401 experimental design Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 229910021389 graphene Inorganic materials 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000012092 media component Substances 0.000 description 1
- 239000012533 medium component Substances 0.000 description 1
- 229910021645 metal ion Inorganic materials 0.000 description 1
- 244000005700 microbiome Species 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 235000013343 vitamin Nutrition 0.000 description 1
- 229940088594 vitamin Drugs 0.000 description 1
- 229930003231 vitamin Natural products 0.000 description 1
- 239000011782 vitamin Substances 0.000 description 1
- 150000003722 vitamin derivatives Chemical class 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
- G16B5/00—ICT specially adapted for modelling or simulations in systems biology, e.g. gene-regulatory networks, protein interaction networks or metabolic networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Physiology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biotechnology (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present application relates to a method, apparatus, computer device, storage medium and computer program product for medium recipe optimization. The method comprises the following steps: performing feature extraction based on formula features of the candidate culture medium formula and cell strain features corresponding to the target cell strain to obtain at least one extracted feature; determining index value prediction features respectively corresponding to the candidate culture medium formula under a plurality of fitness indexes based on at least one extraction feature; predicting index values by utilizing index value prediction features respectively corresponding to the adaptation indexes to obtain prediction index values respectively corresponding to the candidate culture medium formula under the adaptation indexes; and determining the target medium formula corresponding to the target cell strain based on the prediction index values respectively corresponding to the candidate medium formulas under each fitness index.
Description
Technical Field
The present application relates to the biotechnology field, and in particular, to a method, an apparatus, a computer device, a storage medium and a computer program product for optimizing a culture medium formulation.
Background
With the development of biotechnology, a large number of biological experiments are required, and most of biological experiments have been performed before using culture medium to perform cell culture, so as to obtain cells meeting the experimental requirements. The culture medium refers to a nutrient medium which is used for the growth and reproduction of microorganisms, plants or animals and is prepared by combining different nutrient substances. Because of the different nutrients required for culturing different cell lines, the corresponding culture medium formulation needs to be developed for different cell lines to improve the culture effect.
In the traditional technology, a culture medium formula optimization method based on machine learning is adopted, and modeling, prediction, optimization and recommendation are carried out on a single evaluation index corresponding to the culture medium formula, so that a recommended culture medium formula is obtained.
However, since modeling, prediction, optimization and recommendation can be performed only for a single index, interactivity between prediction tasks of different indexes is ignored, and the recommended medium formulation cannot be guaranteed to be an optimal formulation, resulting in inaccurate determination of the medium formulation.
Disclosure of Invention
Based on this, it is necessary to provide a method, an apparatus, a computer device, a computer-readable storage medium and a computer program product for optimizing a culture medium formulation, which are capable of improving accuracy, in view of the above-mentioned technical problems.
In a first aspect, the present application provides a method of optimizing a culture medium formulation. The method comprises the following steps: performing feature extraction based on formula features of the candidate culture medium formula and cell strain features corresponding to the target cell strain to obtain at least one extracted feature; determining index value prediction features respectively corresponding to the candidate culture medium formula under a plurality of fitness indexes based on the at least one extraction feature; predicting index values by utilizing index value prediction features respectively corresponding to the fitness indexes to obtain prediction index values respectively corresponding to the candidate culture medium formulas under the fitness indexes; and determining a target culture medium formula corresponding to the target cell strain based on the prediction index values respectively corresponding to the candidate culture medium formulas under the fitness indexes.
In a second aspect, the present application also provides a culture medium formulation optimizing apparatus. The device comprises: the characteristic extraction module is used for carrying out characteristic extraction based on the formula characteristics of the candidate culture medium formula and the cell strain characteristics corresponding to the target cell strain to obtain at least one extracted characteristic; the prediction feature determining module is used for determining index value prediction features of the candidate culture medium formula corresponding to a plurality of fitness indexes respectively based on the at least one extraction feature; the index value prediction module is used for predicting the index value by utilizing index value prediction characteristics respectively corresponding to the fitness indexes to obtain the prediction index values respectively corresponding to the candidate culture medium formula under the fitness indexes; and the formula determining module is used for determining the target culture medium formula corresponding to the target cell strain based on the prediction index values respectively corresponding to the candidate culture medium formulas under the fitness indexes.
In some embodiments, the feature extraction module is further to: fusing the formula characteristics of the candidate culture medium formula and the cell strain characteristics corresponding to the target cell strain to obtain fusion characteristics; respectively inputting the fusion features into a plurality of feature extraction networks of the fitness prediction model to extract features, so as to obtain a plurality of extracted features; the index value prediction module is further configured to: and inputting the corresponding index value prediction characteristics into the index value prediction network corresponding to the fitness prediction model for predicting the index value aiming at each fitness index, and obtaining the predicted index value of the candidate culture medium formula under the fitness index.
In some embodiments, the at least one extracted feature is a plurality of extracted features, and the fitness prediction model further includes a weight generation network to which the plurality of fitness indexes respectively correspond; the prediction feature determination module is further configured to: inputting the fusion characteristics into a corresponding weight generation network for weight prediction aiming at each fitness index in the plurality of fitness indexes to obtain a weight set corresponding to the fitness index; the weight set comprises weights corresponding to the extracted features respectively; and carrying out feature fusion on each extracted feature by utilizing the weight respectively corresponding to each extracted feature in the weight set of the fitness index to obtain index value prediction features corresponding to the fitness index.
In some embodiments, the plurality of fitness indicators includes protein expression levels and cell densities; the prediction feature determination module is further configured to: inputting the fusion characteristics into a weight generation network corresponding to the protein expression quantity to predict weights, so as to obtain a weight set corresponding to the protein expression quantity; and inputting the fusion characteristics into a weight generation network corresponding to the cell density to predict the weight, so as to obtain a weight set corresponding to the cell density.
In some embodiments, the candidate medium formulation belongs to a candidate medium formulation set; the determining the target culture medium formula corresponding to the target cell strain based on the prediction index values respectively corresponding to the candidate culture medium formulas under the fitness indexes comprises the following steps: counting the corresponding prediction index values of the candidate culture medium formula under each fitness index to obtain a fitness characterization value of the candidate culture medium formula; determining a new candidate medium formula set based on the fitness characterization value of each candidate medium formula and the candidate medium formula set; returning to each candidate medium formula in the candidate medium formula set, and performing feature extraction based on the formula features of the candidate medium formula and the cell strain features corresponding to the target cell strain to obtain at least one extracted feature until an iteration stop condition is met; and selecting a target culture medium formula corresponding to the target cell strain from the candidate culture medium formula set under the condition that the iteration stop condition is met.
In some embodiments, the determining a new candidate medium formulation set based on the fitness characterization value of each of the candidate medium formulations and the candidate medium formulation set comprises: and selecting a culture medium formula with a fitness characterization value larger than a characterization threshold value from the candidate culture medium formula set to obtain a new candidate culture medium formula set.
In a third aspect, the present application also provides a computer device. The computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to realize the steps in the culture medium formula optimization method.
In a fourth aspect, the present application also provides a computer-readable storage medium. The computer readable storage medium has stored thereon a computer program which, when executed by a processor, implements the steps of the above-described medium recipe optimization method.
In a fifth aspect, the present application also provides a computer program product. The computer program product comprises a computer program which, when executed by a processor, implements the steps of the above-described method for optimizing a medium recipe.
According to the culture medium formula optimizing method, the device, the computer equipment, the storage medium and the computer program product, the index value predicted values of the candidate culture medium formula corresponding to the plurality of fitness indexes are determined based on at least one extracted feature, namely, the index value predicted features corresponding to different fitness indexes are predicted simultaneously by sharing at least one extracted feature, and the interactivity among multiple tasks is considered, so that the predicted index values of the candidate culture medium formula corresponding to the candidate culture medium formula under the fitness indexes are more accurate, the accuracy of index value prediction is improved, and the accuracy of optimizing the culture medium formula is improved based on the target culture medium formula determined by the predicted index values.
Drawings
FIG. 1 is a diagram of an application environment for a method of optimizing a medium formulation in one embodiment;
FIG. 2 is a schematic flow chart of a method of optimizing a culture medium formulation in one embodiment;
FIG. 3 is a network architecture diagram of a multitasking algorithm model in one embodiment;
FIG. 4 is a schematic flow chart of a method for optimizing a culture medium formula in another embodiment;
FIG. 5A is a representation of the performance capabilities of a single-tasking algorithm model and a multi-tasking algorithm model in one embodiment;
FIG. 5B is a representation of the performance capabilities of different multi-tasking algorithm models in one embodiment;
FIG. 5C is a table showing the experimental results of the recommended medium formulation in one example;
FIG. 6 is a block diagram of a medium recipe optimizing apparatus in one embodiment;
FIG. 7 is an internal block diagram of a computer device in one embodiment;
fig. 8 is an internal structural view of a computer device in another embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.
The method for optimizing the culture medium formula can be applied to an application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The data storage system may store data that the server 104 needs to process. The data storage system may be integrated on the server 104 or may be located on a cloud or other network server.
Specifically, the terminal 102 transmits the target cell strain and the candidate medium formulation corresponding to the target cell strain to the server 104. The server 104 receives the target cell strain and the candidate culture medium formula corresponding to the target cell strain sent by the terminal 102, and performs feature extraction based on the formula features of the candidate culture medium formula and the cell strain features corresponding to the target cell strain to obtain at least one extracted feature; then determining index value prediction features respectively corresponding to the candidate culture medium formula under a plurality of fitness indexes based on at least one extraction feature; predicting index values by utilizing index value prediction features respectively corresponding to the adaptation indexes to obtain prediction index values respectively corresponding to the candidate culture medium formula under the adaptation indexes; and determining a target culture medium formula corresponding to the target cell strain based on the prediction index values respectively corresponding to the candidate culture medium formulas under each fitness index, wherein one or more target culture medium formulas can be used. After determining the target medium formulation corresponding to the target cell line, the server 104 may send the target medium formulation and the prediction index values corresponding to the target medium formulation under each fitness index to the terminal 102. The terminal 102 receives the target medium formula sent by the server 104 and the prediction index values respectively corresponding to the target medium formula under each fitness index, so that an experiment operator can perform a biological experiment according to the target medium formula, and the validity and the recommended accuracy of the target medium formula are verified based on the experimental result.
The terminal 102 may be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices, among others. The portable wearable device may be a smart watch, smart bracelet, headset, or the like. The server 104 may be implemented as a stand-alone server or as a server cluster of multiple servers.
In some embodiments, as shown in fig. 2, a method for optimizing a culture medium formulation is provided, which may be performed by a terminal or a server, and may also be performed by the terminal and the server together, where the method is applied to the server 104 in fig. 1, and is illustrated as an example, and includes the following steps:
Wherein, the culture medium is a nutrient medium prepared by combining different nutrient substances and is used for cell culture. The formula of the culture medium comprises the components of the culture medium, and the components of the culture medium comprise at least one of amino acid, metal ion, vitamin or buffer system, but not limited to, according to the respective corresponding adding proportion. The candidate culture medium is used for culturing target cell strains, the candidate culture medium formula belongs to a candidate culture medium formula set, the candidate culture medium formula set comprises a plurality of candidate culture medium formulas, the candidate culture medium formula set is obtained based on an original culture medium formula set, and the original culture medium formula set consists of culture medium formulas verified by biological experiments.
The formula characteristics are characteristics obtained based on the candidate culture medium formula, and represent the constituent components of the candidate culture medium formula and the respective corresponding addition proportion of each component. The cell line of interest is a cell line to be subjected to cell culture and the characteristics of the cell line are characteristics obtained based on the cell line of interest, including but not limited to cell type, cell molecular weight. The extracted features are obtained by feature extraction based on the formula features of the candidate medium formula and the cell strain features corresponding to the target cell strain, and at least one of the extracted features comprises one or more than one of the extracted features.
Specifically, the server acquires a target cell strain and a candidate medium formula set corresponding to the target cell strain, determines formula characteristics of the candidate medium formula and cell strain characteristics corresponding to the target cell strain for each candidate medium formula in the candidate medium formula set, and performs characteristic fusion on the formula characteristics of the candidate medium formula and the cell strain characteristics corresponding to the target cell strain to obtain fusion characteristics. And then the server performs feature extraction based on the fusion features to obtain at least one extracted feature, for example, the fusion features can be input into at least one feature extraction network to obtain at least one corresponding extracted feature, and the feature extraction network is used for feature extraction. The fusion characteristics are obtained by performing characteristic fusion based on the formula characteristics of the candidate culture medium formula and the cell strain characteristics corresponding to the target cell strain, for example, the formula characteristics and the cell strain characteristics can be subjected to matrix splicing treatment to obtain the fusion characteristics.
In some embodiments, the server may perform an experimental design on the initial addition amounts of the components and randomly generate a plurality of original medium formulations with the aim of the culture effect of the selected medium formulation according to the usage range of the components of the medium after verification of the biological experiment. And then, performing biological experiments on each original culture medium formula and specific cell types to obtain the culture effect of each original culture medium formula on the cell lines of the cell types, and forming a formula database based on the original culture medium formula, the corresponding cell line characteristics and the culture effect of the culture medium formula. Each row of data in the recipe database consists of the following three parts:the medium components (x) 11 ,x 12 ,···,x 1n ) Characteristics of cell lines (x) 21 ,x 22 ,···,x 2m ) Culture Effect (y) 1 ,y 2 ,···,y k ). Wherein x is 1i Is the ith component of the culture medium formula and can be used as a dense feature in the model training and prediction process; x is x 2i The i-th characteristic in the cell strain characteristics can be used as a sparse characteristic in the model training and prediction process; y is i Is the ith culture effect of the original culture medium formula, namely the index value corresponding to the ith fitness index.
In some embodiments, the server obtains a recipe database, performs feature preprocessing on data in the recipe database, including normalization of dense features, i.e., media components, and emmbedding of sparse features, i.e., cell line features. For example, the normalization process may be: first, the mean and standard deviation of the components in the medium formulation are obtained, for each data x in the dense features 1i Subtracting the corresponding mean value and dividing by the standard deviation, so that the processed data has a fixed mean value and standard deviation, and further, the normalized data with dense features are in the same data dimension. The ebadd operation refers to: and (3) carrying out digital processing on discontinuous numbers or texts, so that sparse features are uniformly converted into numerical vectors. The normalization process may be calculated according to the following formula:
x * ＝(x-μ)/σ；
wherein x is the value of a component of a certain medium recipe of the recipe database; μ is the average of all values of the component in the recipe database; σ is the standard deviation of all values of the component in the recipe database; x is x * Is a value obtained by normalizing the value of the component.
And 204, determining index value prediction features of the candidate culture medium formula corresponding to the plurality of fitness indexes respectively based on at least one extracted feature.
Wherein, the fitness index is used for evaluating the culture effect of the culture medium corresponding to the culture medium formula, and the fitness index comprises at least one of cell density, cell activity rate or protein expression quantity but not limited to. The index value prediction feature is obtained based on a plurality of extraction features corresponding to the candidate medium formula.
Specifically, for each fitness index of the plurality of fitness indexes, the server determines a weight set corresponding to the fitness index, where the weight set includes weights corresponding to the extracted features. The server can use weights corresponding to the extracted features in the weight set of the fitness indexes to perform feature fusion on the extracted features of the candidate medium formula to obtain index value prediction features corresponding to the fitness indexes, and then obtain index value prediction features corresponding to the candidate medium formula under a plurality of fitness indexes.
And 206, predicting the index value by utilizing index value prediction features respectively corresponding to the fitness indexes to obtain the prediction index values respectively corresponding to the candidate culture medium formula under the fitness indexes.
The predictive index value is an index value of an fitness index predicted by an index value prediction feature, and for example, the fitness index is a protein expression level, and the corresponding predictive index value is 511.
Specifically, for the candidate medium recipe, the server predicts the index value by using index value prediction features corresponding to each fitness index respectively, for example, for each fitness index, the server may input the corresponding index value prediction feature into a corresponding index value prediction network in the fitness prediction model, perform index value prediction, obtain a prediction index value of the candidate medium recipe under the fitness index, and then obtain a prediction index value corresponding to each fitness index of the candidate medium recipe respectively. The fitness prediction model is used for predicting the prediction index values of the culture medium formula corresponding to the fitness indexes respectively.
The target culture medium formula is determined from candidate culture medium formulas, and the culture medium corresponding to the target culture medium formula is used for culturing target cell strains.
Specifically, the server may count the prediction index values corresponding to the candidate medium formulas at the fitness indexes respectively to obtain fitness characterization values of the candidate medium formulas, where the fitness characterization values are used to characterize the culture effect of the medium obtained based on the candidate medium formulas, and the higher the fitness characterization values, the better the culture effect. And the server determines a target medium formula corresponding to the target cell strain based on the fitness characterization value of the candidate medium formula.
In some embodiments, the server may use a global optimization algorithm or a heuristic algorithm to search in an addition proportion space corresponding to each component in the candidate medium formulation based on the fitness characterization value of the candidate medium formulation, and iteratively update the candidate medium formulation set, thereby selecting a target medium formulation corresponding to the target cell strain. The addition ratio space is determined based on the corresponding addition range of the components of the medium formulation.
In some examples, as shown in fig. 5C, the results of comparison of the culture effect of the medium obtained based on the recommended formulation, i.e., the target medium formulation, with the culture medium of control group 1 and the culture medium of control group 2 in the biological experiment are shown. Compared with the control group 1, the cell density of the recommended formula is improved by 143 percent, and the protein expression quantity is improved by 84 percent; compared with the control group 2, the cell density of the recommended formula is improved by 104%, and the protein expression quantity is improved by 66%. Wherein, control group 1 and control group 2 are both commercial media.
In some embodiments, as shown in fig. 4, a flow chart of a method for optimizing a culture medium formulation is shown, a server obtains a target cell strain and a candidate culture medium formulation, vectorization processing can be performed on cell strain characteristics of the target cell strain, standardization processing is performed on culture medium components of the candidate culture medium formulation to obtain formulation characteristics, and then splicing processing is performed on the vectorized cell strain characteristics and the formulation characteristics to obtain a spliced characteristic vector. And the server inputs the spliced feature vector into a plurality of feature extraction networks to extract the features, so as to obtain a plurality of extracted features. Under the condition that the fitness index comprises a fitness index A and a fitness index B, a weight set of a task A can be generated by utilizing a weight generation A network, a weight set of a task B can be generated by utilizing a weight generation B network, then the input characteristics of the task A network are obtained by carrying out weighted calculation on a plurality of extracted characteristics based on the weight set of the task A, and the input characteristics of the task B network are obtained by carrying out weighted calculation on a plurality of extracted characteristics based on the weight set of the task B. The output A of the task A network and the output B of the task B network are respectively corresponding prediction index values of the candidate culture medium formula under the fitness index A and the fitness index B. And then the server can recommend the multitask culture medium formula by utilizing the prediction index values respectively corresponding to the candidate culture medium formulas under the fitness index A and the fitness index B to obtain the target culture medium formula corresponding to the target cell strain.
According to the culture medium formula optimizing method, the index value predicted values of the candidate culture medium formula corresponding to the plurality of fitness indexes are determined based on at least one extracted feature, namely, the index value predicted features corresponding to different fitness indexes are predicted simultaneously by sharing at least one extracted feature and adopting a multi-task learning strategy, and the predicted index values of the candidate culture medium formula corresponding to the plurality of fitness indexes are more accurate due to the fact that the interactivity among the plurality of tasks is considered, the accuracy of index value prediction is improved, and therefore the accuracy of optimizing the culture medium formula is improved according to the target culture medium formula determined based on the predicted index values.
In some embodiments, step 202 comprises: fusing the formula characteristics of the candidate culture medium formula and the cell strain characteristics corresponding to the target cell strain to obtain fusion characteristics; respectively inputting the fusion features into a plurality of feature extraction networks of the fitness prediction model to extract features, so as to obtain a plurality of extracted features; step 206 comprises: and inputting the corresponding index value prediction characteristics into a corresponding index value prediction network in the fitness prediction model for predicting the index value aiming at each fitness index to obtain the predicted index value of the candidate culture medium formula under the fitness index.
The fitness prediction model is used for predicting the prediction index value of the candidate culture medium formula under each fitness index, and comprises a plurality of feature extraction networks and index value prediction networks corresponding to each fitness index respectively. The characteristic extraction network is used for extracting characteristics, and the index value prediction network is used for predicting index values of the fitness indexes.
Specifically, for each candidate medium formula in the candidate medium formula set, the server may fuse the formula feature of the candidate medium formula with the cell strain feature corresponding to the target cell strain to obtain a fused feature, and then input the fused feature into a plurality of feature extraction networks of the fitness prediction model to perform feature extraction, for example, as shown in fig. 3, the fitness prediction model includes 3 feature extraction networks, which are feature extraction network 0, feature extraction network 1, and feature extraction network 2, respectively, and input the fused feature into the 3 feature extraction networks to perform feature extraction, so as to obtain extraction feature 0, extraction feature 1, and extraction feature 2 output by the feature extraction network 0, the feature extraction network 1, and the feature extraction network 2, respectively. And then, determining index value prediction features corresponding to the fitness indexes based on the plurality of extraction features for each fitness index, inputting the corresponding index value prediction features into a corresponding index value prediction network in the fitness prediction model to perform index value prediction, and obtaining the prediction index value of the candidate culture medium formula under the fitness indexes.
In some embodiments, the fitness prediction model may be a multi-task prediction model, for example, as shown in fig. 3, which illustrates a network structure of the fitness prediction model, including 3 feature extraction networks, a weight generation network corresponding to task a and task B, and an index value prediction network, where the weight generation network is used to generate a weight set corresponding to each fitness index. The fitness prediction model is built on a MMoE (Multi-gate texture-of-expertise) network structure widely used in the field of Multi-task learning, wherein the shared bottom network of MMoE is the 3 feature extraction networks in fig. 3. As the correlation among different prediction tasks is considered in the multi-task prediction model, the performance capability of the model is improved, namely the prediction accuracy is improved, and as shown in fig. 5A, the comparison result of the single-task prediction model and the multi-task prediction model is shown, wherein MMoE_MTL, MMoE_Y2 and MMoE_Y3 are models for simultaneously predicting cell density and protein expression quantity, only predicting cell density and only predicting protein expression quantity, and besides different prediction targets, the three models comprise input characteristics, network structures, super parameters, training set test set division proportion and other conditions are consistent.
In some embodiments, as shown in fig. 5B, the comparison results of fitness prediction models corresponding to different algorithms are shown, where mmoe_mtl, mmoe_ Single, catboost (Categorical Features + Gradient Boosting, classification feature+gradient lifting), SVR (Support Vector Regression, support vector regression model) are MMoE model for predicting cell density and protein expression level simultaneously, MMoE model for predicting cell density or protein expression level only, catboost model for predicting cell density or protein expression level only, and SVR model for predicting cell density or protein expression level only, respectively. Except for the difference of algorithm models, the input characteristics of the four algorithm models, the dividing proportion of the training set test set and other conditions are consistent.
In some embodiments, the server may perform multi-task joint training on the fitness prediction model to be trained, and by adjusting the hyper-parameters of the fitness prediction model, a trained fitness prediction model is obtained under the condition that the joint loss function converges. The joint loss function is obtained based on the loss function corresponding to each index value prediction network in the fitness prediction model, and the super parameters of the fitness prediction model comprise: the number of feature extraction networks, the number of layers and units per feature extraction network, the number of layers and units per index value prediction network, the Dropout (discard) ratio, and the L2 regularized coefficient. For example, assuming that the fitness prediction model includes an index value prediction a network and an index value prediction B network, the MSE loss function may be used as a first loss function and a second loss function corresponding to the index value prediction a network and the index value prediction B network, respectively, and the joint loss function may be obtained based on the first loss function and the second loss function. The super parameters of the fitness prediction model may specifically be: the number of the feature extraction networks is 3, the number of layers and units of each feature extraction network is 64 x 16, the number of layers and units of each index value prediction network is 64 x 8, the ratio of Dropout is 0.1, and the regularized coefficient of L2 is 1e-5.
In some embodiments, the server adopts python language and Keras framework to build a deep learning model, namely an fitness prediction model, and the culture effects of simultaneous prediction of cell viability, cell density, protein expression and the like can be achieved through inputting the formula characteristics of the culture medium formula and the cell strain characteristics of the target cell strain into the trained fitness prediction model. And the adaptability prediction model to be trained can be trained by adopting a gradient descent method, which comprises the following steps: SGD (Stochastic gradient descent, random gradient descent algorithm), momentum (Momentum algorithm), adagard (Adaptive Gradient, adaptive gradient algorithm), RMSprop (Root Mean Square Propagation, adaptive learning rate algorithm), adam (Adaptive Momentum, adaptive Momentum algorithm). For example, adam's algorithm may be used to set the learning rate to 0.001.
In this embodiment, the fitness prediction model includes a plurality of feature extraction networks and index value prediction networks corresponding to the fitness indexes respectively, that is, different prediction tasks share the same algorithm model, and prediction accuracy of each index value prediction network is effectively improved through information sharing between tasks.
In some embodiments, the at least one extracted feature is a plurality of extracted features, and the fitness prediction model further includes a weight generation network to which the plurality of fitness indicators respectively correspond; the determining index value prediction features of the candidate medium formula corresponding to the plurality of fitness indexes respectively based on the at least one extraction feature comprises: inputting the fusion characteristic into a corresponding weight generation network for weight prediction aiming at each fitness index in a plurality of fitness indexes to obtain a weight set corresponding to the fitness index; the weight set comprises weights corresponding to the extracted features respectively; and carrying out feature fusion on each extracted feature by utilizing weights respectively corresponding to each extracted feature in the weight set of the fitness index to obtain index value prediction features corresponding to the fitness index.
The fitness prediction model further comprises a weight generation network corresponding to the fitness index values respectively. That is, the index value prediction network is in one-to-one correspondence with the weight generation network, for example, for the fitness index a, the index value prediction network is used for performing index value prediction of the fitness index a, and the weight generation network is used for determining a weight set corresponding to the fitness index a, that is, the weight generation network may learn different combination modes of the plurality of feature extraction networks for corresponding prediction tasks, and perform adaptive weighting on the output of the feature extraction network. The weight set comprises weights corresponding to the extracted features respectively, and the weights are used for representing the importance of the extracted features.
Specifically, for each fitness index value in the plurality of fitness index values, the server inputs the fusion feature into the corresponding weight generating network to obtain a weight set corresponding to the fitness index, for example, as shown in fig. 3, the fusion feature may be input into the weight generating a network and the weight generating B network to obtain a weight set corresponding to the fitness index a and a weight set corresponding to the fitness index B. And then the server performs feature fusion on each extracted feature by using the weights corresponding to each extracted feature in the weight set of the fitness index, namely, performs weighted calculation on each extracted feature to obtain index value prediction features corresponding to the fitness index.
In some embodiments, the fitness prediction model may be expressed by the following formula:
y k ＝h k (f k (x))；
wherein y is k Representing the output result of the kth prediction task, h k (x) Index value prediction network f corresponding to kth prediction task k (x) Representing the input of an index value prediction network, which is obtained by weighting and summing the output results of n feature extraction networks, and the weight coefficientAnd generating by a weight generating network corresponding to the kth prediction task. In addition, in the fitness prediction model, the number of weight generation networks is the same as the number of prediction tasks. f (f) i (x) Representing the output of the ith feature extraction network. Both the ownership generation network and the indicator value prediction network are implemented by the same MLP with a RELU activation function. And the weight g generated by the weight generating network corresponding to the kth prediction task k (x) Is obtained by simple linear transformation of the input features and further by softmax. Specifically, the method can be expressed as:
wherein Wgk E and Rn d are matrixes obtained by training the model; n is the number of feature extraction networks; d is the feature dimension.
In this embodiment, since one weight generating network corresponds to one fitness index, the weight generating network may generate a weight set of the corresponding fitness index, so that the weight generating network corresponding to different fitness indexes may learn different combination modes of extracted features, and thus may capture the correlation between tasks, thereby improving the accuracy of index value prediction.
In some embodiments, the plurality of fitness indicators includes protein expression levels and cell densities; for each fitness index of the plurality of fitness indexes, inputting the fusion feature into a corresponding weight generation network to perform weight prediction, and obtaining a weight set corresponding to the fitness index comprises: inputting the fusion characteristics into a weight generation network corresponding to the protein expression quantity to predict the weight, so as to obtain a weight set corresponding to the protein expression quantity; and inputting the fusion characteristics into a weight generation network corresponding to the cell density to predict the weight, so as to obtain a weight set corresponding to the cell density.
Wherein the plurality of fitness indicators include, but are not limited to, cell density, protein expression level, and cell viability.
Specifically, the server may input the fusion feature into a weight generation network corresponding to the protein expression amount to perform weight prediction, so as to obtain a weight set corresponding to the protein expression amount; and inputting the fusion characteristics into a weight generation network corresponding to the cell density to predict the weight, so as to obtain a weight set corresponding to the cell density. And then the server can perform feature fusion on the plurality of extracted features by using a weight set corresponding to the protein expression quantity to obtain index value prediction features corresponding to the protein expression quantity, and perform feature fusion on the plurality of extracted features by using a weight set corresponding to the cell density to obtain index value prediction features corresponding to the cell density.
In the embodiment, the fitness prediction model considers the correlation between the protein expression amount and the cell density, can simultaneously realize the prediction of the protein expression amount and the cell density,
in some embodiments, the candidate medium formulation belongs to a candidate medium formulation set; step 208 includes: counting the corresponding prediction index values of the candidate culture medium formula under each fitness index to obtain a fitness characterization value of the candidate culture medium formula; determining a new candidate medium formula set based on the fitness characterization value of each candidate medium formula and the candidate medium formula set; returning to the step of extracting features based on the formula features of the candidate medium formulas and the cell strain features corresponding to the target cell strain for each candidate medium formula in the candidate medium formula set to obtain at least one extracted feature until the iteration stop condition is met; and selecting a target medium formula corresponding to the target cell strain from the candidate medium formula set under the condition that the iteration stop condition is met.
The fitness characterization value is obtained by counting the corresponding prediction index values under each fitness index and is used for characterizing the culture effect of the culture medium obtained based on the candidate culture medium formula, and the higher the fitness characterization value is, the better the culture effect is. The iteration stop condition may be, for example, that a preset number of iterations is reached, which is preset.
Specifically, for each candidate medium recipe in the candidate medium recipe set, the server may perform adaptive weighted calculation on the prediction index values corresponding to the multiple fitness indexes respectively, to obtain a fitness characterization value corresponding to the candidate medium recipe. And then the server searches the culture medium formula by utilizing a genetic algorithm, determines a new candidate culture medium formula set, and returns to the step of extracting the characteristics based on the formula characteristics of the candidate culture medium formula and the characteristics of the cell strain corresponding to the target cell strain to obtain at least one extracted characteristic until the iteration stop condition is met. The server can select a target medium formula corresponding to the target cell strain from the candidate medium formula set under the condition that the iteration stop condition is met. For example, the candidate medium formulations in the candidate medium formulation set may be sorted in order of the largest prediction index value corresponding to a certain fitness index, and then the first n candidate medium formulations may be used as the target medium formulation.
In some embodiments, in searching the media formulation and obtaining the new candidate media formulation set by the server using the genetic algorithm, the server may target to find an optimal fitness characterization value based on the fitness characterization value of each candidate media formulation and the candidate media formulation set, and continuously optimize the components of the candidate media formulation under the condition that each component in the candidate media formulation is within a given range and the total mass of the media formulation is less than the mass threshold, for example, a plurality of media formulations with higher fitness characterization values may be generated by random selection, crossover and mutation operations, and a new candidate media formulation set may be generated. The quality threshold may be preset or dynamically changed.
In this embodiment, a new candidate medium formulation is determined based on the fitness characterization value of the candidate medium formulation and the candidate medium formulation set until the iteration stop condition is satisfied, so that the medium formulation can be continuously optimized, the accuracy of determining the target medium formulation is improved, and the culture effect of the medium obtained based on the target medium formulation can be improved.
In some embodiments, determining a new candidate medium formulation set based on the fitness characterization value of each candidate medium formulation and the candidate medium formulation set comprises: and selecting a culture medium formula with a fitness characterization value larger than a characterization threshold value from the candidate culture medium formula set to obtain a new candidate culture medium formula set.
The characterization threshold may be a threshold of a preset fitness characterization value.
Specifically, the server may select a medium formula with a fitness characterization value greater than a characterization threshold from the candidate medium set, to obtain a new candidate medium formula set. The server can also perform operations such as random selection, crossover, mutation and the like based on the candidate medium formula set to obtain a new candidate medium formula set.
In this embodiment, by selecting a medium recipe with a fitness characterization value greater than a characterization threshold, the candidate medium recipe in the new candidate medium recipe set obtained each time is better than the candidate medium recipe before optimization, and finally, the target medium recipe can be obtained through continuous iterative updating.
It should be understood that, although the steps in the flowcharts related to the embodiments described above are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in the flowcharts described in the above embodiments may include a plurality of steps or a plurality of stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of the steps or stages is not necessarily performed sequentially, but may be performed alternately or alternately with at least some of the other steps or stages.
Based on the same inventive concept, the embodiment of the application also provides a culture medium formula optimizing device for realizing the culture medium formula optimizing method. The implementation of the solution provided by the device is similar to that described in the above method, so specific limitations in one or more embodiments of the medium formulation optimization device provided below can be found in the limitations of the medium formulation optimization method described above, and are not repeated here.
In some embodiments, as shown in fig. 6, there is provided a medium formulation optimizing apparatus comprising: a feature extraction module 602, a predicted feature determination module 604, an index value prediction module 606, and a recipe determination module, wherein:
the feature extraction module 602 is configured to perform feature extraction based on the formula feature of the candidate medium formula and the cell strain feature corresponding to the target cell strain, so as to obtain at least one extracted feature.
The prediction feature determining module 604 is configured to determine, based on at least one extracted feature, index value prediction features that respectively correspond to the candidate medium formula under a plurality of fitness indexes.
The index value prediction module 606 is configured to predict an index value by using index value prediction features corresponding to each fitness index, so as to obtain a predicted index value corresponding to each fitness index of the candidate medium formula.
The formula determining module 608 is configured to determine a target medium formula corresponding to the target cell strain based on the prediction index values corresponding to the candidate medium formulas under the fitness indexes.
In some embodiments, the feature extraction module is further to: fusing the formula characteristics of the candidate culture medium formula and the cell strain characteristics corresponding to the target cell strain to obtain fusion characteristics; respectively inputting the fusion features into a plurality of feature extraction networks of the fitness prediction model to extract features, so as to obtain a plurality of extracted features; the index value prediction module is further configured to: and inputting the corresponding index value prediction characteristics into a corresponding index value prediction network in the fitness prediction model for predicting the index value aiming at each fitness index to obtain the predicted index value of the candidate culture medium formula under the fitness index.
In some embodiments, the at least one extracted feature is a plurality of extracted features, and the fitness prediction model further includes a weight generation network to which the plurality of fitness indicators respectively correspond; the prediction feature determination module is further configured to: inputting the fusion characteristic into a corresponding weight generation network for weight prediction aiming at each fitness index in a plurality of fitness indexes to obtain a weight set corresponding to the fitness index; the weight set comprises weights corresponding to the extracted features respectively; and carrying out feature fusion on each extracted feature by utilizing weights respectively corresponding to each extracted feature in the weight set of the fitness index to obtain index value prediction features corresponding to the fitness index.
In some embodiments, the plurality of fitness indicators includes protein expression levels and cell densities; the prediction feature determination module is further configured to: inputting the fusion characteristics into a weight generation network corresponding to the protein expression quantity to predict the weight, so as to obtain a weight set corresponding to the protein expression quantity; and inputting the fusion characteristics into a weight generation network corresponding to the cell density to predict the weight, so as to obtain a weight set corresponding to the cell density.
In some embodiments, the candidate medium formulation belongs to a candidate medium formulation set; based on the prediction index values respectively corresponding to the candidate culture medium formulas under the fitness indexes, the method for determining the target culture medium formula corresponding to the target cell strain comprises the following steps: counting the corresponding prediction index values of the candidate culture medium formula under each fitness index to obtain a fitness characterization value of the candidate culture medium formula; determining a new candidate medium formula set based on the fitness characterization value of each candidate medium formula and the candidate medium formula set; returning to the step of extracting features based on the formula features of the candidate medium formulas and the cell strain features corresponding to the target cell strain for each candidate medium formula in the candidate medium formula set to obtain at least one extracted feature until the iteration stop condition is met; and selecting a target medium formula corresponding to the target cell strain from the candidate medium formula set under the condition that the iteration stop condition is met.
In some embodiments, determining a new candidate medium formulation set based on the fitness characterization value of each candidate medium formulation and the candidate medium formulation set comprises: and selecting a culture medium formula with a fitness characterization value larger than a characterization threshold value from the candidate culture medium formula set to obtain a new candidate culture medium formula set.
The various modules in the above described media formulation optimization apparatus may be implemented in whole or in part by software, hardware, and combinations thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In some embodiments, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 7. The computer device includes a processor, a memory, an Input/Output interface (I/O) and a communication interface. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface is connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is used for storing relevant data related to the culture medium formula optimization method. The input/output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by the processor to implement a method of medium formulation optimization.
In some embodiments, a computer device is provided, which may be a terminal, and the internal structure of which may be as shown in fig. 8. The computer device includes a processor, a memory, an input/output interface, a communication interface, a display unit, and an input means. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input/output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless mode can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. The computer program is executed by the processor to implement a method of medium formulation optimization. The display unit of the computer device is used for forming a visual picture, and can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, can also be a key, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.
It will be appreciated by those skilled in the art that the structures shown in fig. 7 and 8 are block diagrams of only some of the structures associated with the present application and are not intended to limit the computer device to which the present application may be applied, and that a particular computer device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.
In some embodiments, a computer device is provided comprising a memory having a computer program stored therein and a processor that when executing the computer program performs the steps of the above-described method of optimizing a medium formulation.
In some embodiments, a computer readable storage medium is provided having a computer program stored thereon, which when executed by a processor, implements the steps of the above-described method of medium recipe optimization.
In some embodiments, a computer program product is provided comprising a computer program which, when executed by a processor, implements the steps of the above-described method of medium formulation optimization.
It should be noted that, the user information (including, but not limited to, user equipment information, user personal information, etc.) and the data (including, but not limited to, data for analysis, stored data, presented data, etc.) referred to in the present application are information and data authorized by the user or sufficiently authorized by each party, and the collection, use and processing of the related data are required to comply with the related laws and regulations and standards of the related countries and regions.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the various embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magnetic random access Memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric Memory (Ferroelectric Random Access Memory, FRAM), phase change Memory (Phase Change Memory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), and the like. The databases referred to in the various embodiments provided herein may include at least one of relational databases and non-relational databases. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic units, quantum computing-based data processing logic units, etc., without being limited thereto.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The above examples only represent a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the present application. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application shall be subject to the appended claims.
Claims (10)
1. A method of optimizing a culture medium formulation, the method comprising:
performing feature extraction based on formula features of the candidate culture medium formula and cell strain features corresponding to the target cell strain to obtain at least one extracted feature;
determining index value prediction features respectively corresponding to the candidate culture medium formula under a plurality of fitness indexes based on the at least one extraction feature;
Predicting index values by utilizing index value prediction features respectively corresponding to the fitness indexes to obtain prediction index values respectively corresponding to the candidate culture medium formulas under the fitness indexes;
and determining a target culture medium formula corresponding to the target cell strain based on the prediction index values respectively corresponding to the candidate culture medium formulas under the fitness indexes.
2. The method of claim 1, wherein the performing feature extraction based on the formula feature of the candidate medium formula and the cell strain feature corresponding to the target cell strain to obtain at least one extracted feature comprises:
fusing the formula characteristics of the candidate culture medium formula and the cell strain characteristics corresponding to the target cell strain to obtain fusion characteristics;
respectively inputting the fusion features into a plurality of feature extraction networks of the fitness prediction model to extract features, so as to obtain a plurality of extracted features;
the predicting the index value by using the index value predicting features respectively corresponding to the fitness indexes, and obtaining the predicting index value respectively corresponding to the candidate culture medium formula under the fitness indexes comprises the following steps:
and inputting the corresponding index value prediction characteristics into the index value prediction network corresponding to the fitness prediction model for predicting the index value aiming at each fitness index, and obtaining the predicted index value of the candidate culture medium formula under the fitness index.
3. The method of claim 2, wherein the at least one extracted feature is a plurality of extracted features, and the fitness prediction model further comprises a weight generation network to which the plurality of fitness indices respectively correspond;
the determining, based on the at least one extracted feature, index value prediction features of the candidate medium formula corresponding to a plurality of fitness indexes respectively includes:
inputting the fusion characteristics into a corresponding weight generation network for weight prediction aiming at each fitness index in the plurality of fitness indexes to obtain a weight set corresponding to the fitness index; the weight set comprises weights corresponding to the extracted features respectively;
and carrying out feature fusion on each extracted feature by utilizing the weight respectively corresponding to each extracted feature in the weight set of the fitness index to obtain index value prediction features corresponding to the fitness index.
4. The method of claim 3, wherein the plurality of fitness indicators comprises protein expression levels and cell densities;
inputting the fusion feature into a corresponding weight generation network for weight prediction aiming at each fitness index of the plurality of fitness indexes, and obtaining a weight set corresponding to the fitness index comprises:
Inputting the fusion characteristics into a weight generation network corresponding to the protein expression quantity to predict weights, so as to obtain a weight set corresponding to the protein expression quantity;
and inputting the fusion characteristics into a weight generation network corresponding to the cell density to predict the weight, so as to obtain a weight set corresponding to the cell density.
5. The method of claim 1, wherein the candidate medium formulation belongs to a candidate medium formulation set;
the determining the target culture medium formula corresponding to the target cell strain based on the prediction index values respectively corresponding to the candidate culture medium formulas under the fitness indexes comprises the following steps:
counting the corresponding prediction index values of the candidate culture medium formula under each fitness index to obtain a fitness characterization value of the candidate culture medium formula;
determining a new candidate medium formula set based on the fitness characterization value of each candidate medium formula and the candidate medium formula set;
returning to the formula characteristics based on the candidate medium formulas and the cell strain characteristics corresponding to the target cell strain for characteristic extraction aiming at each candidate medium formula in the candidate medium formula set to obtain at least one extracted characteristic until the iteration stop condition is met;
And selecting a target culture medium formula corresponding to the target cell strain from the candidate culture medium formula set under the condition that the iteration stop condition is met.
6. The method of claim 5, wherein the determining a new candidate medium formulation set based on the fitness characterization value of each candidate medium formulation and the candidate medium formulation set comprises:
and selecting a culture medium formula with a fitness characterization value larger than a characterization threshold value from the candidate culture medium formula set to obtain a new candidate culture medium formula set.
7. A media formulation optimisation apparatus, the apparatus comprising:
the characteristic extraction module is used for carrying out characteristic extraction based on the formula characteristics of the candidate culture medium formula and the cell strain characteristics corresponding to the target cell strain to obtain at least one extracted characteristic;
the prediction feature determining module is used for determining index value prediction features of the candidate culture medium formula corresponding to a plurality of fitness indexes respectively based on the at least one extraction feature;
the index value prediction module is used for predicting the index value by utilizing index value prediction characteristics respectively corresponding to the fitness indexes to obtain the prediction index values respectively corresponding to the candidate culture medium formula under the fitness indexes;
And the formula determining module is used for determining the target culture medium formula corresponding to the target cell strain based on the prediction index values respectively corresponding to the candidate culture medium formulas under the fitness indexes.
8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 6 when the computer program is executed.
9. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 6.
10. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310102448.3A CN116312741A (en) | 2023-01-28 | 2023-01-28 | Culture medium formula optimization method, device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310102448.3A CN116312741A (en) | 2023-01-28 | 2023-01-28 | Culture medium formula optimization method, device, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116312741A true CN116312741A (en) | 2023-06-23 |
Family
ID=86819505
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310102448.3A Pending CN116312741A (en) | 2023-01-28 | 2023-01-28 | Culture medium formula optimization method, device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116312741A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111291266A (en) * | 2020-02-13 | 2020-06-16 | 腾讯科技（北京）有限公司 | Artificial intelligence based recommendation method and device, electronic equipment and storage medium |
| CN114121163A (en) * | 2021-11-30 | 2022-03-01 | 深圳太力生物技术有限责任公司 | Culture medium prediction system based on ensemble learning, training and culture medium prediction method |
| CN114611386A (en) * | 2022-02-28 | 2022-06-10 | 深圳太力生物技术有限责任公司 | Culture medium mixing proportion optimization method, device, equipment and medium |
| CN114817692A (en) * | 2021-01-20 | 2022-07-29 | 腾讯科技（深圳）有限公司 | Method, device and equipment for determining recommended object and computer storage medium |
-
2023
- 2023-01-28 CN CN202310102448.3A patent/CN116312741A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111291266A (en) * | 2020-02-13 | 2020-06-16 | 腾讯科技（北京）有限公司 | Artificial intelligence based recommendation method and device, electronic equipment and storage medium |
| CN114817692A (en) * | 2021-01-20 | 2022-07-29 | 腾讯科技（深圳）有限公司 | Method, device and equipment for determining recommended object and computer storage medium |
| CN114121163A (en) * | 2021-11-30 | 2022-03-01 | 深圳太力生物技术有限责任公司 | Culture medium prediction system based on ensemble learning, training and culture medium prediction method |
| CN114611386A (en) * | 2022-02-28 | 2022-06-10 | 深圳太力生物技术有限责任公司 | Culture medium mixing proportion optimization method, device, equipment and medium |
Non-Patent Citations (1)
| Title |
|---|
| 薛哲: "基于多视角学习的图像语义分析技术", 北京邮电大学出版社, pages: 4 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111104595B (en) | Deep reinforcement learning interactive recommendation method and system based on text information | |
| Xue et al. | A multi-objective particle swarm optimisation for filter-based feature selection in classification problems | |
| CN110532571A (en) | Text handling method and relevant apparatus | |
| CN107766573B (en) | Commodity recommendation method, commodity recommendation device, commodity recommendation equipment and storage medium based on data processing | |
| Peremezhney et al. | Combining Gaussian processes, mutual information and a genetic algorithm for multi-target optimization of expensive-to-evaluate functions | |
| CN110222838B (en) | Document sorting method and device, electronic equipment and storage medium | |
| Sun et al. | Boosting ant colony optimization via solution prediction and machine learning | |
| Nikolados et al. | Accuracy and data efficiency in deep learning models of protein expression | |
| CN112785005A (en) | Multi-target task assistant decision-making method and device, computer equipment and medium | |
| CN116401555A (en) | Method, system and storage medium for constructing double-cell recognition model | |
| Yang et al. | Prediction of equipment performance index based on improved chaotic lion swarm optimization–LSTM | |
| Safarzadegan Gilan et al. | Active learning in multi-objective evolutionary algorithms for sustainable building design | |
| Guo et al. | Data mining and application of ship impact spectrum acceleration based on PNN neural network | |
| CN108229572B (en) | Parameter optimization method and computing equipment | |
| CN116258923A (en) | Image recognition model training method, device, computer equipment and storage medium | |
| CN116312741A (en) | Culture medium formula optimization method, device, computer equipment and storage medium | |
| CN114821248A (en) | Point cloud understanding-oriented data active screening and labeling method and device | |
| Yang et al. | An improved harris hawks optimization algorithm based on chaotic sequence and opposite elite learning mechanism | |
| CN113779287A (en) | Cross-domain multi-view target retrieval method and device based on multi-stage classifier network | |
| CN111612572A (en) | Adaptive local low-rank matrix approximate modeling method based on recommendation system | |
| Li et al. | Improvement and Application of Fractional Particle Swarm Optimization Algorithm | |
| CN115577618B (en) | Construction method and prediction method of high-pressure converter valve hall environmental factor prediction model | |
| CN116992875B (en) | Text generation method, apparatus, computer device and storage medium | |
| CN116737607B (en) | Sample data caching method, system, computer device and storage medium | |
| Cuevas et al. | An improved evolutionary algorithm for reducing the number of function evaluations |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |