JP5845154B2 - Biological model generation device, biological model generation method, cytometer, and program - Google Patents

Description

  The present invention relates to a biological number model generation device, a biological number model generation method, a cytometer, and a program.

  Conventionally, in order to control the quality of foods and the like, microbial monitoring is performed using a culture method or the like. In the culture method, a diluent such as food is added to a culture medium, and the concentration of microorganisms is examined based on the number of colonies obtained as a result of the culture.

  Here, the culture method is accurate, but it takes several days for results to be obtained, so we cannot ship products such as food until the presence or absence of microbial contamination is determined, and we have the dilemma that product quality deteriorates during that time. Yes. Therefore, in recent years, an apparatus for rapidly monitoring microorganisms by flow cytometry has been developed.

  In the bacterial count measuring apparatus described in Non-Patent Document 1, in order to distinguish from autofluorescence and fluorescence due to non-specific adsorption, samples stained with two types of fluorescent dyes for whole bacteria and one type of fluorescent dye for killed bacteria are used. Fluorescence is detected in time series while flowing through the microchannel. Based on the time series detection data, fluorescence of the fluorescent dye for all bacteria is observed at the same time, and fluorescence of the fluorescent dye for killed bacteria is not observed. It is disclosed that it is determined as a viable bacteria in the case of

Yasuhiko Sasaki & Mitsuo Takei “Introduction of cassette type rapid bacteria count device“ Cassette Lab (R) ONE ”” monthly HACCP June 2012 issue, pp. 38-45

  However, while the conventional technique for measuring the number of viable cells by cytometry can quickly measure the number of bacteria, there is still a problem that there is still a large error in the number of measured cells between the actual measured values by the culture method. Had a point.

  The present invention has been made in view of the above problems, and provides a living body number model generating device, a living body number model generating method, a cytometer, and a program capable of accurately and quickly measuring the living body number. Objective.

  In order to achieve such an object, the living body number model generation device of the present invention is a living body number model generation device including at least a storage unit and a control unit, wherein the storage unit is connected to the living body according to the life or death of the living body. Detection intensity storage means for storing detection intensity data obtained by cytometry using fluorescent labels having different degrees of interaction and emission intensity corresponding thereto, and the control unit is based on the detection intensity data , A histogram generating means for generating a histogram indicating the frequency of each detected intensity section, and multivariate analysis using the frequency value in each section of the histogram as an explanatory variable, and using each coefficient for the obtained explanatory variable And multivariate analysis means for generating a biological number model that is a mathematical expression for determining the presence or absence of the living body or the number of living bodies.

  Further, the living body number model generating apparatus of the present invention is the living body number model generating apparatus described above, wherein the detected intensity storage means is a fluorescence emitted by each of the fluorescent labels obtained by cytometry using a plurality of fluorescent labels. The detected intensity data indicating the detected intensity of the image is stored, and the histogram generating means generates a histogram indicating the frequency for each of the categories based on the detected intensity data in the plurality of fluorescent labels, and the multivariate analyzing means Is characterized in that the frequency for each category of the fluorescent label is used as the explanatory variable.

  Further, the living body number model generating apparatus of the present invention is the living body number model generating apparatus described above, wherein the detected intensity storage means is a fluorescence emitted by each of the fluorescent labels obtained by cytometry using a plurality of fluorescent labels. When the detection intensity data indicating the detection intensity of the plurality of fluorescent labels is stored in the predetermined range based on the detected intensity data, the histogram generation means stores the detected intensity data A histogram showing the frequency of each of the other fluorescent labels for each of the sections is generated.

  Further, the living body number model generating apparatus of the present invention is the living body number model generating apparatus described above, wherein the detected intensity storage means is a fluorescence emitted by each of the fluorescent labels obtained by cytometry using a plurality of fluorescent labels. The detected intensity data indicating the detected intensity is stored, and the histogram generating means adds / subtracts / divides between the detected intensity data of the fluorescent label simultaneously measured, and displays a histogram indicating the frequency for each division of the added / subtracted / divided values. It is characterized by generating.

  Further, the biological number model generation device of the present invention is the biological number model generation device described above, wherein the control unit verifies an error from an actual measurement value for the biological number model obtained by the multivariate analysis unit. The apparatus further comprises model optimization means for optimizing the biological number model by controlling so that the interval between the sections of the histogram repeatedly generated by the histogram generation means is changed. .

  Further, the biological number model generation device of the present invention is the biological number model generation device described above, wherein the histogram generation means converts the detected intensity into a logarithm based on the detected intensity data, and converts the logarithm of the converted logarithm. A histogram showing the frequency for each category is generated.

  Further, the biological number model generation device of the present invention is the biological number model generation device described above, wherein the histogram generation means uses the generated histogram as a spectrum for centering, scaling, smoothing, differentiation, and / or It is characterized by performing standardization processing.

  Further, the biological number model generation device of the present invention is the biological number model generation device described above, wherein the multivariate analysis means includes an unsupervised pattern recognition method, a supervised discriminant analysis method, or a supervised regression analysis. The above-described multivariate analysis is performed using a technique.

  Further, the living body number model generating method of the present invention is a living body number model generating method executed in a computer including at least a storage unit and a control unit, and the storage unit is connected to the living body according to life or death of the living body. Detection intensity storage means for storing detection intensity data obtained by cytometry using fluorescent labels having different degrees of interaction and emission intensity corresponding to the degree of interaction, and the detection intensity data executed in the control unit Based on the histogram generation step for generating a histogram indicating the frequency of each detected intensity section, and the multivariate analysis using the frequency value in each section of the histogram as an explanatory variable. And using a multivariate analysis step to generate a biological number model that is a mathematical formula for determining the presence or absence of the biological body or the number of living bodies. And features.

  Moreover, the cytometer of the present invention is a cytometer that determines the number of living organisms or the presence or absence of the living organism from the detected intensity data using the living organism number model generated by the above-described method for generating the living organism number model. It is characterized by.

  The program according to the present invention is a program for causing a computer including at least a storage unit and a control unit to execute the program, and the storage unit determines the degree of interaction with the living body according to the life and death of the living body, and accordingly. Detection intensity storage means for storing detection intensity data obtained by cytometry using fluorescent labels having different emission intensities, and in the control unit, based on the detection intensity data, A histogram generation step for generating a histogram indicating the frequency, multivariate analysis is performed using the frequency value in each section of the histogram as an explanatory variable, and the presence / absence or the number of living organisms is determined using each coefficient for the obtained explanatory variable. And performing a multivariate analysis step of generating a biological number model as a mathematical expression to be determined.

  According to this invention, the classification of the detection intensity based on the detection intensity data obtained by the cytometry using the fluorescent label in which the degree of the interaction with the living body and the emission intensity corresponding to the living body depending on the life or death of the living body are different. This is a mathematical formula for generating a histogram showing the frequency of each, performing multivariate analysis using the frequency value in each section of the histogram as an explanatory variable, and determining the presence or the number of living organisms using each coefficient for the obtained explanatory variable. Generate a number model. Thereby, this invention has the effect that the biological number model which can measure a biological number correctly and rapidly by cytometry can be shown.

  Further, according to the present invention, in the above, based on the detection intensity data indicating the detection intensity of the fluorescence emitted by each fluorescent label, obtained by cytometry using a plurality of fluorescent labels, the histogram indicating the frequency for each category Are generated in a plurality of fluorescent labels, and multivariate analysis is performed using the frequency of each fluorescent label as an explanatory variable. Thereby, the present invention performs analysis while taking advantage of a plurality of continuous amounts of detection intensity data by performing a multivariate analysis by regarding a histogram based on detection intensities obtained by detecting a plurality of fluorescent labels as one spectrum, There is an effect that the detection rate can be improved.

  Further, according to the present invention, in the above, one of the plurality of fluorescent labels is obtained based on the detection intensity data obtained by cytometry using the plurality of fluorescent labels and indicating the detection intensity of the fluorescence emitted by each fluorescent label. When the detection intensity of one fluorescent label is within a predetermined range, a histogram is generated that indicates the frequency for each of the other fluorescent label sections. Thereby, this invention has the effect that estimation accuracy can be improved using the data in case the fluorescence for killed bacteria is below a threshold value.

  Further, according to the present invention, in the above, based on the detection intensity data obtained by cytometry using a plurality of fluorescent labels and indicating the detection intensity of the fluorescence emitted by each fluorescent label, the fluorescent labels simultaneously measured Addition / subtraction / division / division between the detected intensity data, and a histogram indicating the frequency of each of the divided values is generated. Thereby, this invention produces the effect that the estimation precision can be improved combining the data between different sensors.

  Further, according to the present invention, in the above, the number of living bodies is controlled by changing the interval between the sections of the repeated histogram while verifying an error from the actual measurement value of the living body number model obtained by the multivariate analysis. Optimize the model. Thereby, the present invention increases the risk that the information having the same property is regarded as different when performing analysis using the data of the fine segment intervals, while performing analysis using the data of the coarse segment intervals, There is an effect that it is possible to optimize the interval of the histogram division between the dilemmas that the risk that information of different properties is regarded as the same is increased.

  Further, according to the present invention, in the above, based on the detected intensity data, the detected intensity is converted into a logarithm, and a histogram indicating the frequency of each converted logarithm section is generated. The effect is that the interval can be set.

  Further, according to the present invention, in the above, since the generated histogram is used as a spectrum, centering, scaling, smoothing, differentiation, and / or standardization processing is performed, so that noise included in the spectrum can be reduced or spectrum can be reduced. Effects such as emphasizing the information contained in the image and adjusting the spectrum scale between different samples.

  Further, according to the present invention, in the above, multivariate analysis can be performed using an unsupervised pattern recognition method, a supervised discriminant analysis method, or a supervised regression analysis method. It is possible to perform multivariate analysis by cluster analysis, PLS discriminant analysis, SIMCA, SVM discriminant analysis, multiple regression analysis, PLS regression analysis, SVM regression analysis, or local weighted regression.

FIG. 1 is a block diagram illustrating an example of a configuration of a biological number model generation apparatus 100 to which the present exemplary embodiment is applied. FIG. 2 is a diagram schematically showing a fluorescence detection method for a fluorescently labeled living body using a microchannel. FIG. 3 is a flowchart illustrating an example of processing of the biological number model generation device 100 according to the present embodiment. FIG. 4 is a diagram schematically illustrating a process of converting a histogram from detected intensity data by the histogram generation unit 102b. FIG. 5 is a diagram schematically showing a combining process executed when there is histogram data for a plurality of fluorescent labels. FIG. 6 is a diagram illustrating an error between the biological number model generated according to the present embodiment, an estimated value by the conventional method, and an actually measured value by the culture method. FIG. 7 is a diagram illustrating an example of a data flow according to the present embodiment. FIG. 8 is a diagram showing the combined histogram obtained by the histogram generation process when only the voltage value of channel A is used for the histogram, and the coefficient value for each section of the histogram. FIG. 9 is a graph comparing the biological number model obtained by the histogram generation processing when only the voltage value of channel A is used for the histogram or the estimated value by the conventional method and the actually measured value. FIG. 10 is a diagram showing the combined histogram obtained by the histogram generation process when the voltage values in the channels A and B are used for the histogram, and the coefficient values for each section of the histogram. FIG. 11 is a graph comparing the biological number model obtained by the histogram generation processing when the voltage values in the channels A and B are used for the histogram or the estimated value by the conventional method and the actually measured value. FIG. 12 is a diagram showing the combined histogram obtained by the histogram generation processing when the voltage values in the channels A and C are used for the histogram, and the coefficient values for each section of the histogram. FIG. 13 is a graph comparing the biological model obtained by the histogram generation process when the voltage values in the channels A and C are used for the histogram or the estimated value by the conventional method and the actually measured value. FIG. 14 is a diagram showing the combined histogram obtained by the histogram generation process when the voltage values in all the channels A, B, and C are used for the histogram, and the coefficient values for each section of the histogram. FIG. 15 is a graph comparing the biological model obtained by the histogram generation process when the voltage values in all the channels A, B, and C are used in the histogram or the estimated value by the conventional method and the actually measured value. FIG. 16 shows the combined histogram and histogram obtained by the histogram generation process when the data of channel A and B when B = 0 (the fluorescence of dead bacteria is not detected) is used for the histogram. It is a figure which shows the coefficient value with respect to each division. FIG. 17 shows the biological number model obtained by the histogram generation process when the data of the case where B = 0 (the fluorescence of dead bacteria is not detected) among the voltage values of the channels A and B is used for the histogram, or the conventional method. It is the graph which compared the estimated value and the measured value. FIG. 18 shows each of the combined histogram and the histogram obtained by the histogram generation process when a value (A ÷ C) obtained by dividing the voltage value in channel A by the voltage value of channel C measured simultaneously is used as the histogram. It is a figure which shows the coefficient value with respect to a division. FIG. 19 shows a biological number model obtained by histogram generation processing when a value (A ÷ C) obtained by dividing the voltage value of channel A by the voltage value of channel C simultaneously measured (A ÷ C) is estimated by the histogram generation processing or the conventional method. It is the graph which compared the value and the measured value. FIG. 20 shows each of the combined histogram and the histogram obtained by the histogram generation process in the case where the value (AB) obtained by subtracting the voltage value in channel A by the voltage value of channel B measured simultaneously is used as the histogram. It is a figure which shows the coefficient value with respect to a division. FIG. 21 shows a biological number model obtained by histogram generation processing or a conventional method estimated when a value (AB) obtained by subtracting the voltage value of channel A by the voltage value of channel B measured simultaneously is used for the histogram. It is the graph which compared the value and the measured value. FIG. 22 shows the combined histogram obtained by the histogram generation process in the case where the value (A + C) obtained by adding the voltage value of channel C simultaneously measured to the voltage value of channel A is used for the histogram, and the respective sections of the histogram. It is a figure which shows a coefficient value. FIG. 23 shows the biological number model obtained by the histogram generation processing when the value (A + C) obtained by adding the voltage value of channel C simultaneously measured to the voltage value of channel A is used for the histogram, or an estimated value by the conventional method. It is the graph which compared the measured value. FIG. 24 shows a histogram of combined histograms and histograms obtained by histogram generation processing in the case where a value (A × C) obtained by multiplying the voltage value of channel A simultaneously with the voltage value of channel C is used for the histogram. It is a figure which shows the coefficient value with respect to a division. FIG. 25 shows a biological number model obtained by histogram generation processing or a conventional method estimated when a value (A × C) obtained by multiplying the voltage value of channel A simultaneously with the voltage value of channel C is used for the histogram. It is the graph which compared the value and the measured value. FIG. 26 is a diagram showing the combined histogram obtained by the histogram generation process when values obtained by variously combining the voltage values in the channels A, B, and C are used for the histogram, and coefficient values for each section of the histogram. . FIG. 27 is a graph comparing a biological number model obtained by a histogram generation process when values obtained by variously combining voltage values in channels A, B, and C are used in a histogram, or an estimated value obtained by a conventional method and an actual measurement value. FIG. FIG. 28 is a flowchart illustrating an example of optimization processing in the present embodiment. FIG. 29 is a diagram schematically showing a difference between a case where a histogram is generated with fine segment intervals from voltage data which is detected intensity data and a case where a histogram is generated with coarse segment intervals. FIG. 30 is an isometric view showing errors obtained when the present embodiment is applied using the voltage values of channels A and C in the histogram when the voltage of channel B is not detected (B = 0). . FIG. 31 is a diagram comparing a histogram when the viable cell count estimation error is minimum and a histogram when the viable cell count estimation error is maximum. FIG. 32 is a graph showing the error between the estimated value and the actual measurement value of the living body number model when the viable cell number estimation error is minimum and the living body number model when the viable cell number estimation error is maximum. FIG. 33 is a graph showing the viable cell count estimation error at the time of model creation and verification at each sampling interval of the histogram. FIG. 34 is a graph comparing the conventional method with the biological model when the error is the largest in FIG. 33 and the mathematical model when the error is the smallest. FIG. 35 is a graph (right diagram) showing an error between a histogram (left diagram) when no preprocessing is performed and an estimated value based on the generated biological number model. FIG. 36 is a graph (right diagram) showing a histogram (left diagram) in the case where the preprocessing of centering is performed (left diagram) and an error of an estimated value based on the generated number model. FIG. 37 is a graph (right diagram) showing a histogram (left diagram) in the case of performing preprocessing for scaling (left diagram) and an error of an estimated value based on the generated number model. FIG. 38 is a graph (right diagram) showing a histogram (left diagram) in the case of performing smoothing preprocessing and an error of an estimated value based on the generated number model. FIG. 39 is a graph (right diagram) showing an error in an estimated value based on a histogram (left diagram) and a generated biological number model when pre-processing of first derivative (first derivative) is performed. FIG. 40 is a graph (right diagram) showing an error of an estimated value based on a histogram (left diagram) and a generated biological number model when preprocessing of second derivative (second derivative) is performed. FIG. 41 is a graph (right diagram) showing an error between a histogram (left diagram) and an estimated value based on the generated number model when normalization is performed. FIG. 42 is a graph (left diagram) and a graph (right diagram) showing an error in an estimated value by a generated biological number model when the standardization, second-order differentiation, centering, and scaling preprocessing are performed in combination. is there. FIG. 43 is a diagram illustrating an example in which multivariate analysis is performed by principal component analysis. FIG. 44 is a diagram showing an example in which multivariate analysis was performed by cluster analysis. FIG. 45 is a diagram illustrating an example in which multivariate analysis is performed by PLS discriminant analysis. FIG. 46 is a diagram showing an example in which multivariate analysis was performed by SIMCA. FIG. 47 is a diagram showing an example in which multivariate analysis was performed by SVM regression analysis. FIG. 48 is a diagram illustrating an example in which multivariate analysis is performed by LWR (local weighted regression).

  Hereinafter, embodiments of a biological number model generation device, a biological number model generation method, a cytometer, and a program according to the present invention will be described in detail with reference to the drawings. Note that the present invention is not limited to the embodiments.

  In particular, in the following embodiment, an example in which the present invention is applied to the quality control field of products such as foods may be described. However, the present invention is not limited to this, and the same applies to all technical fields such as medical care and clinical examinations. Can be applied to.

[Outline of this embodiment]
Hereinafter, an outline of an embodiment according to the present invention will be described, and then a configuration, processing, and the like of the present embodiment will be described in detail.

  The present embodiment schematically has the following basic features. First, in the present embodiment, detection intensity data obtained by cytometry using a fluorescent label having a degree of interaction with a living body and a light emission intensity corresponding to the degree of interaction with the living body depending on whether the living body is alive or dead is acquired. Here, in this embodiment, the “living body” means fungi such as bacteria and fungi, individuals of microorganisms, unicellular organisms, individual cells in multicellular organisms, and quasibiological organisms such as mitochondria and chlorophyll. And any organism that can be detected by cytometry. Further, the “fluorescent label” is a label that differs in the degree of interaction with the living body and the light emission intensity corresponding to the living or dead body. Here, the fluorescent label may be a combination that can determine whether a living organism is alive or not by a plurality of fluorescent labels. For example, cells or microorganisms (using a multiple staining method described in JP2011-92104A) may be used. Live bacteria) may be labeled. As an example, DAPI (4 ′, 6-diamidino-2-phenylindole) or the like may be used as a dye that stains all bacteria (live and dead), and PI (Propidium Iodide) or the like as a dye that stains dead bacteria. May be used. Further, “cytometry” is a technique capable of detecting an individual living body, such as flow cytometry and imaging cytometry, and is performed by a flow cytometer, an imaging cytometer, or the like.

  And this embodiment produces | generates the histogram which shows the frequency for every division of detection intensity based on detection intensity data. Details of the method of creating the histogram will be described later.

  The present embodiment performs a multivariate analysis using the frequency value in each section of the histogram as an explanatory variable, and uses a coefficient for the obtained explanatory variable to determine the presence or absence of the living body or the living body number model. Is generated. The term “biological number model” is not limited to a model for determining the number of living bodies, but may mean a model for determining the presence or absence of a living body. Here, in the present embodiment, the number model obtained by the multivariate analysis unit is repeatedly verified while verifying an error with the actual measurement value, and the interval of the above-mentioned histogram segment is changed to optimize the number model. May also be performed.

  The above is the outline of the present embodiment. In addition, although the example which implements this embodiment by the apparatus provided with control parts, such as a computer, is demonstrated below, this invention is not restricted to this, The method of this embodiment may be implemented by a person. is there.

[Configuration of Biological Model Generation Device]
Next, the configuration of the living body number model generation device of the present embodiment according to the present invention will be described with reference to FIG. FIG. 1 is a block diagram showing an example of the configuration of a biological number model generation apparatus 100 to which the present embodiment is applied, and conceptually shows only the portion related to the present invention in the configuration.

  In FIG. 1, the biological number model generation apparatus 100 schematically includes a control unit 102, a communication control interface unit 104, an input / output control interface unit 108, and a storage unit 106. Here, the control unit 102 is a CPU or the like that comprehensively controls the whole body number model generation apparatus 100. The input / output control interface unit 108 is an interface connected to the measurement unit 110, the input unit 112, the output unit 114, and the like. The storage unit 106 is a device that stores various databases and tables. Each unit of the biological number model generation apparatus 100 is connected to be communicable via an arbitrary communication path.

  Various databases and files (detection intensity file 106a, histogram file 106b, biological number model file 106c, etc.) stored in the storage unit 106 are storage means such as a fixed disk device. For example, the storage unit 106 stores various programs, tables, files, databases, and the like used for various processes.

  Among these components of the storage unit 106, the detection intensity file 106a is obtained by cytometry on a sample that has been subjected to fluorescent labeling that differs in the degree of interaction with the living body and the emission intensity corresponding to the living body's life or death. Detection intensity storage means for storing the obtained detection intensity data. For example, the detection intensity file 106a may store raw data acquired by the measurement unit 110 as detection intensity data, or may store detection intensity data received from the external device 200 via the network 300. As an example, the detection intensity data stored in the detection intensity file 106a is fluorescence intensity detection data arranged in a detection order such as a time-series time order or a scanned order according to the nature of the detection means. For example, the detected intensity data obtained by the flow cytometer is, for example, voltage data (electric pulse proportional to the light intensity) associated with the time in each channel in which the wavelength range to be detected is preset. And corresponding intensity data). Moreover, the detection intensity data obtained by the imaging cytometer may be detection value data of a desired wavelength range associated with coordinates (x coordinate, y coordinate) scanned on a two-dimensional plane as an example.

  The histogram file 106b is a histogram data storage unit that stores histogram data generated by a histogram generation unit 102b described later. For example, the histogram data stored in the histogram file 106b is frequency value data associated with each section of the histogram.

  The biological number model file 106c is a biological number model storage unit that stores a biological number model generated by a multivariate analysis unit 102c described later. Here, the data indicating the biological number model stored in the biological number model file 106c may be data in which coefficients corresponding to the histogram classification (channel type, width, range, etc.) are defined.

  In FIG. 1, the input / output control interface unit 108 controls the measurement unit 110, the input unit 112, the output unit 114, and the like. Here, as the output unit 114, in addition to a monitor (including a home television), a speaker can be used (hereinafter, the output unit 114 may be described as a monitor). As the input unit 112, a keyboard, a mouse, a microphone, or the like can be used.

  The measurement unit 110 is a light intensity detection unit that detects the intensity of light in a predetermined wavelength range. For example, the measurement unit 110 may detect the intensity in a range including the wavelength of the fluorescence emitted by the fluorescent label. The measurement unit 110 may detect the intensity of light as a voltage pulse by the photoelectric effect. For example, the peak value or the pulse area of the voltage pulse detected when the fluorescently labeled biological body passes through the detection region. You may detect as detection intensity. As an example, the measurement unit 110 may be a flow cytometer or an imaging cytometer. For example, the measurement unit 110 as an imaging cytometer scans a plane on which a sample containing a living body is scattered while irradiating a laser beam, and the intensity data of the fluorescence emitted by the sample along the scanning line. May be obtained. In addition, the measurement unit 110 as a flow cytometer allows the suspended liquid in which individual living bodies are suspended to flow in a cylindrical sheath liquid and pass over the laser light, thereby increasing the intensity of fluorescence excited by the laser light. You may detect in time series. In addition, the measurement unit 110 as a flow cytometer may detect the intensity of fluorescence by flowing a biological suspension through the microchannel and passing it over the excitation light. Here, FIG. 2 is a diagram schematically showing a fluorescence detection method for a fluorescently labeled living body using a microchannel.

  Unlike flow cytometry using a sheath liquid, microcapillary flow cytometry uses a sequence-controlled air pressure to send a floating liquid into a microchannel in a microcapillary as shown in FIG. Flow). Here, as an example, two wavelengths of excitation light may be irradiated simultaneously, and three wavelengths of fluorescence may be detected simultaneously in each channel. As an example of a microcapillary flow cytometer, a cassette laboratory (R) ONE manufactured by Hitachi Engineering & Service Co., Ltd. described in Non-Patent Document 1 may be used as the measurement unit 110.

  In FIG. 1, the control unit 102 has an internal memory for storing a control program such as an OS (Operating System), a program that defines various processing procedures, and necessary data. And the control part 102 performs the information processing for performing various processes by these programs. The control unit 102 includes a data acquisition unit 102a, a histogram generation unit 102b, a multivariate analysis unit 102c, an optimization unit 102d, and a living body number determination unit 102f in terms of functional concept.

  Among these, the data acquisition unit 102a is detection intensity data acquisition means for acquiring detection intensity data. For example, the data acquisition unit 102a may acquire detection intensity data measured via the measurement unit 110. In addition, when data processing such as analog / digital conversion is necessary for the raw data measured through the measurement unit 110, the data acquisition unit 102a performs measurement through the measurement unit 110 using a known data processing method. The obtained raw data may be converted into detected intensity data. Further, the data acquisition unit 102 a may receive detection intensity data from the external device 200 via the network 300. In the present embodiment, the data acquisition unit 102a stores the acquired detection intensity data in the detection intensity file 106a.

  In addition, the histogram generation unit 102b generates a histogram indicating the frequency of each detection intensity based on the detection intensity data stored in the detection intensity file 106a, and stores the generated histogram data in the histogram file 106b. It is a generation means. Here, the histogram generation unit 102b may generate a histogram indicating the frequency of each detected intensity for each of the plurality of fluorescent labels based on the detected intensity data corresponding to the plurality of fluorescent labels. For example, the histogram generation unit 102b generates and generates a histogram indicating the frequency of each detection intensity category based on the two-wavelength detection intensity data corresponding to the whole bacteria fluorescent dye and the dead bacteria fluorescent dye. The histograms may be combined. For example, since the histogram data divided into 1024 is obtained for each channel of five flow cytometers, the histogram generator 102b connects them in a row to obtain a frequency value that is 5120 variables per sample. Also good. Here, the histogram generation unit 102b may change the interval of the histogram division for each channel of the cytometer, that is, for each detection intensity by a plurality of fluorescent labels.

  Further, the histogram generation unit 102b classifies other fluorescent labels when the detected intensity of one fluorescent label is within a predetermined range based on the detected intensity data corresponding to the plurality of fluorescent labels. You may generate | occur | produce the histogram which shows every frequency. For example, the histogram generation unit 102b indicates the frequency for each category of the other fluorescent label when the detected intensity of one fluorescent label is equal to or lower than a predetermined threshold based on the detected intensity data corresponding to the plurality of fluorescent labels. A histogram may be generated. For example, the detection intensity of fluorescence emitted by each fluorescent label obtained by cytometry using a fluorescent label that easily interacts in a living state and a fluorescent label that easily interacts in a dead state. For the detected intensity data shown, the histogram generator 102b may generate a histogram indicating the frequency of each of the former fluorescent labels when the latter detected intensity of the fluorescent label is equal to or lower than a predetermined threshold. As another example, the histogram generation unit 102b performs death based on detection intensity data indicating the detection intensity of fluorescence emitted by each fluorescent label, which is obtained by cytometry using a fluorescent label for whole bacteria and a fluorescent label for dead bacteria. A histogram may be generated that indicates the frequency for each category of the fluorescent label for all bacteria when the detection intensity of the fluorescent label for bacteria is not more than a predetermined threshold.

  In addition, the histogram generation unit 102b detects the fluorescence label detection intensity data measured simultaneously based on the detection intensity data indicating the detection intensity of the fluorescence emitted by each fluorescence label, which is obtained by cytometry using a plurality of fluorescence labels. It is also possible to generate a histogram indicating the frequency for each category of the value obtained by adding / subtracting / dividing between the values. For example, the histogram generation unit 102b includes the detection intensity data indicating the detection intensity of the fluorescence emitted by each fluorescent label obtained by cytometry using the fluorescent label for whole bacteria and the dead fluorescent label. You may generate | occur | produce the histogram which shows the frequency for every division | segmentation of the value which subtracted thru | or divided the detection intensity by the detection intensity of the fluorescent label for killed bacteria detected at the same time.

  Here, the histogram generation unit 102b may convert the detection intensity into a logarithm based on the detection intensity data, and generate a histogram indicating the frequency for each segment of the converted logarithm. For example, the histogram generation unit 102b does not generate a histogram by dividing the detection intensity at equal intervals, but by dividing the logarithm of the detection intensity (common logarithm, natural logarithm, etc.) at equal intervals. A histogram may be generated. Moreover, it is good also as not only a semilogarithm but a both logarithm. For example, the histogram generation unit 102b may generate a log-log histogram based on the detection intensity data, with the horizontal axis representing the logarithm of the detection intensity and the vertical axis representing the logarithm of the frequency value.

  The histogram generation unit 102b may perform preprocessing on the generated histogram before executing multivariate analysis processing by the multivariate analysis unit 102c described later. For example, the histogram generation unit 102b may perform preprocessing such as centering, scaling, smoothing, differentiation, and / or standardization using the generated histogram as a spectrum.

  In addition, the multivariate analysis unit 102c performs multivariate analysis using the frequency values in each section of the histogram generated by the histogram generation unit 102b as explanatory variables, and uses the coefficients for the obtained explanatory variables to determine the presence / absence of a living body (raw This is a multivariate analysis means for generating a biological number model that is a mathematical expression for determining the presence or absence of bacteria or the number of living organisms (viable bacterial count or the like). For example, the multivariate analysis unit 102c performs multivariate analysis using the frequency value of each section of the histogram data stored in the histogram file 106b as an explanatory variable, and the coefficient of each explanatory variable obtained as a result is used as a section number or the like. Correspondingly, the biological number model file 106c may be stored.

Here, the multivariate analysis unit 102c may generate a biological number model using the following mathematical formula (theoretical formula). Then, the multivariate analysis unit 102c may perform fitting of the theoretical formula by applying each biological count (actually measured value) based on a dilution series of a sample with a known biological count to the above theoretical formula. That is, the multivariate analysis unit 102c may adjust and determine the value of the coefficient of the theoretical formula using the least square method or the like so that the theoretical formula is best fitted (fitted) to the number of living bodies of each actually measured value. Good.

  Here, the multivariate analysis unit 102c may perform multivariate analysis using an unsupervised pattern recognition method, a supervised discriminant analysis method, or a supervised regression analysis method. For example, the multivariate analysis unit 102c performs principal component analysis, cluster analysis, PLS (Partial Last Squares) discriminant analysis, SIMCA (Soft Independent Modeling of Class Analysis), SVM (Support Vector Machine Regression, PL analysis), Multivariate analysis may be performed by analysis, SVM regression analysis, local weighted regression, or the like. In addition, the multivariate analysis unit 102c is disclosed in Japanese Unexamined Patent Application Publication No. 2010-185719, Japanese Unexamined Patent Application Publication No. 2004-294337, Japanese Unexamined Patent Application Publication No. 2010-266380, “Multivariate Data Utilization Method by JMP” (URL: http: // /Go.gl/NjF32), “Data Science by R-From Basics of Data Analysis to Latest Methods” (URL: http://goo.gl/LszNv), etc. A random analysis may be performed. As described above, the multivariate analysis unit 102b may generate a biological number model represented by a mathematical expression or graph such as a regression equation, a polynomial, a calibration curve, or a discriminant by adopting the above-described method.

  In addition, the optimization unit 102d repeats while verifying an error between the estimated value and the actual measurement value obtained from the biological number model obtained by the multivariate analysis unit 102c, and the histogram generation unit 102b changes the interval of the histogram division. It is a model optimizing unit that optimizes the biological number model by controlling. Here, the optimization unit 102d may optimize the histogram interval based on a technique such as cross validation or double cross validation. The optimization unit 102d may perform the histogram interval optimization process for all channels (detection intensities of all fluorescent labels) at once, or may optimize each channel (detection intensity of each fluorescent label) individually. Good.

  In addition, the living body number determination unit 102f, based on the living body number model generated by the multivariate analysis unit 102c, detects the presence or absence of living organisms (such as the presence or absence of viable bacteria) or the number of living organisms (such as the number of living bacteria) from the detected intensity data. It is a living body number judging means for judging.

  The above is an example of the configuration of the living body number model generation apparatus 100. Here, the biological number model generation device 100 may be communicably connected to the network 300 via a communication device such as a router and a wired or wireless communication line such as a dedicated line. In this case, the communication control interface unit 104 is an interface connected to a communication device (not shown) such as a router connected to a communication line or the like, and the biological model generation device 100 and the network 300 (or a communication such as a router). Communication control with the device. That is, the communication control interface unit 104 has a function of communicating data with other terminals via a communication line. In FIG. 1, a network 300 has a function of connecting a biological number model generation apparatus 100 and an external device 200 to each other, such as the Internet.

  The biological number model generation apparatus 100 is communicably connected via the network 300 to an external device 200 that provides an external database relating to measurement data, parameters, and the like, an external program for causing the biological number model generation apparatus to function, and the like. May be.

  In FIG. 1, an external device 200 is mutually connected to the biological number model generation device 100 via a network 300, and an external database or information processing device regarding measurement data, theoretical formulas, parameter values, etc. to the user. Has a function of providing an external program or the like such as a target nucleic acid measurement program for causing the to function as a biological number model generation device. Here, the external device 200 may be configured as a WEB server, an ASP server, or the like. Further, the hardware configuration of the external device 200 may be configured by an information processing apparatus such as a commercially available workstation or a personal computer, and an accessory device thereof. Each function of the external device 200 is realized by a CPU, a disk device, a memory device, an input device, an output device, a communication control device, and the like in the hardware configuration of the external device 200 and a program for controlling them.

[Process of Biological Model Generation Device 100]
Next, an example of the process of the biological number model generation device 100 according to the present embodiment configured as described above will be described in detail with reference to FIGS. Here, FIG. 3 is a flowchart showing an example of processing of the biological number model generation apparatus 100 in the present embodiment.

  As shown in FIG. 3, first, the histogram generation unit 102b generates a histogram indicating the frequency for each category of the detected intensity based on the detected intensity data stored in the detected intensity file 106a, and generates the generated histogram data. Store in the histogram file 106b (step SA-1). Here, FIG. 4 is a diagram schematically showing the conversion processing of the histogram from the detected intensity data by the histogram generation unit 102b. In addition, the detection intensity data of FIG. 4 shows the voltage value in the whole bacteria channel for fluorescent labeling for whole bacteria, and the voltage value in the dead bacteria channel for fluorescent labeling for dead bacteria detected at the time t. The determination item in the detection intensity data is an example of a conventional method for determining that there is one viable cell if the voltage value in the dead cell channel is 0 and the voltage value in the whole cell channel is positive. It is not an indispensable item for carrying out the embodiment.

  As shown in FIG. 4, as an example, the histogram generation unit 102b generates a histogram indicating a frequency value for each voltage value category for each channel of the cytometer (that is, for each detection intensity by a plurality of fluorescent labels). May be. Note that the histogram generation unit 102b is not limited to generating a histogram indicating the frequency of each detected intensity for each of the plurality of fluorescent labels based on the detected intensity data corresponding to the plurality of fluorescent labels. For example, the histogram generation unit 102b determines that the detection intensity of one fluorescent label is within a predetermined range based on detection intensity data corresponding to a plurality of fluorescent labels (for example, when the detection intensity is equal to or lower than a predetermined threshold, You may generate | occur | produce the histogram which shows the frequency for every division | segmentation of another fluorescence label | marker (when it is more than a threshold value). For example, in the detection data example of FIG. 4, the histogram generation unit 102 b detects the detection intensity indicating the detection intensity of the fluorescence emitted by each fluorescent label, which is obtained by cytometry using the fluorescent label for whole bacteria and the fluorescent label for dead bacteria. Based on the data, when the detection intensity of the killing fluorescent label is less than a predetermined threshold (for example, when the voltage value is 0), a histogram indicating the frequency for each category of the fluorescent label for all bacteria is generated. May be. In addition, the histogram generation unit 102b detects the fluorescence label detection intensity data measured simultaneously based on the detection intensity data indicating the detection intensity of the fluorescence emitted by each fluorescence label, which is obtained by cytometry using a plurality of fluorescence labels. It is also possible to generate a histogram indicating the frequency for each category of the value obtained by adding / subtracting / dividing between the values. For example, the histogram generation unit 102b includes the detection intensity data indicating the detection intensity of the fluorescence emitted by each fluorescent label obtained by cytometry using the fluorescent label for whole bacteria and the dead fluorescent label. You may generate | occur | produce the histogram which shows the frequency for every division | segmentation of the value which subtracted thru | or divided the detection intensity by the detection intensity of the fluorescent label for killed bacteria detected at the same time.

  Note that the histogram generation unit 102b may convert the detected intensity into a logarithm based on the detected intensity data, and generate a histogram indicating the frequency for each segment of the converted logarithm. The histogram generation unit 102b may perform preprocessing on the generated histogram before the multivariate analysis processing in the next step SA-2. For example, the histogram generation unit 102b may perform preprocessing such as centering, scaling, smoothing, differentiation, and / or standardization using the generated histogram as a spectrum.

  Subsequently, the multivariate analysis unit 102c performs multivariate analysis using the frequency value in each section of the histogram based on the data generated by the histogram generation unit 102b and stored in the histogram file 106b as an explanatory variable (step SA-2). Here, FIG. 5 is a diagram schematically showing a combining process executed when there is histogram data for a plurality of fluorescent labels.

As shown in FIG. 5, when there is histogram data for a plurality of fluorescent labels, the multivariate analysis unit 102c combines the plurality of histograms, and uses the frequency value in each section of the combined histograms as an explanatory variable for multivariate analysis. May be performed. Here, the multivariate analysis unit 102c may perform multivariate analysis using an unsupervised pattern recognition method, a supervised discriminant analysis method, or a supervised regression analysis method. For example, the multivariate analysis unit 102c performs principal component analysis, cluster analysis, PLS (Partial Last Squares) discriminant analysis, SIMCA (Soft Independent Modeling of Class Analysis), SVM (Support Vector Machine Regression, PL analysis), Multivariate analysis may be performed by analysis, SVM regression analysis, local weighted regression, or the like. In addition, the multivariate analysis unit 102c is disclosed in Japanese Unexamined Patent Application Publication No. 2010-185719, Japanese Unexamined Patent Application Publication No. 2004-294337, Japanese Unexamined Patent Application Publication No. 2010-266380, “Multivariate Data Utilization Method by JMP” (URL: http: // /Go.gl/NjF32), “Data Science by R-From Basics of Data Analysis to Latest Methods” (URL: http://goo.gl/LszNv), etc. A random analysis may be performed. As an example, the multivariate analysis unit 102c may generate a biological number model using the following mathematical formula (theoretical formula). Then, the multivariate analysis unit 102c may perform fitting of the theoretical formula by applying each biological count (actually measured value) based on a dilution series of a sample with a known biological count to the above theoretical formula. That is, the multivariate analysis unit 102c may adjust and determine the value of the coefficient of the theoretical formula using the least square method or the like so that the theoretical formula is best fitted (fitted) to the number of living bodies of each actually measured value. Good.

  The multivariate analysis unit 102c is a living body that is a mathematical expression that determines the presence / absence of a living body (such as the presence / absence of living bacteria) or the number of living bodies (such as the number of living bacteria) using each coefficient for the explanatory variable by multivariate analysis processing. A number model is generated, and the generated number model is stored in the number model file 106c (step SA-3). For example, the multivariate analysis unit 102c supports the coefficient of each explanatory variable obtained as a result of the multivariate analysis using the frequency value of each division of the histogram data stored in the histogram file 106b as an explanatory variable, to the division number or the like. In addition, the biological number model file 106c may be stored.

  Then, the optimization unit 102d verifies an error between the estimated value and the actual measurement value obtained from the biological number model obtained by the multivariate analysis unit 102c, returns the process to Step SA-1, and classifies the histogram by the histogram generation unit 102b. Is controlled to be changed (step SA-4). As described above, the optimization unit 102d verifies the biological number model generated by the multivariate analysis unit 102c and performs control so that the histogram section is repeatedly changed, so that an error from the actual measurement value is reduced. Optimize the number model. Here, the optimization unit 102d may optimize the histogram interval based on a technique such as cross validation or double cross validation. The optimization unit 102d may perform the histogram interval optimization process for all channels (detection intensities of all fluorescent labels) at once, or may optimize each channel (detection intensity of each fluorescent label) individually. Good.

  The above is an example of the processing of the biological number model generation apparatus 100 in the present embodiment. Using the living body number model generated in this way, the living body number model generating apparatus 100 performs the processing of the living body number determination unit 102f to detect the presence / absence of a living body (such as the presence / absence of viable bacteria) or the number of living bodies (from the detected intensity data) Viable counts etc.) can be determined accurately and quickly. Here, FIG. 6 is a diagram illustrating an error between the number model generated by the present embodiment and the estimated value by the conventional method and the actually measured value by the culture method. The horizontal axis of FIG. 6 shows the actual measurement value in the culture method, and the vertical axis shows the estimated value by each method, and the error is smaller as the slope is on the straight line. The rhombus in the figure is a plot by the conventional cytometry method when the fluorescence detection sensitivity is high, the square in the figure is the plot by the conventional cytometry method when the fluorescence detection sensitivity is low, and the triangle in the figure is 3 shows a plot according to this embodiment.

As shown in FIG. 6, in the conventional cytometry method, when the number of viable bacteria is small, there are many false positives, and the detection limit exceeds 10 ³ . Therefore, the presence or absence of viable bacteria cannot be determined correctly. In addition, even when the number of viable bacteria is large, the number of viable bacteria cannot be estimated accurately because it is estimated to be about 1 to 2 digits lower than the actually measured value. According to this embodiment, whether the number of viable bacteria is large or small, the plot is almost on a straight line with a slope of 1, and it is possible to estimate the exact viable cell count with little error from the actual measurement value. It is. In the conventional cytometry method, as shown in the determination item of FIG. 4, in order to discriminate a discrete value whether live or dead from fluorescent intensity data (voltage value) that is a continuous amount, False positives were likely to occur. According to the present embodiment, the multivariate analysis is performed by regarding the histogram as a spectrum, and the analysis is performed while taking advantage of the continuous amount of data. Therefore, an accurate number of living organisms with fewer errors and false positives than the conventional cytometry method. And the presence / absence of a living body can be determined, and the detection rate (estimated value / measured value) can be improved.
The

[Example]
Next, examples according to the present embodiment will be described below with reference to FIGS. Here, FIG. 7 is a diagram illustrating an example of a data flow according to the present embodiment.

  As shown in FIG. 7, the original data MA-1 obtained by cytometry is converted into voltage value data MA-2 that is detected intensity data. Subsequently, by converting the voltage value into a logarithm (MA-3), the voltage value data MA-2 is converted into logarithmic voltage value data MA-4. Then, the histogram is converted (MA-5) according to this embodiment, and the voltage value data MA-4 is converted into a histogram MA-6.

  On the other hand, on the actual measurement value side, based on the summary table MA-7, the viable cell count (number) data MA-8 included in the CSV data is extracted, and the count number MA-9 is prepared as the actual measurement value.

  In this embodiment, both the histogram MA-6 and the count number MA-9 are logarithmically converted (MA-10), and converted to a logarithmic expression histogram MA-11 and a count number MA-12, respectively.

  In this embodiment, multivariate analysis is performed using the frequency value in each section of the histogram MA-11 as an explanatory variable and the actual measurement count MA-12 as an objective variable (MA-13).

  In this embodiment, the biological number model is verified by calculating an error between the estimated value and the actual measurement value obtained from the biological number model obtained by the multivariate analysis (MA-14). For example, according to the cross-validation method, among the count data MA-12 that are actually measured values, a multivariate analysis is performed a plurality of times while shuffling a data set used as teacher data for multivariate analysis and a data set for model verification. Model validation may be repeated.

  In this embodiment, an estimation result (an error evaluation result from the actual measurement value, etc.) of the biological number model generated as a result of the multivariate analysis is obtained (MA-15).

  In this embodiment, the interval between the histogram segments is reset (MA-16), and the above-described processing is repeated to optimize the biological number model.

  The above is a diagram illustrating an example of a data flow according to the present embodiment. Next, details of the histogram generation process, the optimization process, the preprocess, and the multivariate analysis process in this embodiment will be described. In the following examples, unless otherwise specified, PLS regression analysis is used as a multivariate analysis method.

[Histogram generation processing]
An example of the histogram generation process in the present embodiment will be described with reference to FIGS.

  In the conventional flow cytometry method (hereinafter referred to as “conventional method”), voltage is generated in the sensors (A and C channels) that capture the fluorescence of the fluorescent stains a and c for viable bacteria, and the fluorescent stain for dead bacteria is used. When no voltage was generated in the sensor (B channel) that captures the fluorescence of the agent b (that is, in the case of “A and C not B”), it was determined as one viable bacterium (see Non-Patent Document 1).

In this embodiment, the voltage values of the channels (A to C channels) measured in the flow cytometry are histogrammed, and the histograms of the channels are combined. Here, the upper diagrams of FIGS. 8, 10, 12, 14, 16, 18, 20, 20, 24, and 26 show the combined results obtained by the respective histogram generation processes. It is a figure which shows a histogram. In the present embodiment, PLS regression is applied to the combined histogram, and a living body number model that is a mathematical formula for calculating the viable cell count from the histogram is constructed. Here, the living body number model (theoretical formula) as a basis in the present embodiment is the following polynomial that is added by multiplying the frequency value in each section of the histogram by a coefficient. 8, 10, 12, 14, 16, 18, 20, 20, 22, 24, and 26, the lower graphs show the respective sections of the histogram obtained as a result of the PLS regression. It is a figure which shows the coefficient value (regression coefficient) with respect to (division number i = 1-n, object channel AC, range).
coefficient

9, 11, 13, 15, 17, 19, 21, 23, 25, and 27 are biometric models or conventional methods obtained by the respective histogram generation processes. It is the graph which compared the estimated value by and actual value. In addition, the viable count estimation accuracy of the biological number model was evaluated by the following index. The coefficient of determination is a value obtained by squaring the correlation coefficient, and indicates how much linearity there is between the measured value (the number of viable bacteria counted by the culture method) and the estimated value. Larger is better (maximum value is 1). The standard deviation of the error is a value obtained by summing up the estimation errors of each sample, and is obtained by the following formula. The smaller the value, the better.

  Here, FIGS. 8 and 9 are diagrams illustrating an example in which the present embodiment is applied using only the voltage value of the channel A in the histogram. When this data set is used, in the conventional method, the determination coefficient = 0.611 and the standard deviation of the error = 2.53 log (pieces / ml). In this embodiment, the determination coefficient = 0.694 and the error A viable count estimation accuracy of standard deviation = 1.15 log (cells / ml) was obtained.

  10 and 11 are diagrams showing an example in which the present embodiment is applied using the voltage values in channels A and B as histograms. When this data set is used, in the conventional method, the determination coefficient = 0.722, the standard deviation of error = 1.15 log (pieces / ml), but in this embodiment, the determination coefficient = 0.551, A viable cell count estimation accuracy of standard deviation = 1.05 log (cells / ml) was obtained.

  12 and 13 are diagrams showing an example in which the present embodiment is applied using voltage values in channels A and C as histograms. When this data set is used, in the conventional method, the determination coefficient = 0.404 and the standard deviation of the error = 2.54 log (pieces / ml). In this embodiment, the determination coefficient = 0.696 and the error A viable count estimation accuracy of standard deviation = 1.15 log (cells / ml) was obtained.

  14 and 15 are diagrams showing an example in which the present embodiment is applied using the voltage values in all the channels A, B, and C as histograms. When this data set is used, in the conventional method, the determination coefficient = 0.722, the standard deviation of error = 1.15 log (pieces / ml), but in this embodiment, the determination coefficient = 0.552, A viable cell count estimation accuracy of standard deviation = 1.04 log (cells / ml) was obtained.

  FIG. 16 and FIG. 17 are diagrams showing an example in which the present embodiment is applied using the data in the case where B = 0 (the fluorescence of dead bacteria is not detected) among the voltage values of channels A and B as histograms. When this data set is used, the determination coefficient is 0.872 and the standard deviation of error is 1.15 log (pieces / ml) in the conventional method. In this embodiment, the determination coefficient is 0.958 and the error is The viable cell count estimation accuracy of standard deviation = 0.45 log (cells / ml) was obtained.

  FIG. 18 and FIG. 19 are diagrams showing an example in which the present embodiment is applied using a value (A ÷ C) obtained by dividing the voltage value in channel A by the voltage value of channel C, which is simultaneously measured, in the histogram. When this data set is used, in the conventional method, the determination coefficient = 0.404 and the standard deviation of error = 2.54 log (pieces / ml). In this embodiment, the determination coefficient = 0.815, and the error A viable cell count estimation accuracy of standard deviation = 0.91 log (cells / ml) was obtained.

  20 and 21 are diagrams showing an example in which the present embodiment is applied using a value (AB) obtained by subtracting the voltage value in channel A by the voltage value in channel B, which is simultaneously measured, as a histogram. When this data set is used, in the conventional method, the coefficient of determination is 0.872 and the standard deviation of the error is 1.15 log (pieces / ml). In this embodiment, the coefficient of determination is 0.842 and the error is A viable cell count estimation accuracy of standard deviation = 0.83 log (cells / ml) was obtained.

  22 and 23 are diagrams showing an example in which the present embodiment is applied using a value (A + C) obtained by adding the voltage value of channel C, which is simultaneously measured to the voltage value of channel A, for the histogram. When this data set is used, in the conventional method, the coefficient of determination is 0.404 and the standard deviation of the error is 2.54 log (pieces / ml). In this embodiment, the coefficient of determination is 0.587 and the error is A viable count estimation accuracy of standard deviation = 1.34 log (cells / ml) was obtained.

  24 and 25 are diagrams illustrating an example in which the present embodiment is applied using a value (A × C) obtained by multiplying the voltage value of channel A by the voltage value of channel C simultaneously measured in the histogram. When this data set is used, in the conventional method, the determination coefficient = 0.404 and the standard deviation of error = 2.54 log (pieces / ml). In this embodiment, the determination coefficient = 0.812, and the error A viable count estimation accuracy of standard deviation = 0.90 log (cells / ml) was obtained.

  26 and 27 are diagrams showing an example in which the present embodiment is applied using values obtained by variously combining voltage values in channels A, B, and C in a histogram. In addition, in the item of the target channel in the lower diagram of FIG. 26, a combination calculation method of voltage values in channels A, B and C is shown. When this 21-channel data set is used, in the conventional method, the determination coefficient = 0.722, the standard deviation of error = 1.15 log (pieces / ml), but in this embodiment, the determination coefficient = 0. 961, standard deviation of error = 0.43 log (cells / ml) was obtained.

  This is the end of the description of the example of the histogram generation processing in the present embodiment. As described above, the present embodiment can be applied using the luminance information of fluorescence captured by one or a plurality of sensors as a histogram. In particular, accuracy can be improved by performing processing only on data when fluorescence is not captured by a specific sensor. It is also possible to improve accuracy by creating a histogram of values obtained by adding, subtracting, and dividing fluorescence luminance values (voltage values) between different sensors.

[Optimization processing]
An example of optimization processing in the present embodiment will be described with reference to FIGS. Here, FIG. 28 is a flowchart illustrating an example of optimization processing in the present embodiment.

  As shown in FIG. 28, first, in this embodiment, the division interval is set for each channel of the detected intensity data (step SB-1). Then, the obtained data sample is divided into a calibration set and a validation set (step SB-2).

  Then, multivariate analysis is applied to the calibration set (step SB-3), and a biological number model for estimating the viable cell count is generated (step SB-4).

  And the standard deviation of the error of an estimated value is calculated by applying the obtained number model to a validation set (step SB-5).

  Then, the process returns to step SB-2, and the data sample is subdivided into a calibration set and a validation set in another combination, and the above-described steps SB-2 to SB-6 are repeated to perform cross-validation. (Step SB-6).

  When the cross-validation is completed (step SB-6, Yes), the process returns to step SB-1, the division interval for each channel of the detected intensity data is reset, and the processes after step SB-1 are repeated. Then, the interval between segments to be an optimal number model is optimized (step SB-7). Here, FIG. 29 is a diagram schematically showing a difference between a case where a histogram is generated with fine segment intervals from voltage data which is detected intensity data and a case where a histogram is generated with coarse segment intervals. In addition, PLS regression analysis was used as a multivariate analysis method.

  As shown in FIG. 29, in general, when analysis is performed using data of fine segment intervals, there is an increased risk that information having the same property is regarded as different, while analysis is performed using data of coarse segment intervals. Doing so increases the risk that information of different nature will be considered the same. In the former example, it is determined that even the signal of the bacteria stained with dye A is slightly different in brightness, and in the latter example, the bacteria stained with dye C and chlorophyll fluorescence are also determined to be the same. Will be. It is necessary to optimize the interval (class) of the histogram between these extremes.

  In the present embodiment, through the processes of steps SB-1 to SB-7, the optimum segment interval and biological number model that minimize the standard deviation of the error are determined (step SB-8). Here, FIG. 30 shows an equal value indicating an error obtained when the present embodiment is applied using the voltage values of the channels A and C in the histogram when the voltage of the channel B is not detected (B = 0). FIG. In the figure, the horizontal axis indicates the interval between histogram sections of channel A, and the vertical axis indicates the interval between histogram sections of channel C.

  As shown in FIG. 30, the standard deviation of the error of the obtained number model varies greatly depending on the combination of the intervals of the histogram sections of channels A and C. By the optimization processing of this embodiment, a combination of segment intervals that minimizes the error is found. Here, FIG. 31 is a diagram comparing a histogram when the viable cell count estimation error is minimum and a histogram when the viable cell count estimation error is maximum. FIG. 32 is a graph showing the difference between an estimated value and an actual measurement value of the living body number model when the viable cell count estimation error is minimum, and the living body number model when the viable cell count estimation error is maximum. . The horizontal axis of the graph shows the actual value of the viable cell count by the culture method, and the vertical axis shows the estimated value by each biological number model.

  As shown in FIG. 31 and FIG. 32, it is confirmed that the plot of the measured value and the estimated value is close to the straight line with the slope 1 and the error is reduced when the living body number model is generated with the optimum combination of the histogram division intervals. It was done. Here, FIG. 33 is a graph showing the viable cell count estimation error at the time of model creation and verification at each sampling interval of the histogram. The error at the time of model verification is an average value in 100 verifications.

  As shown in FIG. 33, in this embodiment, accurate model evaluation can be performed by repeatedly performing model verification 100 times by the cross-validation method. In FIG. 33, it can be expected that an ideal number model is obtained when the error at the time of model verification is the smallest (near the interval of 1.15). Here, FIG. 34 is a graph compared with the conventional method in each of the biological number model when the error is the largest and the mathematical model when the error is the smallest in FIG.

  As shown in the left diagram of FIG. 34, in the living body number model when the error is the largest, the number of false positives increases when the number of living bacteria is small. That is, even if the histogram interval is too small, false positives increase. As shown in the right diagram of FIG. 34, the biological number model generated at an appropriate histogram interval has few false positives even when the number of viable bacteria is small, and is plotted at a location close to an ideal straight line with a slope of 1. confirmed.

  Above, description of an example of the optimization process in a present Example is finished.

[Preprocessing]
An example of pre-processing in the present embodiment will be described with reference to FIGS.

  In the present embodiment, the spectrum preprocessing is a spectral shape conversion technique used in near infrared spectroscopy, for example, considering a histogram as a spectrum, centering, scaling, smoothing, differentiation, Refers to standardization. This preprocessing has the effect of reducing noise contained in the spectrum, enhancing information contained in the spectrum, and adjusting the spectrum scale between different samples. In the present embodiment, the voltage values of a plurality of channels are converted into a histogram, and a combination of them is regarded as a spectrum, and preprocessing is applied to this spectrum. In the following examples, the number of viable bacteria was estimated by PLS regression using the voltage values of channels A and C when the voltage of channel B was not detected (when B = 0) in the histogram.

  FIG. 35 is a graph (right diagram) showing an error between a histogram (left diagram) when no preprocessing is performed and an estimated value based on the generated biological number model. As shown in FIG. 35, when pre-processing was not performed, the viable cell count estimation accuracy of a determination coefficient = 0.957 and an error standard deviation = 0.35 log (cells / ml) was obtained.

  FIG. 36 is a graph (right diagram) showing a histogram (left diagram) in the case where the preprocessing of centering is performed (left diagram) and an error of an estimated value based on the generated number model. As shown in FIG. 36, as a result of the preprocessing of the centering, the viable cell count estimation accuracy of the determination coefficient = 0.975 and the standard deviation of error = 0.26 log (cells / ml) was obtained.

  FIG. 37 is a graph (right diagram) showing a histogram (left diagram) in the case of performing preprocessing for scaling (left diagram) and an error of an estimated value based on the generated number model. As shown in FIG. 37, as a result of the pre-processing of the scaling, the viable cell count estimation accuracy of determination coefficient = 0.975 and standard deviation of error = 0.28 log (cells / ml) was obtained.

  FIG. 38 is a graph (right diagram) showing a histogram (left diagram) in the case of performing smoothing preprocessing and an error of an estimated value based on the generated number model. As shown in FIG. 38, as a result of pre-processing for smoothing, the viable cell count estimation accuracy of determination coefficient = 0.978 and standard deviation of error = 0.26 log (cells / ml) was obtained.

  FIG. 39 is a graph (right diagram) showing an error in an estimated value based on a histogram (left diagram) and a generated biological number model when pre-processing of first derivative (first derivative) is performed. As shown in FIG. 39, as a result of pre-processing of the first derivative, a viable cell count estimation accuracy of a determination coefficient = 0.975 and an error standard deviation = 0.28 log (cells / ml) was obtained.

  FIG. 40 is a graph (right diagram) showing an error of an estimated value based on a histogram (left diagram) and a generated biological number model when preprocessing of second derivative (second derivative) is performed. As shown in FIG. 40, as a result of pre-processing of the second derivative, a viable cell count estimation accuracy of a determination coefficient = 0.976 and an error standard deviation = 0.28 log (cells / ml) was obtained.

  FIG. 41 is a graph (right diagram) showing an error between a histogram (left diagram) and an estimated value based on the generated number model when normalization is performed. As shown in FIG. 41, as a result of the pre-processing for standardization, the viable cell count estimation accuracy of a determination coefficient = 0.777 and an error standard deviation = 0.27 log (cells / ml) was obtained.

  FIG. 42 is a graph (left diagram) and a graph (right diagram) showing an error in an estimated value by a generated biological number model when the standardization, second-order differentiation, centering, and scaling preprocessing are performed in combination. is there. As shown in FIG. 42, as a result of combining standardization, second-order differentiation, centering, and scaling preprocessing, a determination coefficient = 0.983 and a standard deviation of error = 0.23 log (pieces / ml) are obtained. Bacteria count estimation accuracy was obtained.

  Above, description of the example of the pre-processing in this embodiment is finished.

[Multivariate analysis processing]
An example of multivariate analysis processing in the present embodiment will be described with reference to FIGS.

In the present Example, the sample which added the solution which culture | cultivated E. coli with the tea drink was used for data. A dilution series of the number of viable bacteria was prepared by diluting the concentration of the E. coli solution to 1/10 ³ to 1/10 ⁸ . Then, the actual number of viable bacteria was counted by a culture method, and all 68 samples were measured by a flow cytometer. In this embodiment, the voltage values of channels A and C when the voltage of channel B is not detected (when B = 0) are used in the histogram.

  Here, in this embodiment, various multivariate analysis methods can be used (refer to JP 2010-185719 regarding the multivariate analysis method). As an example, unsupervised pattern recognition methods include pattern classification methods such as principal component analysis, cluster analysis, and self-organizing maps. This unsupervised pattern recognition method is a method for visualizing the characteristics of explanatory variables (in this case, data obtained by spectralizing a histogram), and does not use objective variables (viable cell counts, etc.). Here, FIG. 43 is a diagram illustrating an example in which multivariate analysis is performed by principal component analysis.

  As shown in FIG. 43, clusters were formed for each E. coli addition concentration, and it was confirmed that discrimination of viable cell addition concentration and quantification of viable cell count by discriminant analysis and quantitative analysis became possible. FIG. 44 is a diagram showing an example in which multivariate analysis was performed by cluster analysis.

As shown in FIG. 44, it was confirmed that a group (low concentration) having an E. coli addition concentration of 1/10 ⁷ or less and a cluster other than that (high concentration) formed distinctly different, and both could be discriminated.

  In another example of the multivariate analysis method that can be used in the present embodiment, as a discriminating method with supervision, linear discriminant analysis, Partial Last Squares (PLS) discriminant analysis, Soft Independent Modeling of Class Analysis (SIMCA), Support Vector machine (SVM) discriminant analysis and the like can be applied. This supervised discrimination method is a method for constructing a model for discriminating a qualitative objective variable (presence / absence of viable bacteria) from explanatory variables. In the cross tabulation of the discrimination results based on the number model obtained by the SVM discriminant analysis, it is possible to estimate the low concentration with 100% accuracy when the measured value is low (30 samples), and when the measured value is high ( 38 samples), it was possible to estimate a high concentration with 100% accuracy. Here, FIG. 45 is a diagram showing an example in which multivariate analysis is performed by PLS discriminant analysis.

  In the cross tabulation of the discrimination results based on the number model obtained by the PLS discriminant analysis, it is possible to estimate the low concentration with 100% accuracy when the measured value is low (30 samples), and when the measured value is high ( 38 samples), it was possible to estimate a high concentration with 100% accuracy. Moreover, FIG. 46 is a figure which shows the Example which performed the multivariate analysis by SIMCA.

  In the cross tabulation of the discrimination result by the number model obtained by SIMCA, it is possible to estimate the low concentration with 100% accuracy when the measured value is low (30 samples), and when the measured value is high (38 samples) ) To a high concentration with 100% accuracy.

  In another example of the multivariate analysis method that can be used in the present embodiment, as a supervised regression analysis method, multiple regression analysis, PLS regression analysis, SVM regression analysis, locally weighted regression (local weighted regression), or the like is used. Can be applied. This supervised regression analysis method is a method for constructing a model for quantitatively estimating a quantitative objective variable (viable cell count) from explanatory variables. Here, FIG. 47 is a diagram showing an example in which multivariate analysis was performed by SVM regression analysis.

  As shown in FIG. 47, the SVM regression analysis yielded a biological number model with little error from the actual measurement value at both low and high concentrations (decision coefficient = 0.980, standard deviation of error = 0.19 log). (Pieces / ml)). FIG. 48 is a diagram illustrating an example in which multivariate analysis is performed by LWR (local weighted regression).

  As shown in FIG. 48, LWR (local weighted regression) was used to obtain a biological number model with a small error from an actual measurement value at both low and high concentrations. Determination coefficient = 0.975, standard deviation of error = 0.04 log (pieces / ml).

  As described above, it was confirmed that information on the number of viable cells or the number of viable cells in data can be visualized by performing multivariate analysis using pattern recognition methods such as principal component analysis and cluster analysis. . In addition, it was confirmed that the number of viable bacteria can be determined by performing multivariate analysis by discriminant analysis such as linear discriminant analysis, PLS discriminant analysis, SIMCA, SVM discriminant analysis and the like. It was also confirmed that the viable cell count could be estimated by performing multivariate analysis by regression analysis such as multiple regression analysis, PLS regression analysis, SVM regression analysis, LWR (local weighted regression).

  Above, description of the Example in this embodiment is finished.

[Other embodiments]
Although the embodiments of the present invention have been described so far, the present invention is not limited to the above-described embodiments, but can be applied to various different embodiments within the scope of the technical idea described in the claims. It may be implemented.

  For example, the case where the biological number model generation apparatus 100 performs processing in a stand-alone form has been described as an example, but the biological number model generation apparatus 100 is a client terminal (a separate housing from the biological number model generation apparatus 100). Processing may be performed in response to a request from the client, and the processing result may be returned to the client terminal.

  In addition, among the processes described in the embodiment, all or part of the processes described as being automatically performed can be performed manually, or the processes described as being performed manually can be performed. All or a part can be automatically performed by a known method.

  In addition, unless otherwise specified, the processing procedures, control procedures, specific names, information including registration data for each processing, parameters such as search conditions, screen examples, and database configurations shown in the above documents and drawings Can be changed arbitrarily.

  In addition, regarding the biological number model generation apparatus 100, each illustrated component is functionally conceptual and does not necessarily need to be physically configured as illustrated.

  For example, all or some of the processing functions provided in each device of the biometric model generation device 100, particularly the processing functions performed by the control unit 102, are performed by a CPU (Central Processing Unit) and the CPU. It may be realized by a program to be interpreted and executed as hardware by wired logic. The program is recorded on a non-transitory computer-readable recording medium including a programmed instruction for causing a computer to execute the method according to the present invention, which will be described later. Can be read. That is, in the storage unit 106 such as a ROM or an HDD (Hard Disk Drive), a computer program for giving instructions to the CPU and performing various processes in cooperation with the OS (Operating System) is recorded. This computer program is executed by being loaded into the RAM, and constitutes a control unit in cooperation with the CPU.

  The computer program may be stored in an application program server connected to the biological model generation apparatus 100 via an arbitrary network 300, and the computer program may be downloaded in whole or in part as necessary. Is also possible.

  In addition, the program according to the present invention may be stored in a computer-readable recording medium, and may be configured as a program product. Here, the “recording medium” is any memory card, USB memory, SD card, flexible disk, magneto-optical disk, ROM, EPROM, EEPROM, CD-ROM, MO, DVD, Blu-ray Disc, etc. Of “portable physical media”.

  The “program” is a data processing method described in an arbitrary language or description method, and may be in any format such as source code or binary code. The “program” is not necessarily limited to a single configuration, but is distributed in the form of a plurality of modules and libraries, or in cooperation with a separate program represented by an OS (Operating System). Including those that achieve the function. Note that a well-known configuration and procedure can be used for a specific configuration for reading a recording medium, a reading procedure, an installation procedure after reading, and the like in each device described in the embodiment.

  Various databases and the like (detection intensity file 106a to biological number model file 106c) stored in the storage unit 106 include memory devices such as RAM and ROM, fixed disk devices such as hard disks, flexible disks, and storage means such as optical disks. It stores various programs, tables, databases, web page files, etc. used for various processes and website provision.

  In addition, the biological model generation apparatus 100 may be configured as an information processing apparatus such as a known personal computer or workstation, or may be configured by connecting an arbitrary peripheral device to the information processing apparatus. In addition, the biological number model generation apparatus 100 may be realized by installing software (including programs, data, and the like) that realizes the method of the present invention in the information processing apparatus.

  Furthermore, the specific form of distribution / integration of the devices is not limited to that shown in the figure, and all or a part of them may be functional or physical in arbitrary units according to various additions or according to functional loads. Can be distributed and integrated. That is, the above-described embodiments may be arbitrarily combined and may be selectively implemented.

  As described above in detail, according to the present embodiment, a biological number model generation device, a biological number model generation method, a cytometer, a program, and a program that can accurately and quickly measure the number of living bodies, A recording medium can be provided. Therefore, it is extremely useful for quality control at food production sites such as tea beverage manufacturing factories, and inspection of foods containing mineral water and animal protein. In particular, the present invention can be implemented in a commercially available flow cytometry apparatus, and data can be acquired by a commercially available machine (Cassette Lab ONE, Hitachi Engineering & Service). In addition, it is extremely useful in various fields such as quality control using a flow cytometry apparatus, an imaging cytometry apparatus, and the like, medical treatment, pharmaceuticals, drug discovery, biological research, and clinical examinations.

DESCRIPTION OF SYMBOLS 100 Biological number model production | generation apparatus 102 Control part 102a Data acquisition part 102b Histogram generation part 102c Multivariate analysis part 102d Optimization part 102f Biological number determination part 104 Communication control interface part 106 Storage part 106a Detection intensity file 106b Histogram file 106c Biological number model File 108 Input / output control interface unit 110 Measuring unit 112 Input unit 114 Output unit 200 External device 300 Network

Claims (11)

In a biological number model generation device including at least a storage unit and a control unit,
The storage unit
A detection intensity storage means for storing detection intensity data obtained by cytometry using a fluorescent label in which the degree of interaction with the living body and the emission intensity corresponding to the degree of the living body depending on whether the living body is alive or dead,
With
The control unit
Histogram generation means for generating a histogram indicating the frequency of each detection intensity based on the detection intensity data;
A multivariate analysis is performed using frequency values in each section of the histogram as explanatory variables, and a biological number model that is a mathematical expression for determining the presence or absence of the living body or the number of living bodies is obtained using the coefficients for the obtained explanatory variables. Variable analysis means;
A living body number model generation device characterized by comprising: In the living body number model generating device according to claim 1,
The detected intensity storage means is
Storing the detection intensity data obtained by cytometry using a plurality of fluorescent labels, indicating the detection intensity of the fluorescence emitted by each of the fluorescent labels,
The histogram generation means includes
Based on the detected intensity data, a histogram showing the frequency for each category is generated in the plurality of fluorescent labels,
The multivariate analysis means is:
The frequency for each category of the fluorescent label is the explanatory variable,
A biological number model generation device characterized by the above. In the living body number model generation device according to claim 1 or 2,
The detected intensity storage means is
Storing the detection intensity data obtained by cytometry using a plurality of fluorescent labels, indicating the detection intensity of the fluorescence emitted by each of the fluorescent labels,
The histogram generation means includes
Based on the detected intensity data, a histogram is generated that indicates the frequency of each of the other fluorescent labels for each of the sections when the detected intensity of one of the fluorescent labels is within a predetermined range. about,
A biological number model generation device characterized by the above. In the living body number model generating device according to any one of claims 1 to 3,
The detected intensity storage means is
Storing the detection intensity data obtained by cytometry using a plurality of fluorescent labels, indicating the detection intensity of the fluorescence emitted by each of the fluorescent labels,
The histogram generation means includes
Adding, subtracting, multiplying / dividing between the detected intensity data of the fluorescent label simultaneously measured, and generating a histogram showing the frequency for each category of the added / subtracted / divided values;
A biological number model generation device characterized by the above. In the living body number model generation device according to any one of claims 1 to 4,
The control unit
By controlling the interval between the sections of the histogram repeatedly generated by the histogram generation means while verifying an error from the actual measurement value for the biological number model obtained by the multivariate analysis means, Model optimization means for optimizing the biological number model,
A living body number model generating apparatus, further comprising: In the living body number model generating device according to any one of claims 1 to 5,
The histogram generation means includes
Converting the detected intensity into a logarithm based on the detected intensity data, and generating a histogram showing the frequency for each segment of the converted logarithm;
A biological number model generation device characterized by the above. In the living body number model generation device according to any one of claims 1 to 6,
The histogram generation means includes
Centering, scaling, smoothing, differentiating and / or standardizing the generated histogram as a spectrum,
A biological number model generation device characterized by the above. In the living body number model generation device according to any one of claims 1 to 7,
The multivariate analysis means is:
Performing the above multivariate analysis using unsupervised pattern recognition techniques, supervised discriminant analysis techniques, or supervised regression analysis techniques,
A biological number model generation device characterized by the above. A biological number model generation method executed in a computer including at least a storage unit and a control unit,
The storage unit
A detection intensity storage means for storing detection intensity data obtained by cytometry using a fluorescent label in which the degree of interaction with the living body and the emission intensity corresponding to the degree of the living body depending on whether the living body is alive or dead,
With
Executed in the control unit,
Based on the detection intensity data, a histogram generation step for generating a histogram indicating the frequency of each detection intensity category;
A multivariate analysis is performed using frequency values in each section of the histogram as explanatory variables, and a biological number model that is a mathematical expression for determining the presence or absence of the living body or the number of living bodies is obtained using the coefficients for the obtained explanatory variables. A variable analysis step;
A method for generating a biological number model, comprising:   A cytometer that determines the number of living organisms or the presence or absence of the living organism from the detected intensity data, using the biological number model generated by the living organism number model generating method according to claim 9. A program for causing a computer including at least a storage unit and a control unit to execute the program,
The storage unit
A detection intensity storage means for storing detection intensity data obtained by cytometry using a fluorescent label in which the degree of interaction with the living body and the emission intensity corresponding to the degree of the living body depending on whether the living body is alive or dead,
With
In the control unit,
Based on the detection intensity data, a histogram generation step for generating a histogram indicating the frequency of each detection intensity category;
A multivariate analysis is performed using frequency values in each section of the histogram as explanatory variables, and a biological number model that is a mathematical expression for determining the presence or absence of the living body or the number of living bodies is obtained using the coefficients for the obtained explanatory variables. A variable analysis step;
A program for running