Protein recovery determination system, protein recovery determination method, and recording medium

Abstract

A protein recovery determination system includes: a first determination unit that inputs first image data obtained by imaging a culture of a virus-infected cell population in time series as first input data, and determines, as first output data, whether the virus-infected cell population is in a state suitable for protein recovery; and an acquisition unit that acquires image data by capturing an image of the inside of the culture vessel in time series, the first determination unit performs determination using a first model in which the actual performance value of the first input data and the actual performance value of the first output data are used as first training data, and the first training data are used as second training data. And deeply learning the first neural network to obtain a learned model.

Description

Technical Field

[0001] The present invention relates to a protein recovery determination system, a protein recovery determination method, and a recording medium. Background Art

[0002] In the production of proteins such as antigens, there is a method of using cells that are obtained by introducing the amino acid sequence of a protein into cells through a specific virus having the amino acid sequence of the target protein and then amplifying the cells (see, for example, Non-Patent Document 1). In this method, for example, as Figure 17 shown, a part of the cultured host cells is proliferated, and after a predetermined time, a specific virus is allowed to infect the host cells. After the virus infection, when it is determined that the host cells infected with the virus have sufficiently produced the protein, the protein is recovered. Figure 17 is a diagram for explaining a protein production process of a conventional method. Conventionally, an operator makes a determination of such a recovery timing based on tacit knowledge from the results observed by the operator using a microscope (see Non-Patent Documents 1 to 2).

[0003] Prior Art Documents

[0004] Non-Patent Documents

[0005] Non-Patent Document 1: "Appl Microbiol Biotechnol.", January 2019 (2019 Jan); 103(1): 113 - 123. "Engineering of the baculovirus expression system for optimized protein production."

[0006] Non-Patent Document 2: "Protein Expr Purif.", October 2021 (2021 Oct); 186: 105924. "A review of alternative promoters for optimal recombinant protein expression in baculovirus-infected insect cells" Summary of the Invention

[0007] Problems to be Solved by the Invention

[0008] However, in the prior art, since the operator judges the recovery timing by observing images, there are cases where the recovery timing varies for each operator. For example, when cell death is caused by virus infection, there are problems such as the decomposition of the target protein caused by the proteasome contained in the cell, cell death caused by abnormal infection due to the release of the virus in the cell, and the purification of the target protein becoming difficult due to the contamination of the culture solution caused by components other than the target protein contained in the cell. Therefore, it is required to recover the protein at an appropriate timing (at a stage where there is little cell death and the protein is sufficiently produced from the cells).

[0009] The present invention has been made in view of the above problems, and an object thereof is to provide a protein recovery determination system, a protein recovery method, and a recording medium recording a learned model that can appropriately determine the timing of recovering a protein.

[0010] Technical means for solving the problem

[0011] The present invention includes the following embodiments.

[0012] (1) A protein recovery determination system according to an embodiment of the present invention for achieving the above object includes: a first determination unit that inputs first image data obtained by photographing the culture of a cell population infected with a virus in time series as first input data and determines whether the cell population infected with the virus is in a state suitable for recovering the protein as first output data; and an acquisition unit that acquires image data by photographing the inside of a culture container in time series, and the first determination unit makes a determination using a first model, and the first model is a learned model obtained by deep learning a first neural network using the actual values of the first input data and the actual values of the first output data as first training data.

[0013] (2) The protein recovery determination system according to (1) above further includes a second determination unit. In the second determination unit, second image data obtained by photographing the subculture of a cell population not infected with a virus in time series is input as second input data, and it is determined whether it is a state suitable for infecting the cell population not infected with the virus with the virus as second output data. The second determination unit makes a determination using a second model, and the second model is a learned model obtained by deep learning a second neural network using the actual values of the second input data and the actual values of the second output data as second training data, and the first determination unit makes a determination after the second determination unit makes a determination.

[0014] (3) The protein recovery determination system according to (1) or (2) above, wherein the cells of the cell population are adherent cells.

[0015] (4) In the protein recovery determination system according to (3) above, the adherent cells are insect cells.

[0016] (5) According to the protein recovery determination system according to (4) above, wherein the insect cells are one of Sf (Spodoptera frugiperda) 9 cells, Sf21 cells, Tni (Trichoplusia ni) cells, and H5 (High Five) cells.

[0017] (6) The protein recovery determination system according to (1) or (2) above further includes a first learning unit for learning the first neural network.

[0018] (7) The protein recovery determination system according to (2) above further includes a second learning unit for learning the second neural network.

[0019] (8) In the protein recovery determination system according to (2) above, the acquisition unit acquires information indicating whether the cell population is in a virus-uninfected state or a virus-infected state, and based on the information acquired by the acquisition unit, selects one of the second determination unit and the first determination unit to be used.

[0020] (9) A recording medium according to another embodiment of the present invention records a learned model, which is obtained by deep learning a first neural network using first input data and first output data as first training data. The first input data is first image data obtained by photographing the culture of a virus-infected cell population in time series, and the first output data is data for determining whether the virus-infected cell population is in a state suitable for protein recovery.

[0021] (10) A recording medium according to another embodiment of the present invention records a learned model, which is obtained by deep learning a second neural network using second input data and second output data as second training data. The second input data is second image data obtained by photographing the scale-up culture of a virus-uninfected cell population in time series, and the second output data is data for determining whether it is in a state suitable for infecting the virus-uninfected cell population with a virus.

[0022] (11) In the protein recovery determination method according to another embodiment of the present invention, first image data is obtained by photographing the culture of a cell population infected with a virus in time series, and the first image data is input as first input data into a first determination unit. The first determination unit determines whether the cell population infected with the virus is in a state suitable for protein recovery as first output data. The first determination unit makes the determination using a first model, and the first model is a learned model obtained by deep learning a first neural network using the actual values of the first input data and the actual values of the first output data as first training data.

[0023] (12) In the protein recovery determination method according to (11) above, before making the determination as to whether the cell population is in a state suitable for protein recovery, second image data obtained by photographing the scale-up culture of a cell population not infected with the virus in time series is obtained, and the second image data is input as second input data into a second determination unit. The second determination unit determines whether it is in a state suitable for infecting the cell population not infected with the virus with the virus as second output data. The second determination unit makes the determination using a second model, and the second model is a learned model obtained by deep learning a second neural network using the actual values of the second input data and the actual values of the second output data as second training data.

[0024] Effects of the Invention

[0025] According to (1) to (12) above, the timing of protein recovery can be appropriately determined. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a diagram showing a functional configuration example of the protein recovery determination system according to the embodiment.

[0027] Figure 2 is a diagram showing an example of an image captured by the imaging device at regular intervals (captured using bright-field observation method).

[0028] Figure 3 is an explanatory diagram of an example of a protein manufacturing process based on cell engineering.

[0029] Figure 4 is an example of the production of the first model by the first learning unit.

[0030] Figure 5 is a diagram showing an example of the determination by the first determination unit.

[0031] Figure 6 is when the host cell is an H5 (High Five) cell and at the initial stage of virus infection (for example Figure 3 ​​​​​​The captured image (captured using bright field microscopy) during the period from time t14 to time t15).

[0032] Figure 7 is a captured image (captured using bright field microscopy) during the period between time t16 and time t17 when the host cell is an H5 (High Five) cell and is in a state suitable for recovery (e.g., Figure 3 ).

[0033] Figure 8 is a captured image example (captured using bright field microscopy) during the period from time t14 to time t15 when the host cell is an Sf9 cell and is in the initial stage of virus infection (e.g., Figure 3 ).

[0034] Figure 9 is a captured image example (captured using bright field microscopy) during the period between time t16 and time t17 when the host cell is an Sf9 cell and is in a state suitable for recovery (e.g., Figure 3 ).

[0035] Figure 10 is a flowchart of the processing sequence of the protein recovery determination system.

[0036] Figure 11 is a diagram showing an example of the functional structure when the protein recovery determination system in the embodiment has a protein recovery determination device.

[0037] Figure 12 is an example of the production of the second model based on the second learning unit.

[0038] Figure 13 is a diagram showing an example of the determination of the second determination circuit in the embodiment.

[0039] Figure 14 is a captured image example (captured using bright field microscopy) during the period from time t13 to time t14 when the host cell is an H5 (High Five) cell and is in the infection period (e.g., Figure 4 ).

[0040] Figure 15 is a captured image example (captured using bright field microscopy) during the period from time t13 to time t14 when the host cell is an Sf9 cell and is in the infection period (e.g., Figure 4 ).

[0041] Figure 16 is a flowchart of the processing sequence performed by the protein recovery determination system of the second embodiment.

[0042] Figure 17 is a diagram for explaining the antigen manufacturing process of the conventional method.​​​​​​​​​​​

[0043] Figure 18 is a graph showing the average correct answer rate and the average antibody production of the determination based on the first model and the second model. Detailed implementation mode

[0044] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In addition, in the accompanying drawings used for the following description, the scales of the respective components are appropriately changed so that each component can be recognized in size.

[0045] In addition, in all the accompanying drawings used to illustrate the embodiments, parts having the same function are denoted by the same reference numerals, and repeated descriptions are omitted.

[0046] In addition, "based on XX" in this specification means "at least based on XX", and also includes cases where other elements are also based on in addition to based on XX. In addition, "based on XX" is not limited to the case of directly using XX, and also includes cases based on the case where XX has been calculated or processed. "XX" is an arbitrary element (for example, arbitrary information).

[0047] In addition, "host cell" and "cell of the cell population" in this specification have the same meaning.

[0048] In addition, "image data" in this specification represents a two-dimensional image.

[0049] In addition, "model" and "trained model" in this specification have the same meaning.

[0050] Figure 1 is a diagram showing a structural example of the protein recovery determination system 5 of the present embodiment. As Figure 1 shown, the protein recovery determination system 5 includes, for example: a protein recovery determination device 1, a photographing device 2, and an external device 3. The protein recovery determination device 1, the photographing device 2, and the external device 3 are connected by a wired or wireless line.

[0051] ​The protein recovery determination device 1 includes, for example: an acquisition unit 101, a first determination unit 102, an optional first learning unit 106, an optional storage unit 108, and an output unit 109. The first determination unit 102 executes the first model 104. In addition, the first determination unit 102 may also have the first model 104. The protein recovery determination device 1 is composed of hardware of an information processing device such as a personal computer (PC) using a circuit such as an integrated circuit (IC). The protein recovery determination device 1 is implemented by reading a program that implements a model with a specific function stored in the storage unit 108 etc., and having each unit such as the acquisition unit 101, the first determination unit 102, and the first learning unit 106 execute this program.

[0052] Examples of the acquisition unit 101, the first determination unit 102, the first learning unit 106, and the output unit 109 include processors such as a central processing unit (CPU), a microprocessor unit (MPU), a system on chip (SoC), and a dedicated circuit that execute programs to perform various controls, and ICs having a processor. Examples of the storage unit 108 include memories such as a random access memory (RAM), a read only memory (ROM), a hard disk drive (HDD), and a solid state disk (SSD). The storage unit 108 may also be an external memory of the protein recovery determination system 5.

[0053] The imaging device 2 captures images of the virus-infected cell population in the culture vessel at regular intervals. In addition, the imaging device 2 and the protein recovery determination device 1 are connected to each other via a wired or wireless line. In addition, the imaging time is added to the captured images. Furthermore, identification information for identifying the imaging object may also be associated with the captured images. The imaging device 2 includes an imaging mechanism 21 and a culture vessel 22. Examples of the imaging device 2 include, for example, the device described in Japanese Patent Laid-Open No. 2020-156419. Examples of the imaging mechanism 21 include imaging elements such as a charge coupled device (CCD) and a complementary metal oxide semiconductor (CMOS), and digital cameras having an imaging lens etc.

[0054] The inner surface of the culture vessel 22 can be of any shape, but is preferably planar. In this case, adherent culture of the cell population can be performed on the inner surface of the plane. In addition, the inner surface of the culture vessel 22 can be coated (for example, coated with an extracellular matrix that serves as a cell scaffold) to a surface suitable for adherent culture.

[0055] The photographing device 2 is preferably immobilized in position with respect to the inner surface of the culture vessel 22. When the inner surface of the culture vessel 22 is planar, the photographing direction of the cell population in the culture vessel 22 by the photographing device 2 is preferably a direction substantially perpendicular to the plane.

[0056] The acquisition unit 101 acquires the image (image data a1) photographed by the photographing device 2 at every prescribed time (for example, once every 0.5 hours to 10 hours). The acquisition unit 101 can store the acquired image in the storage unit 108. The image acquired by the acquisition unit 101 can be used as first training data or second training data when producing the first model 104 or the second model 105 (described later).

[0057] The first determination unit 102 executes the first model 104, inputs the first image data (image data a11) into the first model 104 (learned model a12) acquired from the storage unit 108, and determines whether the cell population infected with the virus, which is the first output data, is in a state suitable for protein recovery.

[0058] The first image data preferably includes 10% or more, 20% or more, 30% or more, 40% or more, 50% or more of the culture surface of the cell population, and the upper limit value is 60% or less, 70% or less, 80% or less, 90% or less, or 100% or less. In order to include the number of cells in the cell population in the first image data, the first image data acquired at the same time can be divided into a plurality of parts. In order to divide into a plurality of parts, the number of images photographed at the same time can be multiple (for example, 2 to 100 images).

[0059] As the image data, examples include: image data obtained by photographing using bright field observation method, dark field observation method, phase contrast observation method, differential interference observation method, polarized light observation method, relief phase contrast observation method, fluorescence observation method, and MIX (mixed) observation method. Since it is preferably not to stain the host cells and the shape change of the host cells due to virus infection of the host cells is large, it is easy to extract as a feature amount inherent in the image data in deep learning, and there are also colorless and transparent host cells. Therefore, among these image data, image data obtained by photographing using bright field observation method, dark field observation method, phase contrast observation method, differential interference observation method, and relief phase contrast observation method are preferred.

[0060] The first model 104 is a model that determines a state suitable for protein recovery based on image data of a cell population after virus infection in a culture vessel by performing deep learning based on a neural network (the first neural network) by the first learning unit 106. Since the first input data is unstructured data, deep learning can improve the correct answer rate of the first model. The first model 104 is pre-learned and produced by the first learning unit 106. Examples of learning methods based on deep learning include: Deep Neural Network (DNN), Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Generative Adversarial Networks (GAN), as well as combinations of two or more of them.

[0061] In many cases, host cells locally change their morphology over time due to virus infection, and the images of host cells contain limited regions (local receptive fields). In addition, the first input data is image data of a cell population containing multiple host cells. Therefore, as a learning method based on deep learning, CNN is preferably used. CNN is a learning method that can share weights with respect to local receptive fields.

[0062] The first model 104 can be a model obtained by learning by the first learning unit 106 of the protein recovery determination device 1, or can be a model obtained by learning by another device similar to the protein recovery determination device 1.

[0063] The first model 104 can use the first image data (image data a14) acquired by the acquisition unit 101 as input data, and the data obtained by evaluating the determination of the first determination unit 102 as output data for training data, and can be re-learned by the first learning unit 106 and updated by a first update unit (not shown) included in the first learning unit 106. Regarding re-learning, it can be performed each time the acquisition unit 101 acquires the first image data, or after the acquisition unit 101 acquires the first image data, it can be recorded in the storage unit 108 and performed at any time (for example, once every 1 month to 6 months).

[0064] The storage unit 108 stores, for example, programs, thresholds, feature amounts used in various processes of the protein recovery determination device 1 or the protein recovery determination device 1A described later, image data acquired by the acquisition unit 101, a learned model a15 obtained by learning by the first learning unit 106, and the like.

[0065] The output unit 109 outputs, as data a2 indicating the determination result, information (data a13 indicating the determination result) indicating the result determined by the first determination unit 102 to the external device 3. The output unit 109 may also output image data determined to be in a state suitable for protein recovery, the estimated protein recovery amount, or the estimated amount of protein (HCP, host cell protein) from the host cell to the external device 3.

[0066] The external device 3 presents the information output by the protein recovery determination device 1. The external device 3 is, for example, an image display device, a printing device, a tablet terminal, a smart phone, a personal computer, or the like. The external device 3 and the protein recovery determination device 1 are connected to each other by a wired or wireless line.

[0067] Figure 2 It is a diagram showing an example of an image captured by the imaging device 2 at regular intervals (captured using bright field microscopy). In addition, Figure 2 The host cell is H5 (High Five). The image g11 is an image captured at the Figure 3 time t11. The image g12 is an image captured at a time between the time t12 and the time t13. The image g13 is an image captured at a time between the time t12 and the time t13. The image g14 is an image captured at the time t14. The image g15 is an image captured at a time between the time t15 and the time t16. The image g16 is an image captured at a time between the time t16 and the time t17.

[0068] In addition, Figure 2 The host cell or the captured image shown is an example, and the captured image, the host cell, etc. are not limited thereto.

[0069] Next, the host cell in the protein expression system and the infecting virus for infecting the host cell will be described. Examples of the host cell include bacterial cells, yeast cells, fungal cells, insect cells, and mammalian cells. As the infecting virus, it can be appropriately selected according to the host cell. For example, when the host cell is an insect cell, it is preferably a Nucleopolyhedrovirus.

[0070] The host cell is preferably an adherent cell, more preferably an insect cell. As long as adherent culture of adherent cells is used, since the morphology of the host cell can be fixed before and after virus infection, determination can be performed correctly. Since the morphology of the host cell infected with the virus is not as complicated as the image data of suspension culture, the learning efficiency when creating a learned model is excellent.

[0071] In the case where the host cell is an insect cell, it is characterized in that insect cells are mostly suitable for adherent culture, and the morphological changes of the host cells before and after virus infection are large. Therefore, in the case where the host cell is an insect cell, noise removal or feature quantity selection using deep learning is excellent.

[0072] Examples of insect cells include: Sf9 cells, Sf21 cells, Tni cells, or H5 (High Five) cells. Sf9 cells are a cell line derived from Spodoptera frugiperda. In addition, Tni cells are a cell line derived from Trichoplusia ni.

[0073] Figure 3 It is an explanatory diagram of an example of a protein manufacturing process based on cell engineering. In Figure 3 the horizontal axis is time.

[0074] Time t11 is the time when cells (such as H5 (High Five) cells) that are the target of virus infection of the host cell start to proliferate. The period from time t12 to time t13 is the period during the scale-up culture of the host cell. During the period from time t13 to time t14, the fusion rate increases, and it is in a state suitable for virus infection of the host cell.

[0075] Time t14 is the virus infection start time. The time after time t14 is the post-infection period. The period from time t15 to time t16 is the period during which the virus infection of the host cell is in progress. The period from time t16 to time t17 is in a state suitable for protein recovery.

[0076] Figure 4 It is an example of the production of the first model 104 of the first learning unit 106.

[0077] The first learning unit 106 produces the first model 104 for determining whether it is in a state suitable for recovery by inputting the first training data (actual value) into the first neural network. The first learning unit 106 is preferably trained for each combination of host cell and infecting virus. The combination of host cell and virus, for example, is the combination of host cell being Sf9 cells and virus being nuclear polyhedrosis virus. The training data for whether it is in a state suitable for recovery uses the judgment result of an experienced person.

[0078] The first model 104 can be a model different for each host cell. In this case, even if the infecting virus is different, as long as the host cell is the same, the same model can be used.

[0079] Figure 5It is a diagram showing a determination example of the first determination unit 102. The first determination unit 102 executes the first model 104, inputs the captured image into the first model 104 at regular intervals, and determines whether it is in a state suitable for recovery. In addition, when the first model 104 includes models for each host cell, at the time of determination, the first determination unit 102 can select the first model 104 based on the information indicating the host cell. Regarding the information indicating the host cell, for example, it can be input by an operator operating the external device 3, and the information indicating the host cell is attached to the first image data acquired by the acquisition unit 101 as identification information for identifying the imaging object.

[0080] Next, an example of a captured image before virus infection and an example of a captured image at the recovery time are shown.

[0081] Figure 6 It is a captured image (captured using bright-field microscopy) of the host cell being H5 (High Five) cells and in the early stage of virus infection (for example, Figure 3 during the period from time t14 to time t15). Figure 7 It is a captured image (captured using bright-field microscopy) of the host cell being H5 (High Five) cells and in a state suitable for recovery (for example, Figure 3 during the period between time t16 and time t17).

[0082] As Figure 6 , Figure 7 shown, when the host cell is H5 (High Five), in the early stage of virus infection, it extends a tail like a tadpole, and as the virus infection progresses, it shrinks into a circle and the outline becomes darker.

[0083] Figure 8 It is an example of a captured image (captured using bright-field microscopy) of the host cell being Sf9 cells and in the early stage of virus infection (for example, Figure 3 during the period from time t14 to time t15). Figure 9 It is an example of a captured image (captured using bright-field microscopy) of the host cell being Sf9 cells and in a state suitable for recovery (for example, Figure 3 during the period between time t16 and time t17).

[0084] As Figure 8 , Figure 9 shown, when the host cell is Sf9 cells, in the early stage of virus infection, it is spherical, and as the infection progresses, the outline becomes darker (peeling off from the adhesion surface), and as the virus infection progresses, the outline becomes even darker.

[0085] Figure 10 It is a flowchart of the processing sequence of the protein recovery determination system.

[0086] (Step S1) The imaging device 2 captures images of the cell population infected with the virus at regular intervals.

[0087] (Step S2) The acquisition unit 101 acquires the images captured by the imaging device 2 at regular intervals.

[0088] (Step S3) The first determination unit 102 inputs the first image data obtained by capturing images of the cell population infected with the virus in time series into the learned first model 104, and determines whether the cell population infected with the virus is in a state suitable for protein recovery.

[0089] (Step S4) When the first determination unit 102 determines that it is in a state suitable for recovery (Step S4: Yes (YES)), it proceeds to the process of Step S5. When the first determination unit 102 determines that it is not the recovery period (Step S4: No (NO)), it returns to the process of Step S1.

[0090] (Step S5) The output unit 109 outputs the information indicating the state suitable for protein recovery determined by the first determination unit 102 to the external device 3.

[0091] Thus, according to the present embodiment, it is possible to determine whether it is a state suitable for protein recovery. Moreover, according to the present embodiment, protein can be recovered at an appropriate timing (at a stage with few cell deaths and sufficient protein production from the cells), so the production amount of proteins based on cell engineering can be increased.

[0092] In the first model 104, since the image data obtained by capturing images in time series is used, fluctuations in time caused by operators and fluctuations caused by the imaging ability of operators (such as light quantity, focus, etc.) can be eliminated. Therefore, feature quantities can be determined using deep learning, and the accuracy rate of the output data can be improved.

[0093] In addition, in the first model 104, since time-series image data is used, subtle changes that cannot be noticed by operators can also be determined as feature quantities through deep learning, thereby improving the accuracy rate of the output data.

[0094] Figure 11 It is a diagram showing a structural example when the protein recovery determination system of the present embodiment has the protein recovery determination device 1A.

[0095] The protein recovery determination device 1A includes, for example, on the basis of the structure included in the protein recovery determination device 1, a second determination unit 103, an arbitrary second learning unit 107, and an arbitrary selection unit 110. The second determination unit 103 executes the second model 105. In addition, the second determination unit 103 may also have the second model 105.

[0096] Examples of the second determination unit 103, the second learning unit 107, and the selection unit 110 include processors such as a CPU, an MPU, an SoC, and an application-specific circuit that execute programs to perform various controls, and an IC having a processor.

[0097] The imaging device 2 captures images of the culture of host cells before virus infection in the culture vessel at regular intervals (for example, once every 0.5 hours to 10 hours).

[0098] The second determination unit 103 inputs second image data, which is second input data obtained by capturing images of a cell population that has not been infected with a virus during subculture in time series, into a second model 105 (learned model a22) obtained from the storage unit 108, and determines whether the second output data is a state suitable for infecting the virus-free cell population with a virus.

[0099] The second image data preferably includes 10% or more, 20% or more, 30% or more, 40% or more, 50% or more of the culture surface of the cell population, and the upper limit is 100%. In order to include the number of cells in the cell population in the second image data, the second image data obtained at the same time can be divided into multiple parts. In order to divide it into multiple parts, the number of images captured at the same time can be multiple (for example, 2 to 100 images).

[0100] Examples of the image data include image data obtained by capturing images using bright-field observation, dark-field observation, phase-contrast observation, differential interference observation, polarization observation, relief phase-contrast observation, fluorescence observation, and MIX (mixed) observation. Since it is preferably not to stain the host cells and the shape of the host cells changes greatly due to the formation of colonies during proliferation, it is easy to extract as an inherent feature quantity in the image data in deep learning. There are also colorless and transparent host cells. Therefore, among these image data, image data obtained by capturing images using bright-field observation, dark-field observation, phase-contrast observation, differential interference observation, and relief phase-contrast observation is preferably used.

[0101] The second model 105 is a model that is constructed by deep learning based on a neural network (second neural network) and determines the virus infection period based on the image of the cell population during subculture in the culture vessel. Since the second input data is unstructured data, deep learning can improve the correct answer rate of the second model. Examples of the learning method of deep learning include DNN, RNN, CNN, LSTM, and GAN, and combinations of two or more of them.

[0102] In the proliferation of host cells promoted by colony formation, the morphology of colonies changes in time series. Therefore, the image of the colonies of host cells contains a defined area (local receptive field). In addition, the second input data is image data including a plurality of colonies. Thus, as a learning method based on deep learning, CNN is preferably used. CNN is a learning method that can share weights with respect to the local receptive field.

[0103] The second model 105 can be a model obtained by learning by the second learning unit 107 included in the protein recovery determination device 1A, or can be a model obtained by learning by another device similar to the protein recovery determination device 1.

[0104] The second learning unit 107 can use the second image data (image data a24) acquired by the acquisition unit 101 as input data, and the data obtained by evaluating the determination of the second determination unit 103 as output data for re-learning, and update the second model 105 by a second update unit (not shown) included in the second learning unit 107. Regarding re-learning, it can be performed each time the second image data is acquired by the acquisition unit 101, or after the second image data is acquired by the acquisition unit 101, it can be recorded in the storage unit 108 and performed at any time (for example, once every 1 to 6 months). The second learning unit 107 stores the generated learned model a25 in the storage unit 108.

[0105] In addition to the information indicating the result of the determination by the first determination unit 102 (data a13 indicating the result of the determination), the output unit 109 also outputs the information indicating the result of the determination by the second determination unit 103 (data a23 indicating the result of the determination) as data a2 indicating the result of the determination to the external device 3. The output unit 109 can also output the captured image or the fusion rate determined to be in a state suitable for infection to the external device 3.

[0106] The selection unit 110 determines whether the image data a3 acquired by the acquisition unit 101 is the first image data or the second image data, and thus selects whether to use the first determination unit 102 to determine whether it is in a state suitable for protein recovery, or to use the second determination unit 103 to determine whether it is in a state suitable for infection. When the selection unit 110 selects to perform the determination using the first determination unit 102, it outputs the image data a3 acquired by the acquisition unit 101 to the first determination unit 102 as the image data a11. On the other hand, when the selection unit 110 selects to perform the determination using the second determination unit 103, it outputs the image data a3 acquired by the acquisition unit 101 to the second determination unit 103 as the image data a21.

[0107] Next, an example of the production of the second model 105 based on the second learning unit 107 will be described.

[0108] Figure 12 This is a production example of the second model based on the second learning unit 107. As Figure 12 shown, the second learning unit 107 produces the second model 105 by inputting second training data (actual values) into the second learning unit 107. In addition, the second learning unit 107 can learn for each combination of host cells and infected viruses. The determination of whether it is a suitable state for infection is the result of the judgment by an experienced person.

[0109] The second model 105 may include multiple models produced for each host cell. In this case, even if the infected viruses are different, as long as the host cells are the same, the same model can be used. For example, the host cell of the second model 105-1 may be Sf9 cells, the host cell of the second model 105-2 may be H5 (High Five) cells, the host cell of the second model 105-3 may be Sf21 cells, and the host cell of the second model 105-4 may be Tni cells.

[0110] Figure 13 This is a diagram showing a determination example of the second determination unit 103 of the present embodiment. The second determination unit 103 executes the second model, inputs the captured image into the second model 105 at regular intervals, and determines whether it is a suitable state for infection. When the second model 105 includes models for each host cell, at the time of determination, the second determination unit 103 can select the second model 105 based on the information indicating the host cell. In addition, regarding the information indicating the host cell, for example, it can be input by an operator operating the external device 3, and the information indicating the host cell is attached to the second image data acquired by the acquisition unit 101 as identification information for identifying the captured object.

[0111] Next, an example of a captured image at the infection time will be described.

[0112] Figure 14 This is an example of a captured image (captured using the bright field observation method) of H5 (High Five) cells as host cells during the infection period (for example, Figure 4 during the period from time t13 to time t14).

[0113] Figure 15 This is an example of a captured image (captured using the bright field observation method) of Sf9 cells as host cells during the infection period (for example, Figure 4 during the period from time t13 to time t14).

[0114] Next, an example of the processing sequence performed by the protein recovery determination system 5A will be described.

[0115] Figure 16 This is a flowchart of the processing sequence performed by the protein recovery determination system of the present embodiment.

[0116] (Step S21) The imaging device 2 takes images of the culture of the virus-uninfected cell population at regular intervals.

[0117] (Step S22) The acquisition unit 101 acquires the images taken by the imaging device 2 at regular intervals.

[0118] (Step S23) The second determination unit 103 inputs the second image data obtained by taking images of the culture of the virus-uninfected cell population in time series into the learned second model 105, and determines the recovery time based on the information indicating whether it is a state suitable for virus infection.

[0119] (Step S24) When the second determination unit 103 determines that it is the infection time (Step S24: Yes), it proceeds to the process of Step S25. When the second determination unit 103 determines that it is not the infection time (Step S24: No), it returns to the process of Step S21.

[0120] (Step S25) The output unit 109 outputs the information indicating the time period determined by the second determination unit 103 as being suitable for virus infection to the external device 3. Next, after the virus is introduced into the culture container, the Figure 10 flowchart of the processing sequence shown is executed.

[0121] After the second determination unit 103 of the protein recovery determination device 1A makes a determination, the first determination unit 102 makes a determination. That is, the second determination unit 103 and the first determination unit 102 make determinations in the order of the second determination unit 103 and the first determination unit 102.

[0122] When the timing of virus infection is appropriate, the correct answer rate of the first learning model based on the learning of the first neural network can be increased. When there are many intercellular spaces, not only is the number of cells small, but also due to cell death caused by weak intercellular response, there is a problem of reduced protein recovery. When there are no intercellular spaces, there is a problem of reduced protein recovery due to cell death caused by nutrient deficiency due to insufficient absorption of nutrients from the culture medium. According to the present embodiment, virus infection can be performed at an appropriate time when the cells can tolerate virus infection, and thus an effect of being able to solve these problems can be obtained.

[0123] Since the operator knows the timing of virus infection, for example, the operator can operate the external device 3 to input or select the determination of the infection period, and the acquisition unit 101 acquires the input or selected result. In addition, regarding virus infection, it can also be automatically performed by other external devices 3 based on the result output by the protein recovery determination device 1A. In this case, the protein recovery determination device 1A can perform determination by itself to obtain whether it is before or after infection. For example, it can be that the acquisition unit 101 acquires information indicating whether it is in a virus-uninfected state or a virus-infected state, and based on the result acquired by the selection unit 110, selects whether to determine the protein recovery period by the first determination unit 102 or to determine the virus infection period by the second determination unit 103. The selection circuit can make a selection based on a learning model that discriminates information indicating whether it is in a virus-uninfected state or a virus-infected state.

[0124] Thus, the protein recovery determination device 1A can obtain whether to determine the virus infection period or the protein recovery period, and based on the obtained result, select a determination unit, a model, and a learning unit.

[0125] The protein recovery determination system 5 and the protein recovery determination system 5A are used for the production of proteins using cells (cell engineering).

[0126] A program for implementing all or part of the functions of the protein recovery determination device 1 (or 1A) of the present invention can also be recorded in a computer-readable recording medium, and the computer system reads and executes the program recorded in this recording medium, thereby performing all or part of the processing performed by the protein recovery determination device 1 (or 1A). In addition, the "computer system" mentioned here includes hardware such as an operating system (Operating System, OS) and peripheral devices. In addition, the "computer system" also includes a World Wide Web (WWW) system having a homepage providing environment (or display environment). In addition, the so-called "computer-readable recording medium" refers to a removable medium such as a floppy disk, a magneto-optical disk, a ROM, a compact disc read only memory (CD-ROM), and a storage device such as a hard disk built in the computer system. Furthermore, the so-called "computer-readable recording medium" also includes a medium that holds a program for a certain period of time, such as a volatile memory (RAM) inside a computer system of a server or a client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line.

[0127] In addition, the above program can also be transmitted from a computer system that stores this program in a storage device or the like to other computer systems via a transmission medium or through a transmission wave in the transmission medium. Here, the "transmission medium" for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) like the Internet or a communication line (communication wire) like a telephone line. In addition, the program can also be used to implement a part of the above functions. Furthermore, the program can also be a so-called differential file (differential program) that can implement the above functions in combination with a program already recorded in the computer system.

[0128] Figure 18 It is a graph showing the average correct answer rate and the average antibody production based on the first model and the second model. Figure 18 It shows the results of determining (first determination) whether the cell population infected with the virus is in a state suitable for protein recovery and determining (second determination) whether it is in a state suitable for infecting an uninfected cell population with the virus for each of the three images (n = 3) captured by the imaging device 2. In Figure 18 For comparison, the results when the person in charge A (a highly skilled person in charge with more than 10 years of experience) and the person in charge B (a person in charge with low skills and less than 1 year of experience) made the first determination and the second determination are also shown together. In addition, σ represents the standard deviation of the antibody production.

[0129] As Figure 18 shown, in the results for the combination of the cell line "Sf9" and the virus species "Tif1-r", when the person in charge A made both the first determination and the second determination, the average correct answer rate was "100%", the average antibody production was "15.4", and the standard deviation σ was "2.16". On the other hand, when the first determination and the second determination were made using the first model and the second model, the average correct answer rate was "93%", the average antibody production was "14.1", and the standard deviation σ was "2.54". From this, it can be confirmed that even when using the first model and the second model, high-precision determination can be performed, and furthermore, high antibody production can be achieved. Similarly, it can be confirmed that for the combination of the cell line "Sf9" and the virus species "MD5" and the combination of the cell line "HF" and the virus species "Mi-2", high-precision determination can be performed, and furthermore, high antibody production can be achieved.

[0130] As described above, the embodiments have been used to illustrate the ways of implementing the present invention, but the present invention is not limited by such embodiments in any way, and various modifications and substitutions can be made without departing from the gist of the present invention.

[0131] Explanation of reference numerals

[0132] 5, 5A: Protein recovery determination system

[0133] 1, 1A: Protein recovery determination device

[0134] 2: Imaging device

[0135] 3: External device

[0136] 101: Acquisition unit

[0137] 102: First determination unit

[0138] 103: Second determination unit

[0139] 104, 104-1, 104-2, 104-3, 104-4: First model

[0140] 105, 105-1, 105-2, 105-3, 105-4: Second model

[0141] 106: First learning unit

[0142] 107: Second learning unit

[0143] 108: Storage unit

[0144] 109: Output unit

[0145] 110: Selection unit

[0146] 21: Imaging mechanism

[0147] 22: Culture container

[0148] a1, a3: Image data

[0149] a2: Data representing the determination result

[0150] a11, a21: Image data

[0151] a12, a22: Learned model

[0152] a13, a23: Data representing the determination result

[0153] a14, a24: Image data

[0154] a15, a25: Learned model

Claims

1. A protein recovery determination system, comprising: A first determination unit that inputs first image data obtained by photographing the culture of a cell population infected with a virus in time series as first input data, and determines whether the cell population infected with the virus is in a state suitable for protein recovery as first output data; And An acquisition unit that obtains image data by photographing the inside of a culture vessel in time series, The first determination unit makes a determination using a first model, The first model is a learned model obtained by deep learning a first neural network using the actual values of the first input data and the actual values of the first output data as first training data.

2. The protein recovery determination system according to claim 1, further comprising a second determination unit. In the second determination unit, second image data obtained by photographing the subculture of a cell population not infected with a virus in time series is input as second input data, and it is determined whether it is in a state suitable for infecting the cell population not infected with the virus with the virus as second output data, The second determination unit makes a determination using a second model, The second model is a learned model obtained by deep learning a second neural network using the actual values of the second input data and the actual values of the second output data as second training data, The first determination unit makes a determination after the second determination unit has made a determination.

3. The protein recovery determination system according to claim 1 or 2, wherein The cells of the cell population are adherent cells.

4. The protein recovery determination system according to claim 3, wherein, The adherent cells are insect cells.

5. The protein recovery determination system according to claim 4, wherein, The insect cells are one of Sf (Spodoptera frugiperda (fall armyworm)) 9 cells, Sf21 cells, Tni (Trichoplusia ni (cabbage looper)) cells, and H5 (High Five) cells.

6. The protein recovery determination system according to claim 1 or 2, further comprising a first learning unit for learning the first neural network.

7. The protein recovery determination system according to claim 2, further comprising a second learning unit for learning the second neural network.

8. The protein recovery determination system according to claim 2, wherein, The acquisition unit acquires information indicating whether the cell population is in a state of not being infected with a virus or being infected with a virus, Based on the information acquired by the acquisition unit, one determination unit to be used is selected from the first determination unit and the second determination unit.

9. A recording medium that records a learned model, the learned model being obtained by deep learning a first neural network using first input data and first output data as first training data, The first input data is first image data obtained by photographing the culture of a cell population infected with a virus in time series, The first output data is data for determining whether the cell population infected with the virus is in a state suitable for protein recovery.

10. A recording medium that records a learned model, the learned model being obtained by deep learning a second neural network using second input data and second output data as second training data, The second input data is second image data obtained by photographing the subculture of a cell population not infected with a virus in time series, The second output data is data for determining whether it is a state suitable for infecting the virus-free cell population with the virus.

11. A method for determining protein recovery, wherein first image data is obtained by photographing the culture of the virus-infected cell population in time series, the first image data is input as first input data into a first determination unit, the first determination unit determines whether the virus-infected cell population is in a state suitable for protein recovery as first output data, the first determination unit makes the determination using a first model, the first model is a learned model obtained by deep learning a first neural network using the actual values of the first input data and the actual values of the first output data as first training data.

12. The method for determining protein recovery according to claim 11, wherein before making the determination as to whether the cell population is in a state suitable for protein recovery, second image data obtained by photographing the scale-up culture of the virus-free cell population in time series is obtained, the second image data is input as second input data into a second determination unit, the second determination unit determines whether it is a state suitable for infecting the virus-free cell population with the virus as second output data, the second determination unit makes the determination using a second model, the second model is a learned model obtained by deep learning a second neural network using the actual values of the second input data and the actual values of the second output data as second training data.

Images