Cell line stability prediction method and device, computer equipment and storage medium

Abstract

The application relates to a cell strain stability prediction method and device, computer equipment and a storage medium. The method comprises the following steps: obtaining an observation image of a cell strain to be predicted; inputting the observation image into a stability prediction model for processing to obtain a stability value of the cell strain to be predicted or a stability category determined according to the stability value; the stability prediction model is determined according to images of a plurality of historical cell strains and a stability value or a stability category determined according to the stability value of each historical cell strain; and the stability value or the stability category determined according to the stability value is used to represent the stability of the protein expression amount of the cell strain. When the prediction result of the stability value of the cell strain or the stability category determined according to the stability value is determined by using the prediction method of the application, the manual multi-generation culture and a large amount of protein measurement experiment are not needed, the stability prediction model can be relied on to realize the prediction of the stability of the cell strain, and the stability prediction efficiency of the cell strain is greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biotechnology, and in particular to a cell strain stability prediction method and device, computer equipment and storage medium. BACKGROUND

[0002] At present, in the field of biotechnology, a cell strain is a cell population formed by proliferation of a single cell through single cell separation culture or screening method. The protein expressed by the cell strain can be used for drug production.

[0003] In the traditional technology, the stability of the cell strain can be determined. For example, the cell strain is cultured for multiple generations (usually the cell strain can be subcultured for more than 20 generations), and then the protein expression yield of each generation of the cell strain is measured by manual experiment to obtain the protein yield of the cell strain, and the stability of the cell strain is evaluated based on the protein yield change of each generation of the cell strain.

[0004] However, when using the above method to determine the stability of the cell strain, the experiment amount is huge and time-consuming. Therefore, the current stability prediction method still has the problem of low efficiency. SUMMARY

[0005] Therefore, it is necessary to provide a cell strain stability prediction method, device, computer equipment, computer readable storage medium and computer program product which can improve the efficiency.

[0006] In a first aspect, the present application provides a cell strain stability prediction method, which comprises:

[0007] obtaining an observation image of a cell strain to be predicted;

[0008] inputting the observation image into a stability prediction model for processing to obtain a stability value of the cell strain to be predicted or a stability category determined according to the stability value; the stability prediction model is determined according to images of a plurality of historical cell strains and a stability value or a stability category determined according to the stability value of each historical cell strain; and the stability value or the stability category determined according to the stability value is used to represent the stability of the protein expression amount of the cell strain.

[0009] In one embodiment, the training process of the stability prediction model comprises:

[0010] using the images of the plurality of historical cell strains as model input, using the same stability value or stability category determined according to the stability value of the historical cell strain as the training target of the model output, training the initial model to obtain the stability prediction model.

[0011] In one of the embodiments, the images of the plurality of historical cell strains are input into a model, the training target of the model is the same as the stability value of the historical cell strain or the stability category determined according to the stability value, the initial model is trained to obtain the stability prediction model, including:

[0012] The image of the historical cell strain is subjected to cell segmentation processing to obtain a plurality of single-cell images corresponding to the image of the historical cell strain;

[0013] The plurality of single-cell images corresponding to the image of the historical cell strain are subjected to feature extraction as a whole to obtain a feature set of the image of the historical cell strain;

[0014] The initial model is determined according to the feature set, the training target of the model is the same as the stability value of the historical cell strain or the stability category determined according to the stability value, the initial model is trained to obtain the stability prediction model.

[0015] In one of the embodiments, the historical cell strain is processed, further including:

[0016] The protein expression amount at the early stage of culture of the historical cell strain and the protein expression amount at the late stage of culture of the historical cell strain are obtained;

[0017] The stability value of the historical cell strain or the stability category determined according to the stability value is determined according to the protein expression amount at the early stage of culture and the protein expression amount at the late stage of culture.

[0018] In one of the embodiments, the stability value of the historical cell strain or the stability category determined according to the stability value is determined according to the protein expression amount at the early stage of culture and the protein expression amount at the late stage of culture, including:

[0019] The protein expression amount of the Nth generation at the early stage or the average of the protein expression amounts of several generations before and after the Nth generation and the protein expression amount of the Mth generation at the late stage or the average of the protein expression amounts of several generations before and after the Mth generation are determined, and the stability value of the historical cell strain or the stability category determined according to the stability value is determined according to the ratio of the protein expression amount of the Nth generation at the early stage or the average of the protein expression amounts of several generations before and after the Nth generation to the protein expression amount of the Mth generation at the late stage or the average of the protein expression amounts of several generations before and after the Mth generation.

[0020] In one of the embodiments, the processing of the observation image by the stability prediction model includes:

[0021] The observation image is subjected to cell segmentation processing to obtain a plurality of single-cell images corresponding to the observation image;

[0022] The plurality of single-cell images are subjected to sampling and combination processing to obtain a reference single-cell image combination;

[0023] Feature extraction is performed on the reference single-cell image combination, and a stability value of the observation image or a stability category determined based on the stability value is determined based on a result of the feature extraction.

[0024] In a second aspect, the present application further provides a cell strain stability prediction device, which comprises:

[0025] An acquisition module is configured to acquire an observation image of a cell strain to be tested.

[0026] A processing module is configured to input the observation image into a stability prediction model for processing, so as to obtain a stability value of the cell strain to be tested or a stability category determined based on the stability value. The stability prediction model is determined based on images of a plurality of historical cell strains and a stability value or a stability category determined based on the stability value of each historical cell strain. The stability value or the stability category determined based on the stability value is used to represent stability of protein expression of the cell strain.

[0027] In a third aspect, the present application further provides a computer device, which comprises a memory and a processor. The memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0028] An observation image of a cell strain to be tested is acquired.

[0029] The observation image is input into a stability prediction model for processing, so as to obtain a stability value of the cell strain to be tested or a stability category determined based on the stability value. The stability prediction model is determined based on images of a plurality of historical cell strains and a stability value or a stability category determined based on the stability value of each historical cell strain. The stability value or the stability category determined based on the stability value is used to represent stability of protein expression of the cell strain.

[0030] In a fourth aspect, the present application further provides a computer readable storage medium, which stores a computer program. The computer program is executed by a processor to implement the following steps:

[0031] An observation image of a cell strain to be tested is acquired.

[0032] The observation image is input into a stability prediction model for processing, so as to obtain a stability value of the cell strain to be tested or a stability category determined based on the stability value. The stability prediction model is determined based on images of a plurality of historical cell strains and a stability value or a stability category determined based on the stability value of each historical cell strain. The stability value or the stability category determined based on the stability value is used to represent stability of protein expression of the cell strain.

[0033] In a fifth aspect, the present application further provides a computer program product, which comprises a computer program. The computer program is executed by a processor to implement the following steps:

[0034] an observation image of a cell strain to be predicted is acquired;

[0035] the observation image is input into a stability prediction model for processing to obtain a stability value of the cell strain to be predicted or a stability category determined based on the stability value; the stability prediction model is determined based on images of a plurality of cell strains and a stability value or a stability category determined based on the stability value of each historical cell strain; and the stability value or the stability category determined based on the stability value is used to represent the stability of protein expression of the cell strain.

[0036] The cell strain stability prediction method, device, computer device, storage medium, and computer program product provided in the embodiments of the present application acquire an observation image of a cell strain to be predicted; input the observation image into a stability prediction model for processing to obtain a stability value of the cell strain to be predicted or a stability category determined based on the stability value; the stability prediction model is determined based on images of a plurality of cell strains and a stability value or a stability category determined based on the stability value of each historical cell strain; and the stability value or the stability category determined based on the stability value is used to represent the stability of protein expression of the cell strain. As can be seen, by acquiring cell strain images and stability values or stability categories determined based on the stability values of a plurality of historical generations of cell strains, and using the images and the stability values or the stability categories determined based on the stability values as training samples of a model for model training, a trained stability prediction model can be obtained. Then, the cell strain image can be input into the model for prediction, and a stability value or a stability category prediction result determined based on the stability value of the cell strain can be obtained. This replaces the scheme in the prior art, which not only requires the cell strain to be cultured for more than 20 generations, but also requires a large number of manual protein measurement experiments to finally obtain the stability value of the cell strain or the stability category determined based on the stability value. Therefore, when the prediction method of the present application is used to determine the stability value of the cell strain or the stability category prediction result determined based on the stability value, manual multi-generation culture and a large number of protein measurement experiments are not required, and the stability of the cell strain can be predicted by relying on the stability prediction model, which greatly improves the efficiency of cell strain stability prediction. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 An application environment diagram of the cell strain stability prediction method provided in the embodiments of the present application;

[0038] Figure 2 A flowchart of the cell strain stability prediction method provided in the embodiments of the present application;

[0039] Figure 3 Another flowchart of the cell strain stability prediction method provided in the embodiments of the present application;

[0040] Figure 4 Another flowchart of the cell strain stability prediction method provided by the embodiment of the present application is shown in FIG. 6.

[0041] Figure 5 Another flowchart of the cell strain stability prediction method provided by the embodiment of the present application is shown in FIG. 6.

[0042] Figure 6 Another flowchart of the cell strain stability prediction method provided by the embodiment of the present application is shown in FIG. 6.

[0043] Figure 7 Another flowchart of the cell strain stability prediction method provided by the embodiment of the present application is shown in FIG. 6.

[0044] Figure 8 A structure block diagram of the cell strain stability prediction device provided by the embodiment of the present application is shown in FIG. 7.

[0045] Figure 9 An internal structure diagram of the computer device provided by the embodiment of the present application is shown in FIG. 8. DETAILED DESCRIPTION

[0046] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0047] The cell strain stability prediction method provided by the embodiment of the present application can be applied in the application environment as shown in FIG. 1. Figure 1 The cell strain stability prediction method provided by the embodiment of the present application can be applied in the application environment as shown in FIG. 1. Figure 1 An application environment diagram of the cell strain stability prediction method provided by the embodiment of the present application is shown in FIG. 1. The application environment includes a microscope 10 and a server 20. The microscope 10 can communicate with the server 20. Specifically, the communication can be performed in a wired or wireless manner. The microscope 10 can acquire a cell strain image with unknown stability and transmit the cell strain image to the server 20. The server 20 can process the cell strain image to obtain a cell strain stability value or a stability category prediction result determined according to the stability value. Specifically, the microscope 10 can be an electron microscope or an optical microscope. The server 20 can be a stand-alone server or a server cluster composed of multiple servers.

[0048] Figure 2 A flowchart of the cell strain stability prediction method provided by the embodiment of the present application is shown in FIG. 2. The execution subject of the method can be a server 20 as shown in FIG. 1. Figure 1 The method includes the following steps as shown in FIG. 2. Figure 2

[0049] ​Step 201: obtaining an observation image of a cell strain to be predicted;

[0050] The embodiment of the present application provides a cell strain stability prediction method, which can predict the stability of a cell strain based on an image of the cell strain. For example, the image of the cell strain is input into a model to predict the stability of the cell strain. Therefore, in the embodiment of the present application, the image of the cell strain to be predicted can be obtained first.

[0051] In a possible implementation, the image of the cell strain to be predicted can be an observation image of the cell strain to be predicted.

[0052] Specifically, the observation image of the cell strain to be predicted can be an image of the cell strain collected by an optical observation instrument when the cell strain is observed by the optical observation instrument. For example, the observation image of the cell strain to be predicted can be an image obtained by image collection of the cell strain at any generation in a culture cycle of the cell strain.

[0053] For example, the optical observation instrument is a microscope 10 in the application environment shown in FIG. 1, the microscope 10 can collect the observation image of the cell strain to be predicted, and the server 20 can obtain the observation image from the microscope 10. Figure 1 For example, the optical observation instrument is a microscope 10 in the application environment shown in FIG. 1, the microscope 10 can collect the observation image of the cell strain to be predicted, and the server 20 can obtain the observation image from the microscope 10.

[0054] Step 202: inputting the observation image into a stability prediction model for processing to obtain a stability value of the cell strain to be predicted or a stability category determined according to the stability value; the stability prediction model is determined according to images of a plurality of historical cell strains and a stability value or a stability category determined according to the stability value of each historical cell strain; and the stability value or the stability category determined according to the stability value is used to represent the stability of the protein expression amount of the cell strain.

[0055] Specifically, the historical cell strain can be a cell strain cultured in the past, and the historical cell strain image can be an image obtained by image collection of the historical cell strain in a culture process of the historical cell strain. For example, the historical cell strain can be cultured for multiple generations, and the image of the historical cell strain can be an image of a cell strain at any generation in a culture cycle of the cell strain.

[0056] The stability of the protein expression amount of the cell strain is used to represent the difference between the protein expression amount of a descendant cell strain and the protein expression amount of a predecessor cell strain in a culture cycle, and can be a ratio of the protein expression amount of the descendant cell strain to the protein expression amount of the predecessor cell strain.

[0057] For example, the culture cycle includes X generations, wherein the stability value or the stability category determined according to the stability value of the cell strain is determined according to the difference between the cumulative amount A1 of the pre-R generation protein expression in the X generations and the cumulative amount A2 of the post-S generation protein expression in the X generations, for example, the stability value or the stability category determined according to the stability value can be determined according to the ratio of A2 to A1, less than 70% is unstable, and 70% or more is stable.

[0058] It should be noted that when collecting the images of the historical cell strains, the stability value or the stability category determined according to the stability value (hereinafter referred to as the true stability value or the stability category determined according to the stability value) of the historical cell strains can be determined according to the above description. Further, a set of training samples can be constructed according to the images of the historical cell strains and the corresponding stability values or the stability categories determined according to the stability values, and based on this, as the number of historical cell strains increases, multiple sets of training samples can be obtained. Training the training samples can obtain a stability prediction model.

[0059] In one possible implementation, the above-mentioned multi-generation cell strain images and the true stability values or the stability categories determined according to the stability values corresponding to the images can be used as training samples of a model, and the model is trained until the ratio of the predicted stability value or the stability category determined according to the stability value obtained by the model training to the true stability value or the stability category determined according to the stability value of a certain cell strain image approaches infinity, at which time the training of the model is completed, and a stability prediction model is obtained.

[0060] In one possible implementation, the cell strain image to be predicted collected by the microscope 10 can be input into the stability prediction model, and based on the cell strain image features learned by the stability prediction model, the predicted stability value or the stability category determined according to the stability value of the cell strain image to be predicted is obtained.

[0061] It should be noted that the predicted stability value of the cell strain image to be predicted is used to represent regression, that is, to determine a specific stability value, and the stability category determined according to the stability value is used to represent classification, that is, to determine which interval of the stability value the cell strain to be predicted belongs to.

[0062] The observation image of the to-be-predicted cell strain is obtained, the to-be-observed image is input into the stability prediction model for processing, and a stability value of the to-be-predicted cell strain or a stability category determined according to the stability value is obtained. The stability prediction model is determined according to the images of a plurality of historical cell strains and the stability value or the stability category determined according to the stability value of each historical cell strain. The stability value or the stability category determined according to the stability value is used to represent the stability of the protein expression amount of the cell strain. It can be seen that, by using the cell strain stability prediction method of the present application, the images of the cell strains obtained by culturing a plurality of generations of history and the corresponding true stability values or the stability categories determined according to the stability values are obtained. The above images and stability values or the stability categories determined according to the stability values are used as training samples of the model, and the model is trained to obtain a trained stability prediction model. Then, the image of the to-be-predicted cell strain can be input into the model to predict the stability, and the predicted stability value of the cell strain or the stability category determined according to the stability value can be obtained. Instead of the scheme in the prior art, which needs to culture the cell strain for more than 20 generations and manually measure the protein yield of the cell strain through a large number of experiments to finally obtain the stability value of the cell strain or the stability category determined according to the stability value. Therefore, when the prediction method of the present application is used to determine the predicted stability value of the cell strain or the stability category determined according to the stability value, manual multi-generation culture and a large number of protein measurement experiments are not required, and the stability prediction model can be used to predict the stability of the cell strain, which greatly improves the efficiency of cell strain stability prediction.

[0063] In the foregoing embodiments, the technical scheme of determining the stability value of the to-be-predicted cell strain or the stability category determined according to the stability value according to the observation image of the to-be-observed cell strain and the stability prediction model is introduced. In another embodiment of the present application, the stability prediction model can be obtained according to the images of the historical cell strains and the stability value or the stability category determined according to the stability value. For example, the training process of the stability prediction model involved in step 202 in the foregoing is as follows:

[0064] The images of a plurality of historical cell strains are used as model inputs, the training target is that the model output is the same as the stability value or the stability category determined according to the stability value of the historical cell strain, the initial model is trained, and the stability prediction model is obtained.

[0065] Specifically, the training target is used to represent that, in the model training process, the input of the model is the historical cell strain image and the corresponding real stability value or the stability category determined according to the stability value, the model can automatically learn the characteristics of the historical cell strain image, derive the predicted stability value or the stability category determined according to the stability value of the fine cell strain, there is a certain difference between the predicted stability value or the stability category determined according to the stability value and the real stability value or the stability category determined according to the stability value, the predicted stability value or the stability category determined according to the stability value is compared with the real stability value or the stability category determined according to the stability value, and the model parameters are updated according to the comparison result until the predicted stability value or the stability category determined according to the stability value is infinitely close to or equal to the real stability value or the stability category determined according to the stability value.

[0066] For example, the image of the historical cell strain can be a cell strain cell morphology image cultured for at least 20 generations.

[0067] In one possible implementation, a plurality of historical cell strain images cultured for at least 20 generations and the real stability value or the stability category determined according to the stability value corresponding to the image can be input into an initial model, the predicted stability value or the stability category determined according to the stability value output by the model is compared with the real stability value or the stability category determined according to the stability value corresponding to the image, the model parameters are updated according to the comparison result until the predicted stability value or the stability category determined according to the stability value is infinitely close to or equal to the real stability value or the stability category determined according to the stability value, and the stability prediction model is obtained.

[0068] In the foregoing embodiment, the technical solution of obtaining the stability prediction model according to the image and the stability value or the stability category determined according to the stability value of the historical cell strain is introduced. In another embodiment of the present application, the stability prediction model can be determined according to the single cell image of the historical cell strain. For example, the specific implementation of the foregoing step of "taking the images of a plurality of historical cell strains as the model input, taking the same as the training target of the stability value or the stability category determined according to the stability value output by the model as the model output, training the initial model, and obtaining the stability prediction model" includes the steps of Figure 3

[0069] Step 301, performing cell segmentation processing on the image of the historical cell strain to obtain a plurality of single cell images corresponding to the historical cell strain image;

[0070] Specifically, the cell segmentation processing is used to segment the multiple cells into single cells. For example, a cell strain image contains a plurality of cells, and the plurality of cells are segmented to obtain corresponding single cell images.

[0071] ​In a possible implementation, the historical cell strain image can be segmented into single cell images to obtain a plurality of single cell images corresponding to the historical cell strain image.

[0072] In step 302, feature extraction is performed on the plurality of single cell images corresponding to the historical cell strain image to obtain a feature set of the historical cell strain image.

[0073] Specifically, the feature extraction is used to represent a process of extracting feature information from the historical cell strain image. The feature set is a set composed of the extracted feature information.

[0074] In a possible implementation, the feature information can be extracted from the single cell image of the historical cell strain, and the feature information contains the feature information of the historical cell strain image.

[0075] In step 303, an initial model is determined according to the feature set, and the training target of the initial model is to output the same stability category as the stability value or the stability category determined according to the stability value of the historical cell strain. The initial model is trained to obtain a stability prediction model.

[0076] Specifically, the initial model is used to represent an untrained model. The stability value of the historical cell strain or the stability category determined according to the stability value is used to represent the stability degree of the protein yield of the historical cell strain. The stability prediction model is used to represent that the historical cell strain image and the corresponding stability value or the stability category determined according to the stability value are input into the initial model for training. The initial model can learn the feature information of the cell strain image. The cell strain stability prediction result output by the model is compared with the real stability value or the stability category determined according to the stability value of the corresponding cell strain. The model parameters are updated according to the comparison result until the prediction result approaches or equals the real stability value or the stability category determined according to the stability value, and the stability prediction model is obtained.

[0077] In a possible implementation, the feature information of the single cell image processed by the cell segmentation can be extracted. The cell strain stability prediction result output by the model is compared with the real stability value or the stability category determined according to the stability value of the corresponding cell strain. The model parameters are updated according to the comparison result until the prediction result approaches or equals the real stability value or the stability category determined according to the stability value, and the stability prediction model is obtained.

[0078] In the foregoing embodiment, the technical solution of determining the stability prediction model according to the single-cell images of the historical cell strains is introduced. In another embodiment of the present application, the stability value of the historical cell strains can be determined according to the protein expression amount of the historical cell strains at the early stage of culture and the protein expression amount at the late stage of culture, or the stability category determined according to the stability value. For example, the specific implementation of the "stability value of the historical cell strains or the stability category determined according to the stability value" involved in the foregoing step 402 includes Figure 4 the steps of:

[0079] Step 401, obtaining the protein expression amount of the historical cell strains at the early stage of culture and the protein expression amount of the historical cell strains at the late stage of culture.

[0080] Specifically, the protein expression amount at the early stage of culture is used to represent the protein expression amount of each historical cell strain at the early stage of the culture cycle. The protein expression amount at the late stage of culture is used to represent the protein expression amount of each historical cell strain at the late stage of the culture cycle.

[0081] For example, the culture cycle can be at least 20 generations.

[0082] In one possible implementation, the protein expression amount of each historical cell strain at the early stage of the 20-generation culture cycle and the protein expression amount at the late stage of the culture cycle can be obtained.

[0083] Step 402, determining the stability value of the historical cell strains according to the protein expression amount at the early stage and the protein expression amount at the late stage of culture, or determining the stability category according to the stability value.

[0084] Specifically, the stability value of the historical cell strains or the stability category determined according to the stability value is used to represent the stability degree of the protein production of the cell strains.

[0085] In one possible implementation, the stability degree of the protein production of the historical cell strains can be determined according to the protein expression amount of the historical cell strains at the early stage of the 20-generation culture cycle and the protein expression amount at the late stage of the culture cycle.

[0086] In the foregoing embodiment, the technical solution of determining the stability value of the historical cell strains according to the protein expression amount at the early stage of culture and the protein expression amount at the late stage of culture, or determining the stability category according to the stability value is introduced. In another embodiment of the present application, the stability value of the cell strains can be determined according to the protein expression amount of the historical cell strains at the first N generations at the early stage of culture and the protein expression amount at the last M generations at the late stage of culture, or the stability category determined according to the stability value. For example, the "stability value of the historical cell strains or the stability category determined according to the stability value" involved in the foregoing step 402 includes:

[0087] The average of the protein expression amount of the Nth generation in the early stage or the protein expression amount of several generations before and after the Nth generation and the average of the protein expression amount of the Mth generation in the late stage or the protein expression amount of several generations before and after the Mth generation are determined, and the stability value of the historical cell strain is determined according to the ratio of the average of the protein expression amount of the Nth generation in the early stage or the protein expression amount of several generations before and after the Nth generation to the average of the protein expression amount of the Mth generation in the late stage or the protein expression amount of several generations before and after the Mth generation, or the stability category determined according to the stability value.

[0088] Specifically, the average of the protein expression amount of the Nth generation in the early stage or the protein expression amount of several generations before and after the Nth generation is used to represent the protein expression amount of the historical cell strain in the early stage of culture. For example, the culture period can be 20 generations, the early stage of culture can be the first 3 generations, and the average of the protein expression amount of the first 3 generations can be the average of the protein expression amount of the first 3 generations of each cell strain. The average of the protein expression amount of the Mth generation is used to represent the protein expression amount of the historical cell strain in the late stage of culture. For example, the culture period can be 20 generations, the late stage of culture can be the last 3 generations, and the average of the protein expression amount of the last 3 generations can be the average of the protein expression amount of the last 3 generations of each cell strain. The ratio is used to represent the stability degree of the historical cell strain, and the stability value is determined according to the ratio or the stability category determined according to the stability value.

[0089] In one possible implementation, the culture period is 20 generations, the average of the protein expression amount of the first 3 generations and the average of the protein expression amount of the last 3 generations can be obtained, and the ratio of the average of the protein expression amount of the first 3 generations to the average of the protein expression amount of the last 3 generations is determined as the stability degree of the historical cell strain, and the stability value is determined according to the ratio or the stability category determined according to the stability value.

[0090] In the foregoing embodiments, the technical solutions are introduced that the stability value of the historical cell strain can be determined according to the ratio of the average of the protein expression amount of the Nth generation in the early stage or the protein expression amount of several generations before and after the Nth generation to the average of the protein expression amount of the Mth generation in the late stage or the protein expression amount of several generations before and after the Mth generation, or the stability category determined according to the stability value. In another embodiment of the present application, the stability value of the cell strain to be observed or the stability category determined according to the stability value can be determined according to the observed image and the stability prediction model. For example, the specific implementation of the foregoing step of “processing of the observed image by the stability prediction model” includes the following steps: Figure 5

[0091] Step 501, performing cell segmentation processing on the observed image to obtain a plurality of single-cell images corresponding to the observed image;

[0092] Specifically, the observed image can be an image of the cell strain to be predicted collected by the microscope 10. For example, the observed image of the cell strain to be predicted can be any generation in the culture period of the cell strain.

[0093] ​In a possible implementation, an observation image of a cell strain to be predicted can be acquired by the microscope 10, and a cell segmentation process can be performed on the observation image to obtain a plurality of single-cell images corresponding to the cell strain image.

[0094] In step 502, the plurality of single-cell images are sampled and combined to obtain a reference single-cell image combination.

[0095] Specifically, the sampling process is used to extract a certain number of single-cell images from the plurality of single-cell images. The reference single-cell image combination is used to represent the certain number of single-cell images as input of the stability prediction model.

[0096] In a possible implementation, a certain number of single-cell images can be extracted from the plurality of single-cell images, and the certain number of single-cell images are used as input of the stability prediction model.

[0097] In step 503, feature extraction is performed on the reference single-cell image combination, and a stability value of the observation image or a stability category determined based on the stability value is determined based on a result of the feature extraction.

[0098] Specifically, the feature extraction is used to extract feature information from the single-cell images of the cell strain to be predicted.

[0099] In a possible implementation, feature information can be extracted from the single-cell images of the cell strain to be predicted, and the feature information includes a feature set of the single-cell images.

[0100] In the foregoing embodiments, the technical solutions that the predicted stability value of the observation image of the cell strain to be observed or the stability category determined based on the stability value can be determined based on the observation image and the stability prediction model are introduced. Therefore, by using the cell strain stability prediction method provided in the present application, the prediction of the stability of the cell strain can be achieved by relying on the stability prediction model without manual multi-generation culture and a large number of protein testing experiments, and the efficiency of the prediction of the stability of the cell strain is greatly improved.

[0101] Figure 6 Another flowchart of the cell strain stability prediction method provided in the embodiments of the present application can be used to train a binary classification, multi-classification, or regression training model (i.e., the initial model described in the foregoing embodiments). The specific implementation of the method includes the following steps:

[0102] In step T1, the process starts.

[0103] In step T2, a large number of microscope images of the first three generations of cell strains (i.e., the images of the historical cell strains described in the foregoing embodiments) and the protein expression amount of the supernatant of each generation (i.e., the protein expression amount of the historical cell strains described in the foregoing embodiments) are accumulated. Then, step T3 is performed.

[0104] Step T3, take the ratio of the late protein expression amount (i.e., the culture late protein expression amount described in the foregoing embodiment) to the early protein expression amount (i.e., the culture early protein expression amount described in the foregoing embodiment) as the stable value of the cell strain, and then determine the stable value or the stability category determined according to the stable value (i.e., the stable value of the historical cell strain or the stability category determined according to the stable value described in the foregoing embodiment) from the stable value. The stable value of 70% or more is marked as a stable strain, and the stable value of less than 70% is marked as an unstable strain. Then, step T4 is performed.

[0105] Step T4, cell cutting (i.e., the cell splitting process described in the foregoing embodiment). Then, step T5 is performed.

[0106] Step T5, population characterization.

[0107] It should be noted that the population characterization can stack the single cell pictures of each generation of the cell strain to generate a multi-channel image to represent the respective population input. This is beneficial to simplify the initial model input, so as to reduce the processing load of the initial model.

[0108] Step T6, data augmentation (i.e., the sampling and combination process described in the foregoing embodiment).

[0109] It should be noted that the data augmentation can extract and combine multiple single cell images obtained by cutting the image of each generation of the cell strain multiple times to generate multiple samples, which can ensure the randomness of the training samples and make the training result more accurate. For example, if 20,000 single cell images are obtained after cell cutting of the image of each generation of the cell strain, the server 20 can first randomly extract 3,000 images from the above single cell images, extract 100 times with replacement, and combine the 3,000 images into one sample, thereby obtaining a total of 100 samples. Then, the 3,000 images of each sample are shuffled and recombined to generate new samples, thereby obtaining a total of 1,000 new samples. Step T7, generate a 3D data set.

[0110] It should be noted that the three groups of 1,000 samples obtained by processing the images of the first three generations of the cell strain obtained from step T2 by step T6 are extracted one sample from each group, and three samples are obtained. They are combined into new samples, and 1,000 new samples are generated by extracting and combining 1000 times with replacement. Each sample in the 1,000 new samples is packed in order, generating 1,000 packs, each containing 3 groups of images. The three groups of 1,000 images are shuffled and rearranged 10 times, and the above packing process is repeated. Finally, 1,000 new samples of the cell strain are obtained as the input of the model. Then, step T8 is performed.

[0111] Step T8, binary classification, multi-classification or regression training model (i.e., the initial model described in the foregoing embodiment).

[0112] It should be noted that the binary classification, multi-classification or regression model can be trained according to the cell strain image and the corresponding stability.

[0113] Step T9, end.

[0114] It should be noted that the first three generations of microscope images are used to represent the cell population after preprocessing (i.e., cell cutting and data enhancement described in the foregoing embodiments), and the above cell population is used as a sample for model training.

[0115] Figure 7 Another flowchart of the cell strain stability prediction method provided in the embodiments of the present application is shown in FIG. 6. Figure 7 In the foregoing embodiments, the first three generations of microscope images of the cell strain (i.e., the observation images of the cell strain to be predicted described in the foregoing embodiments) are taken as examples for stability prediction, but the specific number of generations of the observation images of the cell strain is not limited in the embodiments of the present application.

[0116] Figure 7 The observation images of the cell strain to be predicted can be predicted for stability according to the binary classification, multi-classification or regression model (i.e., the stability prediction model described in the foregoing embodiments), and the specific implementation of the method includes the following steps:

[0117] Step P1, start.

[0118] Step P2, obtaining the microscope images of the first three generations of the cell strain with unknown stability (i.e., the observation images described in the foregoing embodiments). Then, step P3 is performed.

[0119] Step P3, cell cutting (i.e., cell segmentation processing described in the foregoing embodiments). Then, step P5 is performed.

[0120] Step P4, population representation.

[0121] It should be noted that the population representation can stack the single cell pictures of each generation of the cell strain to generate a multi-channel image to represent the respective population input. This is beneficial to simplify the input of the binary classification, multi-classification or regression model (i.e., the initial model described in the foregoing embodiments), so as to reduce the processing load of the binary classification, multi-classification or regression model.

[0122] Step P5, data enhancement (i.e., sampling and combination processing described in the foregoing embodiments).

[0123] It should be noted that data augmentation can combine multiple single-cell images cut from each generation of cell strain images to generate multiple samples, which can ensure the randomness of the training samples and make the training results more accurate. For example, if 20,000 single-cell images are obtained after cell cutting from each generation of cell strain images, the server 20 can first randomly extract 3,000 images from the single-cell images, extract 100 times with replacement, and combine the 3,000 images into a sample, thereby obtaining a total of 100 samples. The 3,000 images of each sample are randomly reorganized to generate new samples, thereby obtaining a total of 1,000 samples. Step P6, generating a 3D data set.

[0124] It should be noted that the first three generations of images of the cell strain obtained from step T2 are processed by step T6 to obtain three groups of 1,000 samples. One sample is extracted from each of the three groups of 1,000 samples, and the three samples are combined into a new sample. This is repeated 1,000 times with replacement to generate 1,000 new samples. Each of the 1,000 new samples is sequentially packaged to generate 1,000 packages, each containing 3 groups of images. The three groups of 1,000 images are randomly reorganized 10 times, and the above packaging process is repeated. Finally, 1,000 new samples of the cell strain are obtained as input for the model. Step P7 is then performed.

[0125] Step P7, a binary classification or multi-classification prediction model (i.e., the stability prediction model described in the foregoing embodiments).

[0126] It should be noted that the binary classification, multi-classification, or regression model can complete the prediction of the stability of the cell strain according to the cell strain images.

[0127] Step P8, outputting the cell strain stability prediction result (i.e., the stability value of the cell strain to be predicted or the stability category determined according to the stability value, as described in the foregoing embodiments).

[0128] Step P9, end.

[0129] It should be understood that although each step in the flowchart involved in each of the above-described embodiments is displayed in sequence according to the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in each of the above-described embodiments can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or steps or stages in other steps.

[0130] Based on the same inventive concept, the embodiment of the present application also provides a device for implementing the cell strain stability prediction method, Figure 8 The structural block diagram of the cell strain stability prediction device provided by the embodiment of the present application is shown in the figure, and the device comprises:

[0131] The acquisition module 801 is configured to acquire an observation image of a cell strain to be tested.

[0132] The processing module 802 is configured to input the observation image into a stability prediction model for processing, so as to obtain a stability value of the cell strain to be predicted or a stability category determined according to the stability value; the stability prediction model is determined according to images of a plurality of historical cell strains and a stability value of each historical cell strain or a stability category determined according to the stability value; and the stability value or the stability category determined according to the stability value is used to represent the stability of the protein expression amount of the cell strain.

[0133] In one embodiment, the processing module 802 is configured to take the images of the plurality of historical cell strains as model inputs, take the same stability value of the historical cell strain or the stability category determined according to the stability value as a training target of the model output, train an initial model, and obtain the stability prediction model.

[0134] In one embodiment, the images of the historical cell strains are subjected to cell segmentation processing, so as to obtain a plurality of single-cell images corresponding to the historical cell strain images; the plurality of single-cell images corresponding to the historical cell strain images are subjected to feature extraction as a whole, so as to obtain a feature set of the historical cell strain images; the initial model is determined according to the feature set, the same stability value of the historical cell strain or the stability category determined according to the stability value is taken as a training target of the model output, the initial model is trained, and the stability prediction model is obtained.

[0135] In one embodiment, the initial protein expression amount of the historical cell strain and the late protein expression amount of the historical cell strain are acquired; and the stability value of the historical cell strain or the stability category determined according to the stability value is determined according to the initial protein expression amount and the late protein expression amount.

[0136] In one embodiment, the protein expression amount of the Nth generation in the early stage or the average of the protein expression amounts of several generations before and after the Nth generation and the protein expression amount of the Mth generation in the late stage or the average of the protein expression amounts of several generations before and after the Mth generation are determined, and the stability value of the historical cell strain or the stability category determined according to the stability value is determined according to the ratio of the protein expression amount of the Nth generation in the early stage or the average of the protein expression amounts of several generations before and after the Nth generation to the protein expression amount of the Mth generation in the late stage or the average of the protein expression amounts of several generations before and after the Mth generation.

[0137] In an embodiment, the processing module 802 is configured to perform cell segmentation on the observation image to obtain a plurality of single-cell images corresponding to the observation image; perform sampling and combination on the plurality of single-cell images to obtain a reference single-cell image combination; perform feature extraction on the reference single-cell image combination, and determine a stability value of the observation image based on a result of the feature extraction or a stability category determined based on the stability value.

[0138] Each of the above modules in the cell strain stability prediction apparatus can be implemented wholly or partially by software, hardware, or a combination thereof. Each of the above modules can be embedded in or independent of a processor in a computer device in a hardware form, or can be stored in a memory in the computer device in a software form, so as to be invoked and executed by the processor to perform operations corresponding to each of the above modules.

[0139] In an embodiment, a computer device is provided, which can be a server, Figure 9 A computer device internal structure diagram is provided for an embodiment of the present application. The computer device includes a processor, a memory, and a network interface connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is configured to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement a cell strain stability prediction method.

[0140] Those skilled in the art can understand that, Figure 9 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. A specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0141] In an embodiment, a computer device is provided, which includes a memory and a processor. The memory stores a computer program. The processor executes the computer program to implement the following steps:

[0142] An observation image of a cell strain to be predicted is obtained. The observation image is input into a stability prediction model for processing to obtain a stability value of the cell strain to be predicted or a stability category determined based on the stability value. The stability prediction model is determined based on images of a plurality of historical cell strains and a stability value or a stability category determined based on the stability value of each historical cell strain. The stability value or the stability category determined based on the stability value is used to represent the stability of the protein expression amount of the cell strain.

[0143] In one embodiment, the processor, when executing the computer program, further implements the following steps: inputting images of a plurality of historical cell strains as model inputs, taking the same stability value of the historical cell strains or a stability category determined according to the stability value as a training target of the model output, training the initial model, and obtaining the stability prediction model.

[0144] In one embodiment, the processor, when executing the computer program, further implements the following steps: performing cell segmentation processing on the images of the historical cell strains to obtain a plurality of single-cell images corresponding to the images of the historical cell strains.

[0145] performing feature extraction on the plurality of single-cell images corresponding to the images of the historical cell strains to obtain a feature set of the images of the historical cell strains; determining an initial model according to the feature set, taking the same stability value of the historical cell strains or a stability category determined according to the stability value as a training target of the model output, training the initial model, and obtaining the stability prediction model.

[0146] In one embodiment, the processor, when executing the computer program, further implements the following steps: obtaining the protein expression amount of the historical cell strain at the early stage of culture and the protein expression amount of the historical cell strain at the late stage of culture; determining the stability value of the historical cell strain or a stability category determined according to the stability value according to the protein expression amount at the early stage of culture and the protein expression amount at the late stage of culture.

[0147] In one embodiment, the processor, when executing the computer program, further implements the following steps:

[0148] determining the protein expression amount of the Nth generation at the early stage or the average protein expression amount of several generations before and after the Nth generation, and the protein expression amount of the Mth generation at the late stage or the average protein expression amount of several generations before and after the Mth generation, and determining the stability value of the historical cell strain or a stability category determined according to the stability value according to the ratio of the protein expression amount of the Nth generation at the early stage or the average protein expression amount of several generations before and after the Nth generation to the protein expression amount of the Mth generation at the late stage or the average protein expression amount of several generations before and after the Mth generation.

[0149] In one embodiment, the processor, when executing the computer program, further implements the following steps: performing cell segmentation processing on the observation image to obtain a plurality of single-cell images corresponding to the observation image; performing sampling and combination processing on the plurality of single-cell images to obtain a reference single-cell image combination; performing feature extraction on the reference single-cell image combination, and determining the stability value of the observation image or a stability category determined according to the stability value based on the result of the feature extraction.

[0150] In one embodiment, a computer-readable storage medium is provided, which stores a computer program, and the computer program, when executed by a processor, implements the following steps:

[0151] obtaining an observation image of a cell strain to be predicted;

[0152] inputting the image to be observed into the stability prediction model to obtain a stability value of the cell strain to be predicted or a stability category determined according to the stability value; the stability prediction model is determined according to images of a plurality of historical cell strains and a stability value of each historical cell strain or a stability category determined according to the stability value; and the stability value or the stability category determined according to the stability value is used to represent the stability of the protein expression amount of the cell strain.

[0153] In one embodiment, the computer program, when executed by the processor, further implements the following steps: inputting the images of the plurality of historical cell strains into the model as model inputs, taking the same stability value of the historical cell strain or the stability category determined according to the stability value as the training target of the model output, training the initial model, and obtaining the stability prediction model.

[0154] In one embodiment, the computer program, when executed by the processor, further implements the following steps: performing cell segmentation processing on the image of the historical cell strain to obtain a plurality of single-cell images corresponding to the image of the historical cell strain; performing feature extraction on the plurality of single-cell images corresponding to the image of the historical cell strain as a whole to obtain a feature set of the image of the historical cell strain; determining the initial model according to the feature set, taking the same stability value of the historical cell strain or the stability category determined according to the stability value as the training target of the model output, training the initial model, and obtaining the stability prediction model.

[0155] In one embodiment, the computer program, when executed by the processor, further implements the following steps: obtaining the protein expression amount at the early stage of culture of the historical cell strain and the protein expression amount at the late stage of culture of the historical cell strain; and determining the stability value of the historical cell strain or the stability category determined according to the stability value according to the protein expression amount at the early stage of culture and the protein expression amount at the late stage of culture.

[0156] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0157] determining the protein expression amount of the Nth generation at the early stage or the average of the protein expression amounts of several generations before and after the Nth generation, and the protein expression amount of the Mth generation at the late stage or the average of the protein expression amounts of several generations before and after the Mth generation, and determining the stability value of the historical cell strain or the stability category determined according to the stability value according to the ratio of the protein expression amount of the Nth generation at the early stage or the average of the protein expression amounts of several generations before and after the Nth generation to the protein expression amount of the Mth generation at the late stage or the average of the protein expression amounts of several generations before and after the Mth generation.

[0158] In an embodiment, the computer program, when executed by the processor, further implements the following steps: performing cell segmentation processing on the observation image to obtain a plurality of single-cell images corresponding to the observation image; performing sampling and combination processing on the plurality of single-cell images to obtain a reference single-cell image combination; performing feature extraction on the reference single-cell image combination, and determining a stability value of the observation image based on a result of the feature extraction or a stability category determined according to the stability value.

[0159] In an embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the following steps:

[0160] obtaining an observation image of a cell strain to be predicted;

[0161] inputting the observation image to be observed into a stability prediction model for processing to obtain a stability value of the cell strain to be predicted or a stability category determined according to the stability value; the stability prediction model is determined according to images of a plurality of historical cell strains and a stability value of each historical cell strain or a stability category determined according to the stability value; and the stability value or the stability category determined according to the stability value is used to represent the stability of the protein expression amount of the cell strain.

[0162] In an embodiment, the computer program, when executed by the processor, further implements the following steps: taking the images of the plurality of historical cell strains as model inputs, taking the same stability value of the historical cell strain or the stability category determined according to the stability value as a training target of the model output, training an initial model to obtain the stability prediction model.

[0163] In an embodiment, the computer program, when executed by the processor, further implements the following steps: performing cell segmentation processing on the images of the historical cell strains to obtain a plurality of single-cell images corresponding to the images of the historical cell strains; performing feature extraction on the plurality of single-cell images corresponding to the images of the historical cell strains as a whole to obtain a feature set of the images of the historical cell strains; determining an initial model according to the feature set, taking the same stability value of the historical cell strain or the stability category determined according to the stability value as a training target of the model output, training the initial model to obtain the stability prediction model.

[0164] In an embodiment, the computer program, when executed by the processor, further implements the following steps: obtaining a protein expression amount at an early stage of culture of the historical cell strain and a protein expression amount at a late stage of culture of the historical cell strain;

[0165] determining the stability value of the historical cell strain or the stability category determined according to the stability value according to the protein expression amount at the early stage of culture and the protein expression amount at the late stage of culture.

[0166] In an embodiment, the computer program, when executed by the processor, further implements the following steps:

[0167] The average of the protein expression amount of the Nth generation in the early stage or the protein expression amount of several generations before and after the Nth generation and the average of the protein expression amount of the Mth generation in the late stage or the protein expression amount of several generations before and after the Mth generation are determined, and the stability value of the historical cell strain is determined according to the ratio of the average of the protein expression amount of the Nth generation in the early stage or the protein expression amount of several generations before and after the Nth generation to the average of the protein expression amount of the Mth generation in the late stage or the protein expression amount of several generations before and after the Mth generation, or the stability category determined according to the stability value.

[0168] In one embodiment, the computer program, when executed by the processor, further implements the following steps: performing cell segmentation processing on the observation image to obtain a plurality of single-cell images corresponding to the observation image; performing sampling and combination processing on the plurality of single-cell images to obtain a reference single-cell image combination; performing feature extraction on the reference single-cell image combination, and determining the stability value of the observation image or the stability category determined according to the stability value based on the result of the feature extraction.

[0169] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. Any reference to memory, storage, database or other medium used in the embodiments provided by the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0170] Any combination of the technical features of the above embodiments can be made. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combination of the technical features does not exist, it should be considered as the scope of the present application.

[0171] The above-described embodiments are merely illustrative of several embodiments of the present application, which are described in more detail and in a specific and detailed manner, but should not be construed as limiting the scope of the patent. It should be noted that for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present application, and these are all within the scope of the present application. Therefore, the scope of protection of the patent of the present application should be subject to the appended claims.

Claims

1. A method for predicting cell line stability, characterized in that, The method includes: Obtain observation images of the cell line to be predicted; The observed images are input into a stability prediction model for processing to obtain the stability value of the cell line to be predicted or a stability category determined based on the stability value. The stability prediction model is determined based on images of multiple historical cell lines and the stability value or stability category of each historical cell line. The stability value or stability category is used to characterize the stability of cell line protein expression levels, and the stability of cell line protein expression levels or the stability category is used to characterize the difference between the protein expression levels of progeny cell lines and the protein expression levels of previous generations of cell lines in the culture cycle. The images of the historical cell lines are images of any generation of cell lines in a culture cycle in which the historical cell lines have been cultured for multiple generations. The stability value is equal to the proportion of protein expression levels in the later stages of culture to the proportion of protein expression levels in the early stages of culture in the historical cell line protein expression levels obtained in the culture cycle. The stability prediction model processes the observed images of the cell line to be predicted as follows: Cell segmentation processing is performed on the observed images of the cell line to be predicted to obtain multiple single-cell images corresponding to the observed images of the cell line to be predicted; the observed images of the cell line to be predicted are images obtained by image acquisition of the cell line at any time during the cell culture cycle; The multiple single-cell images are sampled and combined to obtain a reference single-cell image combination; the reference single-cell image combination includes a preset number of single-cell images extracted from the multiple single-cell images of the cell line to be predicted. Feature extraction is performed on the reference single-cell image combination, and the stability value of the observed image of the cell line to be predicted or the stability category determined based on the stability value is determined based on the feature extraction result.

2. The method according to claim 1, characterized in that, The training process of the stability prediction model includes: Using the images of the multiple historical cell lines as model input, and with the model output being the same as the stability value of the historical cell lines or the stability category determined based on the stability value as the training objective, the initial model is trained to obtain the stability prediction model.

3. The method according to claim 2, characterized in that, The process of training the initial model using images of the multiple historical cell lines as model input and aiming to obtain the stability prediction model by ensuring that the model output matches the stability value of the historical cell lines or the stability category determined based on the stability value, includes: The images of the historical cell lines are subjected to cell segmentation processing to obtain multiple single-cell images corresponding to the images of the historical cell lines. Feature extraction is performed on multiple single-cell images corresponding to the historical cell line image to obtain the feature set of the historical cell line image; The model output is determined based on the feature set. The initial model is trained with the model output being the same as the stability value of the historical cell line or the stability category determined based on the stability value as the training target, to obtain the stability prediction model.

4. The method according to claim 2, characterized in that, The protein expression level in the early stage of culture includes the protein expression level of the historical cell line in the early stage of the culture cycle at generation N, or the average of the protein expression levels of several generations before and after generation N. The protein expression level in the later stage of culture includes the protein expression level of the historical cell line in the later stage of the culture cycle at generation M, or the average of the protein expression levels of several generations before and after generation M. The stability value of the historical cell line or the stability category derived from the stability value is determined based on the protein expression levels at the initial stage of culture and the protein expression levels at the later stage of culture, including: The average protein expression levels of the Nth generation or several generations before and after the Nth generation are determined, as well as the average protein expression levels of the Mth generation or several generations before and after the Mth generation. The stability value of the historical cell line or the stability category determined based on the ratio of the average protein expression levels of the Nth generation or several generations before and after the Nth generation to the average protein expression levels of the Mth generation or several generations before and after the Mth generation is determined.

5. A cell line stability prediction device, characterized in that, The device includes: The acquisition module is used to acquire observation images of the cell line to be predicted; The processing module is used to input the observed images into a stability prediction model for processing, to obtain the stability value of the cell line to be predicted or the stability category determined based on the stability value; the stability prediction model is determined based on images of multiple historical cell lines and the stability value or stability category determined based on the stability value of each historical cell line; the stability value or the stability category determined based on the stability value is used to characterize the stability of cell line protein expression levels, and the stability of cell line protein expression levels or the stability category is used to characterize the difference between the protein expression levels of progeny cell lines and the protein expression levels of previous generations of cell lines in the culture cycle; the images of the historical cell lines are images of any generation of cell lines in a culture cycle in which the historical cell lines have been cultured for multiple generations, and the stability value is equal to the proportion of protein expression levels in the later stages of culture to the proportion of protein expression levels in the early stages of culture in the historical cell line protein expression levels obtained in the culture cycle. The stability prediction model processes the observed images of the cell line to be predicted as follows: Cell segmentation processing is performed on the observed images of the cell line to be predicted to obtain multiple single-cell images corresponding to the observed images of the cell line to be predicted; the observed images of the cell line to be predicted are images obtained by image acquisition of the cell line at any time during the cell culture cycle; The multiple single-cell images are sampled and combined to obtain a reference single-cell image combination; the reference single-cell image combination includes a preset number of single-cell images extracted from the multiple single-cell images of the cell line to be predicted. Feature extraction is performed on the reference single-cell image combination, and the stability value of the observed image of the cell line to be predicted or the stability category determined based on the stability value is determined based on the feature extraction result.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images