Method for evaluating drug efficacy by fusing target molecule and time-concentration dependent cell phenotype

Abstract

The invention relates to a drug effect evaluation method for fusing target molecules and time-concentration dependent cell phenotypes, which comprises the following steps: S1, preparing samples including a modeling sample and a to-be-detected sample; s2, carrying out FRET imaging and cell fluorescence imaging; s3, carrying out FRET image processing and FRET efficiency calculation; s4, cell fluorescence image processing and feature extraction; s5, calculating a phenotype characterization value; s6, calculating an FRET characterization value; s7, basic drug response value calculation; and S8, comprehensive drug effect evaluation. Target molecule information and time-concentration dependent cell phenotypic response are fused, a comprehensive drug effect evaluation model is constructed, the effect of drugs on whole cells is reflected, and whether the drugs target specified molecules or not is specifically indicated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pharmacodynamic evaluation and live cell image analysis, and particularly relates to a pharmacodynamic evaluation method fusing target molecules and time-concentration-dependent cell phenotypes. BACKGROUND

[0002] Pharmacodynamic evaluation is a key link in biomedical research and clinical application, and is of great significance for in-depth understanding of drug action mechanism, optimization of treatment plan and guidance of personalized medication. Existing live cell pharmacodynamic evaluation methods mainly rely on analysis of cell phenotype changes or target molecule effects. In terms of cell phenotype analysis, high-content fluorescence image analysis (HCS) visualizes and quantifies cell and subcellular phenotypes, measures phenotypes in single cells and performs heterogeneity reaction analysis [Liberali P et al. Single-cell and multivariate approaches in genetic perturbation screens. Nature Reviews Genetics, 2015, 16(1): 18-32.]. For example, Alexandre et al. extracted more than 40 parameters including cell morphological changes by HCS for predicting pharmacodynamic effects of drugs [Bouzekri A et al. Multidimensional profiling of drug-treated cells by imaging mass cytometry. FEBS open bio, 2019, 9(9): 1652-1669.].

[0003] However, only cell phenotype changes can often only reflect the overall response of cells to chemical disturbances such as drugs, lack specificity, and are difficult to directly reveal the action mechanism between drugs and specific molecular targets. Given that targeted drugs generally trigger downstream cascade reactions by regulating key targets in signal pathways, and then affect cell phenotypes, therefore, analyzing drug target molecule information is crucial for achieving personalized and precise medication. Fluorescence resonance energy transfer (FRET) technology is currently the only imaging technology that can monitor weak, reversible and dynamic molecular events in live cells in situ and in real time, and can sensitively capture specific molecular target conformation changes or interaction strength changes induced by drugs, thereby directly revealing the molecular mechanism of drug binding or regulation with targets, and can provide molecular level evidence for pharmacodynamic evaluation. Combining cell phenotype information with target molecule action information can more comprehensively analyze drug effects, and help to distinguish drug action mechanisms.

[0004] Although the pharmacodynamic evaluation method of fusing cell phenotype information and target molecule information has been developed, the drug action is a dynamic time-concentration dependent process. The existing method often focuses on the analysis at a single time point or a fixed concentration in the experimental design or model construction, and fails to fully quantify the combined effects of the two key kinetic parameters of time and concentration on the pharmacodynamics. Therefore, there is an urgent need for a new method that can screen specific cell phenotype characteristics, systematically quantify the time-concentration effect, and realize the quantitative pharmacodynamic evaluation of fusing cell phenotype and target molecule information. SUMMARY

[0005] To solve the technical problems existing in the prior art, the present application provides a pharmacodynamic evaluation method of fusing target molecules and time-concentration dependent cell phenotypes. By deeply fusing the FRET information of target molecules and phenotype response, a comprehensive pharmacodynamic value calculation model is constructed, so that the final comprehensive pharmacodynamic value not only reflects the overall cell effect of the drug, but also specifically indicates the targeted action, which is significantly better than the existing evaluation method which only focuses on a single concentration or time point or has a simple fusion method. This enables the present application to efficiently distinguish and identify effective drugs with target specificity, thereby providing more reliable scientific basis for personalized precision medicine.

[0006] The method of the present application realizes the following technical scheme: a pharmacodynamic evaluation method of fusing target molecules and time-concentration dependent cell phenotypes, comprising the following steps:

[0007] S1, sample preparation, including modeling samples and test samples; selecting and culturing target cells to the logarithmic growth phase and dividing them into control and experimental groups; wherein the modeling samples include time gradient and concentration gradient modeling samples; the experimental group of the time gradient modeling samples is set to three or more time gradients, and the same positive drug is added at a fixed concentration, and the control group is set to the corresponding same time gradient without drug treatment, only adding the same volume of culture medium; the concentration gradient modeling sample is set to three or more concentration gradients with the same drug acting time, and the same positive drug is added, and the control group is set to the same time without drug treatment, only adding the same volume of culture medium; the experimental group of the test sample is set to any drug acting time and concentration, and different test drugs are added, and the control group is set to the corresponding same time without drug treatment, only adding the same volume of culture medium;

[0008] S2, FRET imaging and cell fluorescence imaging, for all modeling samples and test samples, N fields of view are selected for each sample to perform FRET three-channel imaging and cell fluorescence imaging, wherein N is greater than or equal to 15; the AA, DA and DD three-channel FRET images and organelle fluorescence images are taken for each field of view; finally, each sample has multiple fluorescence images in N fields of view;

[0009] S3, the FRET three-channel fluorescence images of the modeling samples and the to-be-tested samples obtained in step S2 are processed, the regions of interest in the images are selected, the three-channel gray values are obtained after background subtraction, and the donor angle FRET efficiency of the samples is calculated;

[0010] S4, cell fluorescence image processing is performed, a mask is generated using the FRET three-channel images, and is applied to the cell fluorescence images after noise removal in the same field of view for segmentation, and then cell phenotype features are extracted;

[0011] S5, the abnormal values in the cell phenotype features extracted in step S4 are removed, the correlation between the cell phenotype features and the drug action time and concentration is calculated, the phenotype features higher than the threshold are screened out and normalized, the weight of each phenotype feature is calculated, and the cell phenotype representation value is obtained by weighted summation;

[0012] S6, for the to-be-tested samples, the average values of the donor angle FRET efficiencies of the experimental group and the control group are calculated, and the absolute value of the relative change amount between the two is taken as the FRET representation value;

[0013] S7, a target molecule drug response model and a phenotype drug response model are established respectively, and the target molecule response model and the cell phenotype response model are fused to construct a basic drug response model;

[0014] S8, a comprehensive pharmacodynamic evaluation model is established based on the basic drug response model, the normalized comprehensive pharmacodynamic value EFF is calculated according to the to-be-tested drug concentration and action time, and the pharmacodynamic values of different drugs are evaluated.

[0015] The system adopts the following technical scheme: a pharmacodynamic evaluation system fusing target molecules and time-concentration-dependent cell phenotypes, comprising a sample preparation module, a FRET imaging and cell fluorescence imaging module, a FRET image processing module, a cell fluorescence image processing and feature extraction module, a cell phenotype representation value calculation module, a FRET representation value acquisition module, a basic drug response model construction module and a comprehensive pharmacodynamic evaluation module.

[0016] The sample preparation module includes a modeling sample and a to-be-tested sample; target cells are selected and cultured to the logarithmic growth phase and divided into a control group and an experimental group; the modeling sample includes time gradient and concentration gradient modeling samples; the experimental group of the time gradient modeling sample is set to three or more time gradients, and a fixed concentration of the same positive drug is added, and the control group is set to the same time gradient and not treated with the drug, only adding an equal volume of culture medium; the concentration gradient modeling sample is set to the same drug action time, three or more concentration gradients of the same positive drug are added, and the control group is set to the same time and not treated with the drug, only adding an equal volume of culture medium; the experimental group of the to-be-tested sample is set to any drug action time and concentration, and different to-be-tested drugs are added, and the control group is set to the same time and the cells are not treated with the drug, only adding an equal volume of culture medium;

[0017] The FRET imaging and cell fluorescence imaging module is used for FRET three-channel imaging and cell fluorescence imaging of N fields of view of each sample, wherein N is greater than or equal to 15; the AA, DA and DD three channels of each field of view are respectively imaged to obtain FRET images and organelle fluorescence images; and finally, each sample has multiple fluorescence images in N fields of view;

[0018] The FRET image processing module is used for selecting a region of interest in the FRET three-channel fluorescence images of the modeling sample and the to-be-tested sample obtained by the FRET imaging and cell fluorescence imaging module, obtaining three-channel gray values after background subtraction, and calculating the donor angle FRET efficiency of the sample;

[0019] The cell fluorescence image processing and feature extraction module is used for generating a mask by using the FRET three-channel images, applying the mask to the cell fluorescence images after noise removal in the same field of view, and then extracting cell phenotype features;

[0020] The cell phenotype characterization value calculation module is used for removing abnormal values in the cell phenotype features extracted by the cell fluorescence image processing and feature extraction module, calculating the correlation between the cell phenotype features and the drug action time and concentration, screening out phenotype features higher than a threshold value and performing normalization, calculating the weight of each phenotype feature, and performing weighted summation to obtain a cell phenotype characterization value;

[0021] The FRET characterization value acquisition module is used for calculating the average value of the donor angle FRET efficiency of the experimental group and the control group in the to-be-tested sample, and taking the absolute value of the relative change amount between the two as a FRET characterization value;

[0022] The basic drug response model construction module is used for establishing a target molecule drug response model and a phenotype drug response model, fusing the target molecule response model and the cell phenotype response model to construct a basic drug response model;

[0023] The comprehensive drug efficacy evaluation module establishes a comprehensive drug efficacy evaluation model through a basic drug response model, and calculates a normalized comprehensive drug efficacy value EFF according to a concentration of a drug to be tested and an action time, so as to evaluate drug efficacy of different drugs.

[0024] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0025] 1. The present application integrates cell morphological information and drug target molecule information, and overcomes the limitation that the existing method cannot fully consider the combined influence of drug action time and concentration. In the aspect of drug efficacy model construction, a time-concentration double Logistic cell phenotype drug response model is proposed. The model realizes the quantification of the drug action dynamic process through the fitting of the change of the phenotype characteristics with time and concentration and the reverse mapping of the phenotype representation value. More importantly, the present application creatively fuses the target molecule FRET information and the phenotype response, and constructs a comprehensive drug efficacy value (EFF) calculation model. The RES FRET of the phenotype response (RES PT ) in the model plays a key targeted correction role, so that the final comprehensive drug efficacy value EFF not only reflects the overall cell effect of the drug, but also specifically indicates the targeted action, which is significantly better than the existing evaluation method which only focuses on a single concentration or time point or has a simple fusion mode. This enables the present method to efficiently distinguish and identify effective drugs with target specificity, thereby providing a more reliable scientific basis for personalized precision medicine.

[0026] 2. The present application significantly improves the specificity of drug efficacy evaluation at the data acquisition and preprocessing level. Unlike the existing high-content screening method which directly extracts the phenotype characteristics of all cells in the field of view, the present application innovatively introduces a cell phenotype screening mechanism based on FRET images. The cells successfully transfected with the target protein are accurately identified and selected for phenotype characteristic extraction. The present method effectively eliminates the interference of non-responsive or background cells, greatly improves the targeted specificity of the analyzed phenotype data, and enables the subsequently constructed drug efficacy model to more accurately reflect the target effect of the drug, which is significantly better than the existing method in which a large amount of non-specific information is mixed in the phenotype characteristics. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 is a method flowchart of the present application;

[0028] Figure 2 is a cell phenotype image processing and feature extraction flowchart of the present application;

[0029] Figure 3 is a main workflow diagram of the calculation of the phenotype representation value of the present application;

[0030] Figure 4This is an example diagram of the modeling sample setting scheme according to an embodiment of the present invention;

[0031] Figure 5 (a) is a schematic diagram of the DD channel of a FRET image;

[0032] Figure 5 (b) is a schematic diagram of the DA channel of a FRET image;

[0033] Figure 5 (c) is a schematic diagram of the AA channel of the FRET image;

[0034] Figure 5 (d) is a schematic diagram of the cell nuclear channel;

[0035] Figure 5 (e) is a schematic diagram of mitochondrial channels;

[0036] Figure 6 (a) is a generalized logistic fitting curve of phenotypic characterization values ​​versus normalized time under the action of the drug afatinib in this embodiment;

[0037] Figure 6 (b) is a generalized logistic fitting curve of phenotypic characterization values ​​versus normalized concentration under the action of the drug afatinib in this embodiment;

[0038] Figure 7 The FRET efficiency E of the four test drugs in this embodiment is shown from the donor perspective. D Statistical box plot;

[0039] Figure 8 This is a schematic diagram showing the combined pharmacodynamic values ​​of the four drugs to be tested in this embodiment. Detailed Implementation

[0040] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0041] Example 1

[0042] like Figure 1 As shown, this embodiment illustrates a pharmacodynamic evaluation method that integrates target molecules and time-concentration-dependent cell phenotypes, comprising the following steps:

[0043] S1, sample preparation is performed, including modeling samples and samples to be tested; target cells are selected and cultured to the logarithmic growth phase and divided into control and experimental groups; wherein, the modeling samples include time gradient and concentration gradient modeling samples; the experimental group of the time gradient modeling samples is set to three or more time gradients, and a fixed concentration of the same positive drug is added, and the control group is set to the corresponding same time gradient and is not treated with drugs, but only an equal volume of culture medium is added; the concentration gradient modeling samples are set to three or more concentration gradients of the same drug for a certain time, and the same positive drug is added, and the control group is set to the same time and is not treated with drugs, but only an equal volume of culture medium is added; the experimental group of the samples to be tested is set to any drug action time and concentration, and different drugs to be tested are added, and the control group is set to the corresponding same time, and the cells are not treated with drugs, but only an equal volume of culture medium is added;

[0044] S2, FRET imaging and cell fluorescence imaging are performed, for all modeling samples and samples to be tested, N fields of view are selected for each sample to perform FRET three-channel imaging and cell fluorescence imaging, wherein N≥15; the AA, DA, and DD three-channel FRET images and organelle fluorescence images are photographed for each field of view; finally, each sample has multiple fluorescence images in N fields of view;

[0045] S3, for the FRET three-channel fluorescence images of the modeling samples and the samples to be tested obtained in step S2, the region of interest in the image is selected, the three-channel gray values are obtained after the background is deducted, and the donor angle FRET efficiency of the sample is calculated;

[0046] S4, cell fluorescence image processing is performed, a mask is generated using the FRET three-channel image, and is applied to the cell fluorescence image after noise removal in the same field of view to segment, and then the cell phenotype features are extracted;

[0047] S5, the abnormal values in the cell phenotype features extracted in step S4 are removed, the correlation of the cell phenotype features with the drug action time and concentration is calculated, the phenotype features higher than the threshold value are screened out and normalized, the weight of each phenotype feature is calculated, and a weighted sum is obtained to obtain a cell phenotype representation value;

[0048] S6, for the samples to be tested, the average values of the donor angle FRET efficiencies of the experimental group and the control group are calculated, and the absolute value of the relative change amount between the two is taken as the FRET representation value;

[0049] S7, a target molecule drug response model and a phenotype drug response model are respectively established, and a basic drug response model is constructed by fusing the target molecule response model and the cell phenotype response model;

[0050] S8. Establish a comprehensive pharmacodynamic evaluation model based on the basic drug response model, calculate the normalized comprehensive pharmacodynamic value EFF according to the concentration of the drug to be tested and the duration of action, and evaluate the pharmacodynamics of different drugs.

[0051] Specifically, in this embodiment, the sample preparation process in step S1 is as follows:

[0052] S11. Modeling Samples: Select and culture target cells to the logarithmic growth phase and divide them into control and experimental groups. All cells in the culture dishes are transfected with the selected drug target molecule FRET plasmid. After successful expression of the transfected plasmid, the corresponding subcellular organelles are labeled with fluorescent dyes. A known positive target drug D0 is selected through preliminary experiments. The experimental group includes time-gradient and concentration-gradient modeling samples. For time-gradient modeling samples, the drug concentration is fixed at the IC50 of the known positive drug, and then three or more time gradients are set, denoted as t1, t2, t3, ... t i For concentration gradient modeling samples, a fixed treatment time is established, determined through preliminary experiments, which is the time at which significant changes in cell phenotype occur due to drug action. Then, three or more concentration gradients are set, denoted as c1, c2, c3, ... c j Each petri dish is labeled as follows: This indicates that the time for adding the positive drug is t. x The concentration is c y The sample consists of 0 < x ≤ i and 0 < y ≤ j; t0 and c0 represent the control group.

[0053] S12. Sample to be tested, based on the drug to be tested (D1, D2, D3, ... D z The number of cultured cell samples (z) and the methods for plasmid transfection and subcellular organelle staining are consistent with those in step S11; all test drugs are fixed at the same time (t). ε and concentration c ε Different drugs D m The petri dishes were labeled as follows: Where m∈{0,1,2,...,z}.

[0054] Specifically, in this embodiment, step S3 further includes the following steps:

[0055] The FRET efficiency E of the donor angle in the experimental and control groups was calculated using the three-channel grayscale values ​​of FRET. D and acceptor-donor concentration ratio R c Specifically:

[0056]

[0057] F c =I DA -a(I AA -cIDD )-d(I DD -bI AA )

[0058] wherein, I DA is the fluorescence intensity detected in the acceptor channel using donor excitation light; I AA is the fluorescence intensity detected in the acceptor channel using acceptor excitation light; I DD is the fluorescence intensity detected in the donor channel using donor excitation light; a, b, c, d are spectral crosstalk coefficients, G is a sensitized quenching conversion factor, and the fluorescence intensity ratio k of the donor and the acceptor at the same concentration is measured in advance in the system calibration link; F C is the sensitized emission intensity;

[0059] According to the distribution of the donor-acceptor concentration ratio R c , the measurement range is selected, and the average value of the drug target donor angle FRET efficiency E D of each experimental group sample in the measurement range is calculated The average value of the donor angle FRET efficiency of the control group sample

[0060] Specifically, in this embodiment, the extraction process of the cell phenotype characteristics in step S4 includes:

[0061] S41, pre-process the FRET three-channel fluorescence images of the modeling samples and the to-be-tested samples obtained in step S2, including merging, removing noise, and enhancing contrast;

[0062] S42, segment and identify the successfully transfected cells from the FRET three-channel fluorescence images pre-processed in step S41, and generate a mask from the results;

[0063] S43, pre-process the subcellular organelle fluorescence images of the modeling samples and the to-be-tested samples obtained in step S2, including removing noise and enhancing contrast;

[0064] S44, segment and identify the subcellular organelles from the subcellular organelle fluorescence images processed in step S43, apply the mask generated in step S42 to the subcellular organelle fluorescence images obtained in step S43, screen the subcellular organelles of the successfully transfected cells, and extract the phenotype characteristics of the screened subcellular organelles.

[0065] Specifically, in this embodiment, the calculation process of the cell phenotype characteristic value in step S5 is as follows:

[0066] S51, clean the extracted cell phenotype characteristics, eliminate abnormal values in the cell phenotype characteristics, and calculate the correlation between the average value of the phenotype characteristics and the drug action time and concentration under the action of the known positive targeted drug based on the Pearson correlation coefficient.

[0067] S52, retaining f phenotypic characteristics with absolute values of the correlation between drug action time and concentration greater than a set threshold, calculating the change amount of the phenotypic characteristics from the control group and normalizing to obtain wherein 0 < p ≤ f, and f is an integer, i.e., the normalized characteristic value of the pth phenotype of drug m with drug concentration c and drug action time t;

[0068] S53, screening the maximum value of the above f phenotypic characteristics in all samples, denoted as The feature importance analysis is performed by calculating the change rate of the feature from the control group and assigning a weight, specifically:

[0069]

[0070] wherein W p represents the weight of the pth phenotypic characteristic, Ctrl p is the value of the pth phenotypic characteristic of the control group;

[0071] S54, weighted sum of the normalized characteristic values of each group of samples to obtain the phenotypic representation value of the sample, specifically:

[0072]

[0073] wherein PT m,t,c is the phenotypic representation value of drug m with drug concentration c and drug action time t, and the phenotypic representation values of all samples are calculated.

[0074] Specifically, in the present embodiment, the FRET signal representation value FRET m,t,c is calculated by the following formula:

[0075]

[0076] wherein FRET m,t,c represents the donor angle FRET efficiency of drug m with action time t and concentration c.

[0077] Specifically, in the present embodiment, the specific process of step S7 is as follows:

[0078] S71, establishing a phenotypic drug response model, using the phenotypic representation values in the modeling samples to construct a time-concentration double Logistic model, specifically:

[0079] For the concentration gradient, the PT value of the known positive targeted drug D0 used for modeling is subjected to Logistic fitting with the normalized drug concentration, specifically:

[0080]

[0081] where c norm is the normalized concentration, calculated by dividing the concentration by the selected maximum concentration, the fitted parameters include: k Pc is the concentration-dependent phenotype change growth rate parameter, a Pc is the offset parameter, n Pc is the shape parameter;

[0082] For the time gradient, the PT value of the known positive targeted drug D0 is modeled with a Logistic fit to the normalized action time, specifically:

[0083]

[0084] where t norm is the normalized time, calculated by dividing the time by the selected maximum time, the fitted parameters include: k Pt is the time-dependent phenotype change growth rate parameter, a Pt is the offset parameter, n Pt is the shape parameter;

[0085] The time weight β t and the concentration weight β c are determined by the parameter change rate or parameter vector length method, specifically:

[0086] (k Pt , a Pt , n Pt ) and (k Pc , a Pc , n Pc ) are considered as three-dimensional vectors, and the lengths of these two vectors are calculated:

[0087]

[0088] The time weight β t and the concentration weight β c are calculated:

[0089]

[0090] According to the measured phenotype characterization values of each drug to be tested in step S54, the values are respectively substituted into the formula PT(t norm ), PT(c norm ) of the known parameters in step S71, and the characterization values of the drug to be tested are inversely solved to correspond to the equivalent normalized time and the equivalent normalized concentration

[0091] The equivalent normalized time and concentration of the inverse mapping back to the modeling drug are weighted and their Euclidean distances are calculated, then normalized to obtain the phenotypic drug response model:

[0092]

[0093] S72. Establish a FRET drug response model, normalize the FRET signal representation values, and eliminate the dimensions of the phenotypic drug response values, specifically as follows:

[0094]

[0095] S73. Integrate the phenotypic drug response model with the FRET drug response model to establish a basic drug response model that integrates phenotype and target, specifically as follows:

[0096]

[0097] Specifically, in this embodiment, the process of obtaining the comprehensive efficacy value (EFF) in step S8 is as follows:

[0098] By substituting the normalized drug action time t and concentration c, along with their corresponding time and concentration weights, a comprehensive pharmacodynamic evaluation model is constructed to quantify the overall drug effect more comprehensively and specifically. Specifically:

[0099]

[0100] Here, EFF is the normalized comprehensive efficacy value. Although the denominator of the comprehensive efficacy evaluation model mentioned above is theoretically not 0 for any sample with added drugs, in order to ensure the stability of numerical calculation, a very small positive number, such as 1e-9, is added to the denominator in actual calculation to avoid potential problems. This processing method is often used in statistical and machine learning modeling.

[0101] Once the model is established, it can be used for quantitative efficacy evaluation of the drug to be tested. To ensure comparability among the drugs, the same time and concentration should be fixed.

[0102] Sample D to be tested m,t,c After imaging, a fluorescence image is obtained;

[0103] The FRET characterization value of the sample was obtained by calculation. m,t,c ;

[0104] Extracting cell phenotypic features;

[0105] The normalized eigenvalues ​​of the samples are obtained through calculation.

[0106] Obtain the phenotypic characterization value PT of the sample m,t,c ;

[0107] establishing a phenotype drug response model to obtain a phenotype drug response value

[0108] calculating a FRET drug response value

[0109] fusing and fusing to establish a basic drug response model to obtain a basic drug response value

[0110] normalizing the selected time and concentration to obtain t norm and c norm After that, and fusing together to calculate the comprehensive drug efficacy value EFF of the drug m when the drug concentration is c and the time is t in the comprehensive drug efficacy evaluation model m,t,c .

[0111] By comparing the comprehensive drug efficacy values EFF of different drugs or different treatment groups, the efficacy of the drugs is evaluated.

[0112] The time points of the time gradient samples should be 3 or more, i.e. i≥3; at this time, the fixed concentration is preferably the IC50 of the modeling drug;

[0113] The concentration of the concentration gradient sample should be 3 or more, i.e. j≥3; preferably, the concentration gradient with the IC50 of the drug as the median, for example: 0.5 times IC50, IC50 and 1.5 times IC50; preferably, the fixed time is the time when the cell phenotype changes stably positively, such as the time after the earliest time point when the cell phenotype changes significantly;

[0114] The drug target molecule FRET plasmid is a FRET plasmid connected with fluorescent proteins according to the drug target; preferably, one drug target molecule is connected with CFP fluorescent protein, and the other drug target molecule is connected with YFP fluorescent protein; such as the CFP-EGFR / YFP-GRB2 plasmid verified in the early stage of the present application; the mixed FRET plasmid with the mass (concentration) ratio of the donor and the acceptor being 1:X or X:1, X≥1, is transfected; preferably, the mixed FRET plasmid with the mass (concentration) ratio of the donor and the acceptor being 1:2 or 2:1; the donor fluorescent protein of the drug target FRET plasmid is ECFP; preferably, the donor plasmid ECFP purchased from the addgene plasmid library of the United States; the acceptor fluorescent protein of the drug target FRET plasmid is YFP; preferably, the acceptor plasmid YFP purchased from the addgene plasmid library of the United States;

[0115] The fluorescent protein includes but is not limited to CFP and YFP fluorescent proteins, and other pairs of donor and acceptor fluorescent probes and fluorescent proteins capable of FRET phenomenon can also be selected.

[0116] The sample container carrier is a container carrier capable of cell culture and imaging, including but not limited to cell culture dishes, well plates, glass slides, etc.

[0117] The corresponding subcellular organelles are labeled with fluorescent dyes, such as Hoechst 33258 fluorescent dye for labeling the nucleus and MitoTracker red dye for labeling mitochondria, but not limited to these two organelles, which can stain cell membranes or endoplasmic reticulum and other subcellular organelles; preferably one or more subcellular organelles; wherein the combination of two subcellular organelles is preferably the nucleus and mitochondria;

[0118] The cells are cancer cells, including but not limited to lung cancer, breast cancer, liver cancer, etc., and the cell lines are preferably A549 and H1975 cells;

[0119] The N fields are preferably selected by placing the sample container carrier (such as a culture dish) containing the transfected plasmid into a wide-field fluorescence microscope, and selecting N FRET three-channel fields with more than 3 cells in each field, preferably 5-10 cells.

[0120] The FRET three-channel and subcellular organelle images are 2048x2048 pixel, 1024x1024 pixel or 512x512 pixel images; preferably 2048x2048 pixel images;

[0121] The FRET three-channel image analysis can be completed by ZEN software of Carl Zeiss or other methods;

[0122] Cell phenotype image processing and feature extraction can be completed using machine learning models or image processing software, preferably CellProfiler software, which is a commercial open source image processing software that can extract features from subcellular organelle fluorescence images, mainly including area features, texture features, intensity features, intensity gradient features and correlation features, etc.

[0123] The data cleaning operation uses the Interquartile Range (IQR) method, which is a method used in statistics to describe the dispersion of data. It is calculated based on the quartiles of the data. First, sort the data from small to large, then divide the data into four equal parts, and the three partition points are the quartiles, which are the first quartile (Q1), the second quartile (Q2, i.e. the median) and the third quartile (Q3). The interquartile range is the difference between the third quartile and the first quartile (IQR = Q3-Q1), which can be used to identify outliers in the data. Generally, data points less than Q1-1.5xIQR or greater than Q3+1.5xIQR are considered outliers.

[0124] Phenotype features with correlation absolute value greater than a set value are retained, usually only features with correlation absolute value greater than 0.9 in single dimension of concentration or time are retained, considering feature extraction error, in both time and concentration dimensions, features with sum of time correlation and concentration correlation absolute value greater than 1.8 are preferably retained;

[0125] The normalization method is Min-Max normalization, the principle of which is to perform linear transformation on the original data to map the data to the interval [0, 1];

[0126] Logistic fitting is based on a generalized Logistic model (modified Richard model), which is a growth model widely used in biology, ecology and agricultural science, etc.

[0127] The comprehensive pharmacodynamic evaluation model, although the denominator is not 0 in theory for any drug-added test sample, in order to ensure the stability of numerical calculation, a very small positive number (for example, 1e-9) is added to the denominator in actual calculation to avoid potential problems;

[0128] When evaluating the pharmacodynamics of the test drug, the drug action time and concentration of different test samples should be consistent, the concentration is preferably the IC50 of the modeling drug, and the time is preferably an earlier time point at which the phenotype and FRET information of the modeling sample both change significantly, which can improve the efficiency of pharmacodynamic evaluation.

[0129] Example 2

[0130] 1. Plasmid construction

[0131] The given fluorescent protein donor-acceptor pair: the donor is a gene-encoded fluorescent protein CFP, and the acceptor is a gene-encoded fluorescent protein YFP.

[0132] Plasmids CFP-EGFR and YFP-GRB2: customized according to research needs, and the sequence of CFP-EGFR is verified by sequencing to be correct.

[0133] 2. Wide-field fluorescence microscopy imaging system

[0134] In this embodiment, a multimodal FRET fully automatic imaging analyzer (IX73, Olympus, Japan) is used as the wide-field fluorescence microscope in this embodiment to operate, which is equipped with a 60x / 1.42NA oil objective lens (Olympus, Japan) objective lens and a CMOS camera (Flash 4.0, Hamamatsu Photonics, Japan); a bright field LED light source; the excitation light source is a 130W mercury lamp (U-HGLGPS, Olympus, Japan), including 100%, 50%, 25%, 12%, 6% and 3% six gears.

[0135] 3. Cell culture and plasmid transfection

[0136] This embodiment selects non-small cell lung cancer (NSCLC) cell line A549, and the A549 cell line is obtained from Wuhan Sheng Biological Co., Ltd. in the experiment, and is cultured in DMEM medium containing 10% FBS (fetal bovine serum), 100 U / mL penicillin and 100 μg / mL streptomycin double antibody. During the experiment, the cells are cultured in a cell incubator with a constant temperature of 37°C and 5% carbon dioxide. The cell strains used in the test are early passages, the culture time is less than 3 months, and the mycoplasma contamination of the cells used for testing is periodically detected by using the luciferase mycoplasma detection kit. When transfection, the cells are first digested with trypsin (Xin Saimei), resuspended by blowing 1000 μL of a pipette, and transferred to a confocal culture dish for 24 h to the logarithmic growth phase, and Turbofect TM transfection. When preparing EGFR-CFP and GRB2-YFP plasmids, 200 ng of plasmid is transfected into each well, and 2 μL of Turbofect TM is used for each plasmid for 24 hours.

[0137] The specific steps of transfection are as follows:

[0138] Take a sterilized EP tube, add 100 μL of serum-free DMEM medium, then add 1-2 μL of transfection reagent, then add plasmid, gently blow 6-8 times, and then stand for 20 minutes;

[0139] After 20 minutes, add 100 μL of serum-free DMEM medium to the EP tube, and mix gently;

[0140] Move the mixture to the culture dish, and put the culture dish back into the incubator for 6 h, and then replace it with serum-containing DMEM medium.

[0141] 4. Drug treatment

[0142] 4.1, modeling sample

[0143] The modeling sample is divided into time gradient samples and concentration gradient samples. Through the previous study, Afatinib is determined as the positive targeted drug for modeling, the IC50 is 25.84 μM, and the phenotype changes significantly when the drug is used for 4 h at this concentration. In this embodiment, the fixed concentration of the time gradient sample is 25.84 μM, and three time points are set at 4 h, 8 h and 12 h. The fixed drug time of the concentration gradient sample is 8 h, and three concentrations are set at 12.92 μM, 25.84 μM and 38.76 μM. In addition to the experimental group, a control group also needs to be set, and an example of the sample setting scheme is shown in Figure 4 .

[0144] 4.2, sample to be tested

[0145] Four drugs were selected in this example, in addition to the second generation epidermal growth factor receptor tyrosine kinase inhibitor (EGFR-TKI) afatinib, the third generation EGFR-TKI almonertinib, NSCLC chemotherapy drug vinorelbine, and the first-line treatment drug for liver cancer, multi-target tyrosine kinase inhibitor sorafenib. The four drugs were treated with a concentration of 25.84 μM for 8 h as experimental group samples.

[0146] 5. Nucleus and mitochondria markers

[0147] In serum-free DMEM medium, fluorescent dyes Hoechst 33258 and MitoTracker red were added at a ratio of 1:1000. 30 min before sample imaging, the original medium in the sample culture dish was discarded, 200-400 μL of serum-free DMEM medium containing fluorescent dyes was added to each culture dish and placed in the incubator, and the nucleus and mitochondria were labeled after 10-20 min of staining. At this time, the serum-free DMEM medium was discarded and fresh DMEM medium containing serum was supplemented.

[0148] 6. Sample imaging

[0149] The motor-driven excitation rotating disc (IX3-RFACA) selected in this embodiment is matched with the IX73 rack, which carries 8 fluorescence excitation mirror groups (Cube1 to Cube8) and can be matched with 25 mm or 32 mm (diameter) filters. One excitation filter (Ex), one dichroic mirror (DM) and one emission filter (Em) can be placed in each Cube at the same time. The switching time of adjacent two Cubes is only 500 ms. A CCD camera is connected to the analyzer. The excitation light intensity is adjusted by selecting different levels of light source output and different attenuation levels of attenuation filter. The donor channel (DD channel) is composed of a 436 / 20 nm excitation filter and a 480 / 20 nm emission filter; the FRET channel (DA channel) is composed of a 436 / 20 nm excitation filter and a 530 / 20 nm emission filter; and the acceptor channel (AA channel) is composed of a 510 / 20 nm excitation filter and a 530 / 20 nm emission filter. A total of 3 imaging is recorded as DD, DA and AA as E-FRET data of one field of view. Under the same field of view, the nucleus channel (Hoechst channel) is switched by a 365 / 10 nm excitation filter and a 440 / 40 nm emission filter to shoot the nucleus fluorescence image. Under the same field of view, the mitochondria channel (Mito channel) is switched by a 630 / 20 nm excitation filter and a 660 / 20 nm emission filter to shoot the mitochondria fluorescence image. 15 fields of view are shot for each culture dish to obtain typical images as shown in Figure 5 (a)、 Figure 5 (b)、 Figure 5 (c)、 Figure 5 (d)、 Figure 5 (e).

[0150] 7. Cell phenotype image processing and feature extraction process

[0151] Cell phenotype image processing and feature analysis are performed by CellProfiler software, and the specific process is as shown in Figure 2Firstly, the FRET three-channel fluorescence images were preprocessed, and the three-channel images were merged, denoised and contrast-stretched through the built-in pipeline. Then, the preprocessed FRET images were segmented and the transfected cells were identified, and the mask was generated. Next, the nucleus and mitochondria fluorescence images of the same field of view were denoised and the contrast was enhanced. Finally, the nucleus and mitochondria fluorescence images were segmented and the objects were identified, and the mask generated before was applied to the identification results of the nucleus and mitochondria to screen out the nucleus and mitochondria corresponding to the successfully transfected cells, and the shape, intensity and other phenotype characteristics of the nucleus and mitochondria were extracted, and the phenotype characteristic data of the nucleus and mitochondria were output to the set path. The pipeline and its functions used in the CellProfiler software are shown in Table 1:

[0152] Table 1 CellProfiler pipeline and its functions required for extracting cell phenotype characteristics

[0153] Pipeline Function ImageMath Perform mathematical operations on images, such as image merging ReduceNoise Reduce image noise and smooth the image RescaleIntensity Adjust the intensity range of image pixels, such as stretching contrast IdentifyPrimaryObjects Identify and segment primary objects, such as cells, nuclei, mitochondria ConvertObjectsToImage Convert identified objects into images MaskObjects Use masks to retain or cover image regions MeasureObjectIntensity Measure pixel intensity characteristics of specified objects MeasureObjectSizeShape Measure size and shape characteristics of specified objects ExportToSpreadsheet Export all measurements to a spreadsheet file

[0154] 8. Phenotype characterization value calculation

[0155] Firstly, the data was preprocessed, and the exported CSV format table file was renamed as “sample name_Nuclei” or “sample name_Mito” according to the nucleus and mitochondria. The prefix “NU_” was added before the phenotype characteristics in the first row of all nucleus table files, and the prefix “MITO_” was added before the phenotype characteristics in the first row of the mitochondria table files, so as to distinguish the different subcellular organelle sources of the same characteristics. Each cell extracted 103 nucleus phenotype characteristics and 103 mitochondria phenotype characteristics, a total of 206 characteristics. 17 nucleus phenotype characteristics and 17 mitochondria phenotype characteristics related to number (Number) and location (Location) were removed, and then the remaining 86 nucleus phenotype characteristics and 86 mitochondria phenotype characteristics were combined into the same table for subsequent processing.

[0156] The main workflow of phenotype characterization value calculation is shown in Figure 3 The data was cleaned using the interquartile range method to remove outliers in all features, and then the mean value of each feature of the sample was calculated, and the mean value of each feature of each sample in the time gradient and concentration gradient was summarized to a new data table. Based on the Pearson correlation coefficient, the correlation of all features with time and concentration was calculated, and the key phenotype characteristics with the sum of the absolute values of the time correlation and the concentration correlation greater than or equal to 1.8 were screened out. The calculation method of the Pearson coefficient is as follows:

[0157]

[0158]

[0159] Screening | pp,t |+|ρ p,c The seven features with a value ≥ 1.8 are key phenotypic features.

[0160] Calculate the changes in the seven key phenotypic features of each sample compared to the control group, and normalize the data using Min-Max to map the data to the [0,1] interval.

[0161] The maximum value corresponding to the above 7 key phenotypic features selected from all samples is denoted as: Feature importance analysis and weighting are performed by calculating the rate of change of this feature compared to the control group. Specifically:

[0162]

[0163] The correlations and weights between the selected key phenotypes and drug action time and concentration are shown in Table 2.

[0164] Table 2. Correlation and weights of key phenotypes with drug action time and concentration.

[0165] Feature Name Time Correlation Concentration Correlation Weight NU_AreaShape_Eccentricity -0.9824 -0.9167 0.1884 NU_AreaShape_Zernike_2_2 -0.9287 -0.9495 0.2668 NU_AreaShape_Zernike_7_1 0.9739 0.8843 0.5315 NU_AreaShape_Zernike_8_4 -0.9942 -0.8927 0.3502 NU_AreaShape_Zernike_8_8 -0.9980 -0.9353 0.4615 MITO_AreaShape_MeanRadius 0.9877 0.8239 0.3062 MITO_AreaShape_MedianRadius 0.9927 0.8107 0.3148

[0166] Based on this, the weighted sum of the normalized values ​​of each feature is calculated and then divided by the sum of all weights to obtain the normalized phenotypic representation value. The phenotypic representation values ​​of the modeling samples are shown in Table 3.

[0167] Table 3. Phenotypic characteristics of samples with time gradient and concentration gradient.

[0168]

[0169] 9. Establishment of phenotypic drug response model

[0170] Using phenotypic values ​​from the modeling samples, a time-concentration dual logistic model is constructed. First, the selected time and concentration are subjected to Min-Max normalization, as follows:

[0171]

[0172]

[0173] For a fixed time point, the PT value of the known positive-target drug D0 used for modeling is fitted with a Logistic regression to the normalized drug concentration, specifically:

[0174]

[0175] The concentration-related phenotypic change rate of growth parameter k was obtained. Pc Offset parameter a Pc Shape parameter n Pc ;

[0176] For the fixed concentration point, the PT value of the known positive targeted drug D0 used for modeling was fitted with the normalized action time by Logistic, specifically:

[0177]

[0178] The time-dependent phenotype change growth rate parameter k Pt , the offset parameter a Pt , and the shape parameter n Pt were obtained; the generalized Logistic fitting curve results of the phenotype characterization values under the action of the drug Afatinib with respect to the normalized time and concentration in this embodiment are shown in Figure 6 (a) and Figure 6 (b). The specific values of the parameters are shown in Table 4:

[0179] Table 4. Parameters of the time and concentration gradient model of phenotype change

[0180] k Pt ]]> a Pt ]]> n Pt ]]> k Pc ]]> a Pc ]]> n Pc ]]> 4.2440 0.5583 0.5021 1.9800 0.3057 -0.5251

[0181] (k Pt , a Pt , n Pt ) and (k Pc , a Pc , n Pc ) were regarded as three-dimensional vectors, and the lengths of the two vectors were calculated:

[0182]

[0183] The time weight β t and the concentration weight β c were calculated:

[0184]

[0185] The calculated time weight and concentration weight are shown in Table 5:

[0186] Table 5. Time weight and concentration weight

[0187] β t ]]> β c ]]> 0.6754 0.3246

[0188] The measured phenotype characterization values of each test drug were respectively substituted into the above formulae PT(t norm ) and PT(c norm ) with known parameters, so that the characterization value of the test drug was inversely solved to correspond to the equivalent normalized time and the equivalent normalized concentration

[0189] The Euclidean distance is calculated by weighting the equivalent normalized time and concentration of the modeling drug respectively, and then the phenotype drug response model is obtained by normalization:

[0190]

[0191] 10. FRET drug response model establishment

[0192] The FRET three-channel images (I DD , I DA , I AA ) are selected from the acquired FRET three-channel images (I DD , I DA , I AA ), and the three-channel gray values of the ROI are obtained after background subtraction. The donor angle FRET efficiency (E D ) and the acceptor-donor concentration ratio (R c ) of each ROI are calculated, and the specific calculation is as follows:

[0193] The FRET efficiency of the donor angle is calculated according to the system calibration parameters and the acceptor-donor concentration ratio wherein, F c = I DA -a(I AA -cI DD )-d(I DD -bI AA ), I DA is the fluorescence intensity detected in the acceptor channel when the donor excitation light is used; I AA is the fluorescence intensity detected in the acceptor channel when the acceptor excitation light is used; I DD is the fluorescence intensity detected in the donor channel when the donor excitation light is used. The spectral crosstalk coefficients a, b, c, d, the sensitized quenching conversion factor G, and the fluorescence intensity ratio k of the same concentration of the donor and the acceptor are system calibration parameters, which can be measured in advance in the system calibration link.

[0194] The measurement range is selected according to the distribution of the acceptor-donor concentration ratio R c , and in this embodiment, 1.5≤R c ≤4 is selected as the measurement range. The average value of the drug target donor angle FRET efficiency E D of each sample in the experimental group in the measurement range is calculated, and the donor angle FRET efficiency E of the control sample is calculated. The results are as follows: Figure 7 .

[0195] The FRET signal characterization value FRET m,t,c is calculated by the following formula:

[0196]

[0197] The FRET signal characterization value is normalized to eliminate the dimension of the phenotype drug response value, and a FRET drug response model is established, which is specifically:

[0198]

[0199] 11. Establishment of the basic drug response model and the comprehensive drug efficacy evaluation model

[0200] The basic drug response model is composed of the FRET drug response model and the phenotype drug response model. The overall definition of drug efficacy here is the size of the effect of inducing drug death by the selected target, that is, only when both the phenotype and the target information change, we consider the drug to be an effective target drug, so logically they are "and" relationship, and in order to avoid some drugs with high toxicity leading to false positive results, the RES FRET of FRET information plays a key role in the targeted correction of the phenotype response (RES PT ), that is:

[0201]

[0202] The normalized drug action time t and concentration c, and the corresponding time and concentration weights are substituted to construct the comprehensive drug efficacy evaluation model, which is specifically:

[0203]

[0204] The final comprehensive drug efficacy value EFF not only reflects the overall cell effect of the drug, but also specifically indicates its targeted effect. Although the denominator is not zero in theory for any added drug sample, a very small positive number (e.g. 1e-9) is added to the denominator to avoid potential problems during actual calculation. Such processing method is commonly used in statistics and machine learning modeling.

[0205] 12. Application of drug efficacy evaluation of the test drug

[0206] After the above model is established, any test drug only needs to calculate the values of RES PT and RES FRET , and then calculate the comprehensive drug response value to perform quantitative efficacy evaluation. In this embodiment, the phenotype characterization values, FRET characterization values, drug response values and comprehensive drug efficacy values of the four test drugs, afatinib, amuvratinib, vinorelbine and sorafenib, under the same drug concentration and drug addition time are shown in Table 6, and the schematic diagram of the comprehensive drug efficacy value is shown in Figure 8 .

[0207] Table 6 Characterization values, drug response values and comprehensive drug efficacy values of test drugs

[0208] Drug PT FRET RES PT ]]> RES FRET ]]> RES SUM ]]> EFF Afatinib 0.5651 10.99％ 0.6667 49.47％ 21.80％ 35.80％ Almonertinib 0.4073 16.65％ 0.5189 74.98％ 31.26％ 51.34％ Vinorelbine 0.6253 4.48％ 0.7218 20.18％ 4.01％ 6.59％ Sorafenib 0.2601 4.61％ 0.3848 20.76％ 0.21％ 0.34％

[0209] According to the calculation results, under the cell line, culture environment and FRET target protein selection of the embodiment, the third generation targeted drug Almonertinib and the second generation targeted drug Afatinib are both positive targeted drugs, and the comprehensive drug efficacy value calculated in the system of the method is much greater than that of the chemotherapeutic drug Vinorelbine and the liver cancer drug Sorafenib; at the same time, the drug efficacy of Almonertinib is greater than that of Afatinib. Therefore, under the conditions of the embodiment, it is judged that the effect of Almonertinib and Afatinib on the EGFR-GRB2 target protein of the A549 cell line is stronger than that of the chemotherapeutic drug Vinorelbine and the liver cancer drug Sorafenib, which is consistent with the biological pharmacology theory.

[0210] Embodiment 3

[0211] Based on the same inventive concept, the present application also provides a drug efficacy evaluation system based on fusion target molecules and time-concentration dependent cell phenotypes, which comprises a sample preparation module, a FRET imaging and cell fluorescence imaging module, a FRET image processing module, a cell fluorescence image processing and feature extraction module, a cell phenotype characterization value calculation module, a FRET characterization value acquisition module, a basic drug response model construction module and a comprehensive drug efficacy evaluation module.

[0212] The sample preparation module comprises modeling samples and test samples; the target cells are selected and cultured to the logarithmic growth phase and divided into a control group and an experimental group; wherein the modeling samples include time gradient and concentration gradient modeling samples; the experimental group of the time gradient modeling samples is set to three or more time gradients, and a fixed concentration of the same positive drug is added, and the control group is set to the corresponding same time gradient and not treated with drugs, only adding an equal volume of culture medium; the concentration gradient modeling samples are set to three or more concentration gradients, and the same positive drug is added for a fixed time, and the control group is set to the same time and not treated with drugs, only adding an equal volume of culture medium; the experimental group of the test samples is set to any drug action time and concentration, and different test drugs are added, and the control group is set to the corresponding same time and the cells are not treated with drugs, only adding an equal volume of culture medium;

[0213] The FRET imaging and cell fluorescence imaging module is used for FRET three-channel imaging and cell fluorescence imaging of N fields of each sample, wherein N is greater than or equal to 15; the AA, DA and DD three-channel FRET images and organelle fluorescence images of each field are photographed; finally, each sample has multiple fluorescence images in N fields;

[0214] a FRET image processing module, configured to select a region of interest in a FRET three-channel fluorescence image of a modeling sample and a to-be-tested sample acquired in the FRET imaging and cell fluorescence imaging module, obtain three-channel gray values after background subtraction, and calculate a donor angle FRET efficiency of the sample;

[0215] a cell fluorescence image processing and feature extraction module, configured to generate a mask by using the FRET three-channel image, apply to a cell fluorescence image after noise removal in the same field of view, and then extract cell phenotype features;

[0216] a cell phenotype characterization value calculation module, configured to remove outliers in the cell phenotype features extracted in the cell fluorescence image processing and feature extraction module, calculate a correlation between the cell phenotype features and a drug action time and concentration, filter out phenotype features higher than a threshold value and perform normalization, calculate a weight of each phenotype feature, and perform weighted summation to obtain a cell phenotype characterization value;

[0217] a FRET characterization value acquisition module, configured to calculate an average value of donor angle FRET efficiencies of an experimental group and a control group in the to-be-tested sample, and take an absolute value of a relative change amount between the two as a FRET characterization value;

[0218] a basic drug response model construction module, configured to construct a basic drug response model by establishing a target molecule drug response model and a phenotype drug response model, and fusing the target molecule response model and the cell phenotype response model;

[0219] a comprehensive pharmacodynamic evaluation module, configured to establish a comprehensive pharmacodynamic evaluation model by the basic drug response model, calculate a normalized comprehensive pharmacodynamic value EFF according to a to-be-tested drug concentration and an action time, and evaluate pharmacodynamic effects of different drugs.

[0220] The above embodiments are preferred embodiments of the present application, but the embodiments of the present application are not limited by the above embodiments, and any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present application should be equivalent replacement methods, and are all included in the protection scope of the present application.

Claims

1. A method for the pharmacodynamic evaluation of a fusion target molecule and a time- concentration dependent cell phenotype, characterized in that, The method comprises the following steps: S1, sample preparation, including modeling samples and test samples; selecting and culturing target cells to the logarithmic growth phase and dividing into a control group and an experimental group; wherein the modeling samples include time gradient and concentration gradient modeling samples; the experimental group of the time gradient modeling samples is set to three or more time gradients, and a fixed concentration of the same positive drug is added respectively, and the control group is set to the corresponding same time gradient and is not treated with drugs, but only an equal volume of culture medium is added; the concentration gradient modeling samples are set to three or more concentration gradients of the same drug for a fixed time, and the same positive drug is added, and the control group is set to the same time and is not treated with drugs, but only an equal volume of culture medium is added; the experimental group of the test samples is set to any drug action time and concentration, and different test drugs are added respectively, and the control group is set to the corresponding same time, and the cells are not treated with drugs, but only an equal volume of culture medium is added; S2, FRET imaging and cell fluorescence imaging are performed, for all modeling samples and test samples, N fields of view are selected for each sample to perform FRET three-channel imaging and cell fluorescence imaging, wherein N≥15; the FRET images and organelle fluorescence images of AA, DA and DD three channels are photographed respectively for each field of view; finally, each sample has multiple fluorescence images in N fields of view; S3, for the FRET three-channel fluorescence images of the modeling samples and the test samples obtained in step S2, the region of interest in the image is selected, the three-channel gray values are obtained after the background is deducted, and the donor angle FRET efficiency of the sample is calculated; S4, cell fluorescence image processing is performed, a mask is generated using the FRET three-channel image, and is applied to the cell fluorescence image after noise removal in the same field of view to segment, and then cell phenotype features are extracted; S5, the abnormal values in the cell phenotype features extracted in step S4 are removed, the correlation of the cell phenotype features with the drug action time and the concentration is calculated, the phenotype features higher than a threshold value are screened out and normalized, the weight of each phenotype feature is calculated, and a weighted sum is obtained to obtain a cell phenotype representation value; S6, for the test samples, the average values of the donor angle FRET efficiencies of the experimental group and the control group are calculated, and the absolute value of the relative change amount between the two is taken as the FRET representation value; S7, a target molecule drug response model and a phenotype drug response model are established respectively, and the target molecule response model and the cell phenotype response model are fused to construct a basic drug response model; S8, a comprehensive drug efficacy evaluation model is established based on the basic drug response model, a normalized comprehensive drug efficacy value EFF is calculated according to the concentration and action time of the test drug, and the efficacy of different drugs is evaluated.

2. The method for evaluating the efficacy of a fusion target molecule and time-concentration dependent cell phenotype according to claim 1, wherein, The specific process of sample preparation in step S1 is as follows: S11. Modeling Samples: Select and culture target cells to the logarithmic growth phase and divide them into control and experimental groups. All cells in the culture dishes are transfected with the selected drug target molecule FRET plasmid. After successful expression of the transfected plasmid, the corresponding subcellular organelles are labeled with fluorescent dyes. A known positive target drug D0 is selected through preliminary experiments. The experimental group includes time-gradient and concentration-gradient modeling samples. For time-gradient modeling samples, the drug concentration is fixed at the IC50 of the known positive drug, and then three or more time gradients are set, denoted as t1, t2, t3, ... t i For concentration gradient modeling samples, a fixed processing time is used, and then three or more concentration gradients are set, denoted as c1, c2, c3, ... c j Each petri dish is labeled as follows: This indicates that the time for adding the positive drug is t. x The concentration is c y The sample consists of 0 < x ≤ i and 0 < y ≤ j; t0 and c0 represent the control group. S12. Sample to be tested, based on the drug to be tested (D1, D2, D3, ... D z The number of cultured cell samples (z) and the methods for plasmid transfection and subcellular organelle staining are consistent with those in step S11; all test drugs are fixed at the same time (t). ε and concentration c ε Different drugs D m The petri dishes were labeled as follows: Where m∈{0,1,2,...,z}.

3. The method for evaluating the efficacy of a fusion target molecule and a time-concentration-dependent cell phenotype according to claim 1, characterized in that, The specific process of step S3 further includes: FRET three-channel gray value calculation of the experimental group and the control group donor angle FRET efficiency E D and acceptor concentration ratio R c Specifically: F c = I DA -a(I AA -cI DD )-d(I DD -bI AA ) where I DA is the fluorescence intensity detected in the acceptor channel using donor excitation light; I AA is the fluorescence intensity detected in the acceptor channel using acceptor excitation light; I DD is the fluorescence intensity detected in the donor channel using donor excitation light; a, b, c, d are spectral crosstalk coefficients, G is the sensitized quenching conversion factor, and k is the fluorescence intensity ratio of equal concentrations of donor and acceptor, which is measured in advance during system calibration; F C is the sensitized emission intensity; According to the donor-acceptor concentration ratio R c The measurement range is selected according to the distribution, and the average value of the drug target donor angle FRET efficiency E of each experimental group sample in the measurement range is calculated D The average value of the drug target donor angle FRET efficiency of the control group sample The average value of the drug target donor angle FRET efficiency of the control group sample 4. The method for evaluating the efficacy of a fusion target molecule and a time-concentration-dependent cell phenotype according to claim 1, characterized in that, The extraction process of the cell phenotype features in step S4 includes: S41, pre-processing the FRET three-channel fluorescence images of the modeling samples and the test samples obtained in step S2, including merging, removing noise and enhancing contrast; S42, segmenting and identifying the successfully transfected cells of the FRET three-channel fluorescence images pre-processed in step S41, and generating a mask from the results; S43, preprocessing the subcellular organelle fluorescence images of the modeling samples and the to-be-tested samples obtained in step S2, including removing noise and enhancing contrast; S44, segmenting and identifying the subcellular organelles of the subcellular organelle fluorescence images processed in step S43, and applying the mask generated in step S42 to the subcellular organelle fluorescence images obtained in step S43 to screen the subcellular organelles of the successfully transfected cells; and extracting the phenotype features of the screened subcellular organelles.

5. The method for evaluating the efficacy of a fusion target molecule and time-concentration dependent cell phenotype according to claim 4, wherein, The calculation process of the cell phenotype characterization value in step S5 is as follows: S51, performing data cleaning on the extracted cell phenotype features to eliminate outliers in the cell phenotype features, and calculating the correlation between the average value of the phenotype features and the drug action time and concentration based on the Pearson correlation coefficient under the action of the known positive targeted drug; S52, retaining f phenotypic characteristics with absolute values of the correlation between the drug action time and the drug concentration greater than a set threshold, calculating the change amount of the phenotypic characteristics from the control group and normalized to obtain wherein 0 wherein 0 is an integer, i.e., the normalized characteristic value of the pth phenotype of the drug m with the drug action time t and the drug concentration c; S53、In all samples, the maximum value corresponding to the above-mentioned f phenotype characteristics is recorded as The feature importance analysis is performed by calculating the change rate of the feature in the control group, and the weight is assigned, specifically as follows: wherein W p represents the weight of the pth phenotypic trait, Ctrl p is the value of the pth phenotypic trait for the control group; S54, weighted sum normalization of the normalized feature values of each group of samples to obtain the phenotype characterization value of the sample, specifically: where PT is the phenotypic characterization value of the drug m with a drug concentration of c and a drug administration time of t, and the phenotypic characterization values of all samples are calculated. m,t,c where PT is the phenotypic characterization value of the drug m with a drug concentration of c and a drug administration time of t, and the phenotypic characterization values of all samples are calculated.

6. The method for evaluating the efficacy of a fusion target molecule and a time-concentration-dependent cell phenotype according to claim 3, characterized in that, The FRET signal characteristic value FRET in step S6 m,t,c is calculated from the following formula: where FRET m,t,c represents the donor angle FRET efficiency of drug m at time of action t and concentration c.

7. The method for evaluating the efficacy of a fusion target molecule and a time-concentration-dependent cell phenotype according to claim 4, characterized in that, The specific process of step S7 is as follows: S71, establishing a phenotype drug response model, using the phenotype characterization values in the modeling samples to construct a time-concentration double Logistic model, specifically: For the concentration gradient, the PT value of the known positive targeted drug D0 used for modeling is subjected to Logistic fitting with the normalized drug concentration, specifically: where c norm is the normalized concentration, calculated by dividing the concentration by the selected maximum concentration, and the fitted parameters include: k Pc is the concentration-dependent rate of change of the phenotype, a Pc is the offset parameter, n Pc is the shape parameter; For the time gradient, the PT value of the known positive targeted drug D0 used for modeling is subjected to Logistic fitting with the normalized action time, specifically: where t norm is the normalized time, calculated by dividing the time by the selected maximum value of time, the fitted parameters include: k Pt is the time-dependent rate of change of the phenotype, a Pt is the offset parameter, n Pt is the shape parameter; Determine the time weight β by parameter change rate or parameter vector length method t And the concentration weight β c Specifically: (k Pt ,a Pt ,n Pt ) and (k Pc ,a Pc ,n Pc ) are considered as three-dimensional vectors, the lengths of these two vectors are calculated: Computing the time weight β t and the concentration weight β c : According to the phenotypic characterization value of each drug to be tested measured in step S54, the value is substituted into the formula PT(t norm ), PT(c norm ) of the known parameters in step S71 respectively, and the characterization value of the drug to be tested is inversely solved to correspond to the equivalent normalized time and the equivalent normalized concentration The equivalent normalized time and concentration of the reverse mapping back to the modeling drug are weighted to obtain the Euclidean distance, and then normalized to obtain the phenotype part of the drug response model: S72, establishing a FRET drug response model, normalizing the FRET signal characterization value to eliminate the dimension of the phenotype drug response value, specifically: S73, fusing the phenotype drug response model and the FRET drug response model to establish a basic drug response model that fuses the phenotype and the target, specifically:

8. The method for evaluating the efficacy of a fusion target molecule and a time-concentration-dependent cell phenotype according to claim 1, characterized in that, The acquisition process of the comprehensive efficacy value EFF in step S8 is as follows: Substitute the normalized drug action time t and concentration c, and the corresponding time and concentration weights to construct a comprehensive efficacy evaluation model to quantify the comprehensive action effect of the drug, specifically: Wherein, EFF is the normalized comprehensive efficacy value.

9. The pharmacodynamic evaluation system based on the fusion target molecule of claim 1 and the pharmacodynamic evaluation method of time-concentration dependent cell phenotype, characterized by, The system comprises a sample preparation module, a FRET imaging and cell fluorescence imaging module, a FRET image processing module, a cell fluorescence image processing and feature extraction module, a cell phenotype characterization value calculation module, a FRET characterization value acquisition module, a basic drug response model construction module, and a comprehensive efficacy evaluation module; The sample preparation module comprises a modeling sample and a to-be-tested sample; target cells are selected and cultured to the logarithmic growth phase and divided into a control group and an experimental group; the modeling sample comprises time gradient and concentration gradient modeling samples; the experimental group of the time gradient modeling sample is set to three or more time gradients, and a fixed concentration of the same positive drug is added, and the control group is set to the same time gradient and not treated with the drug, but only with the same volume of culture medium; the concentration gradient modeling sample is set to the same drug action time, three or more concentration gradients of the same positive drug are added, and the control group is set to the same time and not treated with the drug, but only with the same volume of culture medium; the experimental group of the to-be-tested sample is set to any drug action time and concentration, and different to-be-tested drugs are added, and the control group is set to the same time and the cells are not treated with the drug, but only with the same volume of culture medium; The FRET imaging and cell fluorescence imaging module is used for FRET three-channel imaging and cell fluorescence imaging of N fields of view of each sample, wherein N is greater than or equal to 15; the AA, DA and DD three channels of each field of view are respectively imaged to obtain FRET images and organelle fluorescence images; finally, each sample has multiple fluorescence images in N fields of view; The FRET image processing module is used for selecting a region of interest in the FRET three-channel fluorescence images of the modeling sample and the to-be-tested sample obtained by the FRET imaging and cell fluorescence imaging module, obtaining three-channel gray values after background subtraction, and calculating the donor angle FRET efficiency of the sample; The cell fluorescence image processing and feature extraction module is used for generating a mask by using the FRET three-channel images, applying the mask to the cell fluorescence images after noise removal in the same field of view, and then extracting cell phenotype features; The cell phenotype characterization value calculation module is used for removing abnormal values in the cell phenotype features extracted by the cell fluorescence image processing and feature extraction module, calculating the correlation between the cell phenotype features and the drug action time and concentration, screening out phenotype features higher than a threshold value, normalizing the phenotype features, calculating the weight of each phenotype feature, and performing weighted summation to obtain a cell phenotype characterization value; The FRET characterization value acquisition module is used for calculating the average value of the donor angle FRET efficiency of the experimental group and the control group in the to-be-tested sample, and taking the absolute value of the relative change amount between the two as a FRET characterization value; The basic drug response model construction module is used for establishing a target molecule drug response model and a phenotype drug response model, fusing the target molecule response model and the cell phenotype response model to construct a basic drug response model; The comprehensive pharmacodynamic evaluation module is used for establishing a comprehensive pharmacodynamic evaluation model by the basic drug response model, calculating a normalized comprehensive pharmacodynamic value EFF according to the to-be-tested drug concentration and action time, and evaluating the pharmacodynamics of different drugs.

Images