Drug effect evaluation method combining drug target FRET information and bright field phenotype information

Abstract

The invention discloses a drug effect evaluation method combining drug target FRET information and bright field phenotype information. The method comprises the following steps: constructing a sample; carrying out three-channel imaging and bright field imaging on the sample and carrying out single cell segmentation; extracting and analyzing region and texture features; calculating a phenotype characterization value and an FRET characterization value; calculating a pharmacodynamic correction factor and pharmacodynamic parameters; calculating a drug effect parameter weight factor and a drug effect score; and taking the corresponding to-be-detected drug with the maximum drug effect score as the most effective targeted drug for the target spot. According to the method, the bright field phenotype characterization value and the FRET characterization value of the in-vitro living cells after drug action are measured in situ in real time on the basis of the image, the drug efficacy score is calculated, the most effective targeted drug for the target spot can be screened out for different drug target spots, and accurate drug efficacy evaluation is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of in vitro drug efficacy detection based on live cell image analysis, and particularly relates to a method for evaluating drug efficacy by combining FRET information of drug targets and bright-field phenotype information. Background Art

[0002] The method for analyzing phenotypic information based on bright-field microscopic images of live cells has been widely used in pathological diagnosis, drug efficacy evaluation, disease treatment, and other related fields [Zhao Y, Gu L., et al. Physical Cytometry: Detecting Mass-Related Properties of Single Cells. ACS Sens. 2022 Jan 28;7(1):21-36.]. Bright-field microscopy of live cells can perform non-invasive imaging of single cells, and the analysis of phenotypic information of bright-field microscopic images of live cells can obtain phenotypic information such as the area, shape, and texture of live cells. These bright-field phenotypic information of live cells are crucial for the morphological detection, diagnosis, and treatment of diseases. Perlman [Perlman ZE, Slack MD., et al. Multidimensional drug profiling by automated microscopy. Science. 2004 Nov 12;306(5699):1194-8.] et al. proposed a method for high-throughput cytological analysis through bright-field microscopic images. This method can determine the starting point of the relationship between drug actions at the system level based on the quantitative multidimensional phenotypic information of single cells, and can also predict the toxicity of new drugs. Ando [Ando D M, Mclean C, Berndl M. Improving Phenotypic Measurements in High-Content Imaging Screens[J]. 2017.] et al. reduced the bright-field phenotypic information of cells by the t-SNE method to detect subtle differences between cells after drug action.

[0003] However, previous methods for analyzing bright-field phenotypic information lacked a connection to the mechanisms of basic cellular processes [Amadoz A, Hidalgo MR, Çubuk C, Carbonell-Caballero J, Dopazo J. A comparison of mechanistic signaling pathway activity analysis methods. Brief Bioinform. 2019 Sep 27;20(5):1655-1668.]. In fact, the changes in bright-field phenotypic information after drug action can be understood as alterations in the operating modes of functional modules in cells. Such alterations are caused by different combinations of perturbations (mutations or gene expression changes) in the information of target molecules related to the functional modules. Usually, one change in bright-field phenotypic information may correspond to multiple different perturbations, and due to the real-time and diverse nature of the perturbation effects, one perturbation may also correspond to multiple different changes in bright-field phenotypic information. Therefore, it is of great significance to simultaneously analyze the information of target molecules and bright-field phenotypic information in situ and in real time in the same cell.

[0004] Among the existing techniques for measuring the information of drug target molecules, only the fluorescence resonance energy transfer (FRET) technique can achieve in situ, real-time, and quantitative measurement of drug target molecule information in living cells. Currently, relatively mature techniques such as gene sequencing and proteomics can only measure the information of drug target molecules after cell lysis. The quantitative FRET measurement method (E-FRET) based on 3-cube proposed by Zal [Zal, Tomasz, and Nicholas RJ Gascoigne. "Photobleaching-corrected FRET efficiency imaging of live cells." Biophysical journal 86.6 (2004): 3923-3939.] in 2004 can overcome both acceptor excitation crosstalk and donor emission crosstalk simultaneously. The E-FRET method eliminates spectral crosstalk through the combination of three channels: the donor detection channel (DD) when the donor excitation light is excited, the acceptor detection channel (DA) when the donor excitation light is excited, and the acceptor detection channel (AA) when the acceptor excitation light is excited, and simultaneously quantitatively calculates the FRET efficiency. E D The change in the information of target molecules after drug action can be characterized by the change in FRET efficiency. E D Therefore, it is of great significance to provide a method for evaluating the efficacy of drugs by combining the FRET information of drug target molecules and the bright-field phenotypic information of living cells. Summary of the Invention

[0005] The main object of the present invention is to overcome the disadvantages and deficiencies of the prior art, and to provide a method for evaluating the efficacy of drugs by combining the FRET information of drug targets and the bright-field phenotype information.

[0006] To achieve the above object, the present invention adopts the following technical solutions: One aspect of the present invention provides a method for evaluating the efficacy of drugs by combining the FRET information of drug targets and the bright-field phenotype information, including the following steps: Culturing cell samples and transfecting the selected FRET plasmid of drug target molecules, adding t a time gradient, c a concentration gradient, and n a solution of D n test drugs to obtain n × t × c test samples and t × c reference samples, where n , t, c ≥2 and is an integer; For the z th test drug, for the test sample at the i th time gradient and the j th concentration gradient and the reference sample at the i th time gradient and the j th concentration gradient C ij Select M fields of view for E-FRET three-channel imaging and bright-field imaging, and perform single-cell segmentation to obtain the three-channel images and bright-field images of each single-cell region in each field of view of the test sample and the reference sample C ij , where M ≥10 and is an integer; Extract the region and texture features of the bright-field images of the single-cell regions of the test sample and the reference sample C ij , and perform PCA principal component analysis to obtain the principal component matrix and principal component vector weight matrix of the bright-field images of each single-cell region in each field of view of the normalized test sample and the reference sample C ij ; Calculate the average FRET efficiency of the three-channel images of each single-cell region in each field of view of the test sample and the reference sample C ij respectively; According to the sample to be measured and the reference sample C ij for the principal component matrix and the average FRET efficiency of the bright-field images of each single-cell region in each field of view, calculate the phenotypic characterization value and the FRET characterization value of the sample to be measured ; According to the phenotypic characterization value and the FRET characterization value of the sample to be measured calculate the efficacy correction factor of the z th drug to be measured D z and the efficacy parameter of the sample to be measured ; According to the efficacy parameter of the sample to be measured calculate the efficacy parameter weight factor ; According to the efficacy parameter weight factor calculate the efficacy score z of the D z th drug to be measured , and use the drug to be measured corresponding to the maximum efficacy score as the most effective targeted drug for this target.

[0007] As a preferred technical solution, the single-cell segmentation is specifically as follows: For the sample to be measured and the reference sample C ij Select M fields of view for E-FRET three-channel imaging and bright-field imaging, and perform Gaussian smoothing, downsampling compression, and gray value compression operations on the generated M three-channel images in the fields of view in sequence, and then input them into a pre-trained CNN segmentation model to obtain single-cell segmentation images; upsample and restore the single-cell segmentation images as the mask images mask for the corresponding fields of view; mask the original three-channel images and bright-field images according to the mask images mask for the corresponding fields of view to obtain the three-channel images and bright-field images of each cell region in the corresponding fields of view; The Gaussian smoothing is specifically as follows: randomly generate a Gaussian kernel that conforms to a two-dimensional Gaussian distribution and has a size of n × n pixels, and then convolve the three-channel images with the Gaussian kernel to obtain the Gaussian-smoothed three-channel images; The specific implementation of masking is as follows: the background pixel value of the mask image mask is 0, and the pixel values of each single-cell region are assigned p values according to the number of single-cell regions; the mask image mask and the image to be masked are multiplied by the corresponding elements of the matrix to obtain the masked image.

[0008] As a preferred technical solution, the extraction of the sample to be measured and the reference sample C ij of the bright-field image region and texture features, and perform PCA principal component analysis on the obtained feature vectors to obtain the normalized sample to be measured and the reference sample C ij of the principal component matrix and principal component vector weight matrix of the bright-field image of each single-cell region, specifically: Extract the p th single-cell region of the bright-field image q feature vectors and denote them as , and calculate the sample mean vector ; Subtract q feature vectors from the mean value, that is, each feature vector subtracts the corresponding mean value, as shown in the following formula: ; ; Calculate the covariance matrix of the p th single-cell region after centering processing X p,centered , as shown in the following formula: ; Perform eigenvalue decomposition on the covariance matrix Cov p to obtain eigenvalues and the corresponding eigenvectors ; select the eigenvectors corresponding to the first k eigenvalues as the principal components to form a projection matrix W p , each column is an eigenvector; Project the data set into the vector subspace formed by the first k eigenvectors: Y p = X p,centered * W p ; Obtain the principal component matrix Y p after dimensionality reduction, with a size of N× k ; Each principal component is expressed as a linear combination of the original features: , where is the p -th weight vector corresponding to the i’’ -th principal component of the W p single-cell region; where the principal component vector weight matrix represents the contribution ratio of each principal component, sorted from largest to smallest; and ; Similarly, the principal component matrix Y 0 and the principal component weight matrix W 0 of the reference sample are obtained.

[0009] As a preferred technical solution, the calculation of the FRET efficiency is specifically as follows: According to the donor and acceptor fluorescence intensities of the three channels in the three-channel image I DD 、I AA 、I DA , we get: F C = I DA - a*I AA - d *I DD ; ; where F C is the fluorescence intensity sensitized to the acceptor in the FRET sample, a is the direct excitation correction factor; d is the leakage correction factor; G is the correction factor for different detection efficiencies in the two channels; E D is the FRET efficiency.

[0010] As a preferred technical solution, the phenotypic characterization value of the sample to be measured C ij and the FRET characterization value are calculated according to the principal component matrix, the principal component vector weight matrix, and the average FRET efficiency of the bright-field images of each single-cell region in each field of view of the reference sample and the FRET characterization value , specifically as follows: ; ; Wherein, S m-p and FERT m-p are the phenotypic characterization value and the FRET characterization value of the th m single-cell region in the p th field of view of the sample to be measured, m ∈[1, M and is an integer; is the principal component matrix of the reference sample, is the i’’ th principal component vector of the reference sample; is the principal component matrix of the p th single-cell region of the sample to be measured, is the p th i’’ th principal component vector of the th single-cell region of the sample to be measured; p is the weight factor of the i’’ th principal component vector of the E Dp and E D0 are the average FRET efficiencies of the th p single-cell region of the sample to be measured and the reference sample C ij ; Calculate the averages of S m-1 , S m-2 ,…, S m-p and FERT m-1 , FERT m-2 ,…, FERT m-p as the average phenotypic characterization value and the average FRET characterization value of the m th field of view, denoted as and ; Calculate the mean of , as the phenotypic characterization value and the FRET characterization value of the sample to be measured .

[0011] As a preferred technical solution, said according to the sample to be measured Phenotypic characterization value and FRET characterization value Calculate the z th drug to be tested D z pharmacodynamic correction factor and the sample to be tested pharmacodynamic parameters , specifically: Equation 1: ; Equation 2: ; Substitute Equation 1 into Equation 2 and calculate under the constraint condition α + β = 1 of α , β values as the pharmacodynamic correction factor of the drug to be tested D z ; Substitute the pharmacodynamic correction factor of the drug to be tested D z into Equation 1 to obtain the pharmacodynamic parameter α , β of the sample to be tested .

[0012] As a preferred technical solution, the pharmacodynamic parameter weight factor is calculated according to the pharmacodynamic parameter of the sample to be tested , specifically: ; wherein, c i represents the value of the i th concentration gradient, t j represents the value of the j th time gradient.

[0013] As a preferred technical solution, the pharmacodynamic score of the z th drug to be tested D z is calculated according to the pharmacodynamic parameter weight factor , specifically: ; wherein, represents the pharmacodynamic parameter weight factor at c i concentration and t j time, represents at c i ​Concentration and t j pharmacodynamic parameters at different time points.

[0014] Another aspect of the present invention further provides a pharmacodynamic evaluation system that combines drug target FRET information and bright-field phenotype information, which is applied to the above-mentioned pharmacodynamic evaluation method that combines drug target FRET information and bright-field phenotype information, and includes a sample preparation module, an imaging and segmentation module, a feature extraction and analysis module, a pharmacodynamic parameter calculation module, and a pharmacodynamic scoring module; The sample preparation module is used to culture cell samples and transfect the selected drug target molecule FRET plasmid, and add t a time gradient, c a concentration gradient, and n several test drugs D n to obtain n × t × c test samples and t × c reference samples, where n , t, c ≥2 and is an integer; The imaging and segmentation module is used to perform E-FRET three-channel imaging and bright-field imaging on the z th test drug at the i th time gradient and the j th concentration gradient of the test sample and the reference sample at the i th time gradient and the j th concentration gradient C ij Select M fields of view for E-FRET three-channel imaging and bright-field imaging, and perform single-cell segmentation to obtain the three-channel images and bright-field images of each single-cell region in each field of view of the test sample and the reference sample C ij , where M ≥10 and is an integer; The feature extraction and analysis module is used to extract regional and texture features from the bright-field images of the single-cell regions of the test sample and the reference sample C ij , and perform PCA principal component analysis to obtain the principal component matrix and principal component vector weight matrix of the bright-field images of each single-cell region in each field of view of the normalized test sample and the reference sample C ij ; respectively calculate the test sample and the reference sample Cij The average FRET efficiency of the three-channel images of each single-cell region under each field of view; respectively according to the sample to be tested and the reference sample C ij the principal component matrix and the average FRET efficiency of the bright-field images of each single-cell region under each field of view in the sample to be tested to calculate the phenotypic characterization value and the FRET characterization value ; The pharmacodynamic parameter calculation module is used to calculate the pharmacodynamic correction factor of the th test drug and the FRET characterization value of the sample to be tested according to the phenotypic characterization value z and the pharmacodynamic parameter D z of the sample to be tested; calculate the pharmacodynamic parameter weight factor according to the pharmacodynamic parameter of the sample to be tested; ; The pharmacodynamic scoring module is used to calculate the pharmacodynamic score of the th test drug according to the pharmacodynamic parameter weight factor z and use the test drug corresponding to the maximum pharmacodynamic score as the most effective targeted drug for this target. D z Another aspect of the present invention also provides a storage medium storing a program, which when executed by a processor, implements the above-mentioned pharmacodynamic evaluation method combining drug target FRET information and bright-field phenotypic information.

[0015] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0016] (1) The present invention first proposes to calculate the pharmacodynamic score by in-situ real-time measurement of the bright-field phenotypic characterization value and the FRET characterization value of in-vitro living cells after the action of drugs based on images, realizing accurate pharmacodynamic evaluation of drugs.

[0017] (2) Using the pharmacodynamic score obtained by the present invention, the most effective targeted drug for a specific drug target can be screened out, and subsequent popularization and application of this method to clinical practice can guide doctors in drug selection.

[0018] ​​(3) The drug efficacy scores obtained using the present invention are applicable to different drug targets, cell lines, cancers, and targeted drugs. Therefore, the present invention can greatly promote the scope of the field of in vitro live cell drug efficacy detection technology based on image analysis, and further improve the application scope of FRET detection technology in drug efficacy detection.

[0019] (4) Compared with the traditional drug efficacy detection method based only on phenotype, the method of the present invention can not only identify the targets of drug action, but also accurately give the drug efficacy scores of each drug, which is convenient for horizontal comparison between different drugs targeting the same target. In addition, the method of the present invention has realized modular program calculation, and only by inputting an image, the corresponding drug efficacy score can be directly calculated, reducing the use threshold and improving the efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is the algorithm flowchart of a drug efficacy evaluation method combining drug target FRET information and bright field phenotype information according to an embodiment of the present invention; Figure 2 is the schematic diagram of the image preprocessing result according to an embodiment of the present invention; Figure 3 is the box plot of the phenotypic characterization values of 10 dishes of cells in 5 groups S of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of the present application.

[0022] Example 1: As Figure 1 shown, this embodiment provides a drug efficacy evaluation method combining drug target FRET information and bright field phenotype information, including the following steps: (1) Sample preparation.

[0023] S1. Samples to be tested: According to n types of drugs to be tested ( D 1, D 2,..., D z ,..., D n ), cell samples are cultured, and each drug to be tested is cultured according to t time gradients and c concentration gradients designed in the experimentt × c Dish cells were cultured until the cells adhered to the dish. Each dish of cells was transfected with the selected FRET plasmid of the drug target molecule. After the successful expression of the transfected plasmid, according to the respective concentration gradients of each drug c to t × c dish cells, the solution of each drug to be tested was added to obtain n × t × c test samples, and the test sample numbers were recorded as , indicating the test sample of the z-th drug to be tested at the i -th time gradient and the j -th concentration gradient; where n , t , c ≥2 (and is an integer); S2. Reference samples: Culture t × c dish cells, and then transfect the selected FRET plasmid of the drug target molecule into t × c dish cells. After the successful expression of the transfected plasmid, t × c reference samples were obtained, and the reference sample numbers were recorded as C ij , indicating the reference sample of the i -th time gradient and the j -th concentration gradient; (2)E-FRET imaging and bright-field imaging.

[0024] S3. For the test samples obtained in step 1 (the test sample of the 1st drug at the 1st concentration gradient at the 1st time gradient D ), select M fields of view for E-FRET three-channel imaging and bright-field imaging to obtain three-channel and bright-field images: DD -1 , DD -2 , …, DD -M , AA -1 , AA -2 , …, AA -M , DA -1 , DA -2 , …, DA -M , BF-1 , BF -2 , …, BF -M ; wherein, M ≥10 (and is an integer); S4. Replace the sample to be tested with the sample to be tested , and repeat step 3 to obtain the three-channel images and bright-field images of fields of view of the sample to be tested; and so on until the three-channel images and bright-field images of M fields of view of the sample to be tested (the sample to be tested when adding the th concentration gradient of the i rd drug z at the D z th time gradient) are obtained; j M S5. Select C M 11 fields of view for E-FRET three-channel imaging and bright-field imaging of the reference sample obtained in step 2 (the reference sample at the 1st concentration gradient and the 1st time gradient), and obtain three-channel and bright-field images: DD -1 , DD -2 , …, DD -M , AA -1 , AA -2 , …, AA -M , DA -1 , DA -2 , …, DA -M , BF -1 , BF -2 , …, BF -M , -M ; wherein, M ≥10 (and is an integer); S6. Replace the reference sample C 11 with the reference sample C 12 , and repeat step 5 to obtain the C 12 M fields of view of the E-FRET three-channel images and bright-field images of the reference sample; and so on until the reference sample is obtained​​C ij of M a three-channel E-FRET image and a bright-field image of the field of view; (3)Image preprocessing and image segmentation.

[0025] S7. Obtain the three-channel and bright-field images of the first field of view of the sample to be tested in step 3 : DD -1 , AA -1 , DA 1, BF -1 . First, perform Gaussian smoothing, downsampling compression, and grayscale value compression operations on the three-channel images DD -1 , AA -1 , DA 1, in sequence to obtain a smoothed three-channel image with a size of 512×512 pixels and a grayscale range between 0-255; then input the smoothed three-channel image into the trained CNN segmentation model to obtain an accurate single-cell segmentation image; then upsample the single-cell segmentation image to restore it to a single-cell region image with a size of 2048×2048 pixels of the original three-channel image size, which is used as the mask image mask for this field of view. Finally, mask the original three-channel and bright-field images according to the mask image mask of this field of view to obtain the three-channel and bright-field images of each cell region under this field of view, denoted as DD -1-1 , DD -1-2 ,…, DD -1-p , AA -1-1 , AA -1-2 ,…, AA -1-p , DA -1-1 , DA -1-2 ,…, DA -1-p , BF -1-1 , BF -1-2 ,…, BF -1-p ; where p is the number of single cells in this field of view ( p is an integer); (4)Image feature extraction and image analysis.

[0026] S8. For the three-channel and bright-field images of the single-cell regions obtained in step 7, determine whether the image is a bright-field image. If it is, execute step 9; if not, execute step 10. S9. For the first single-cell region in the first field of view of the sample to be tested BF -1-1 , input the bright-field image into CellProfiler to extract the regional features and texture features of the image, and obtain q feature vectors, denoted as x 1, x 2,..., x q ; where q >10 (an integer); then perform PCA principal component analysis on the q feature vectors to obtain the principal component matrix and the principal component weight matrix, with sizes of m × k , 1× k respectively; then take the first k principal component vectors among the first m principal component vectors of the principal component matrix as the representative features of the bright-field image, and simplify the principal component matrix and the principal component vector weight matrix to sizes of ×10, 1×10; finally, normalize the principal component vector weight matrix to obtain the normalized principal component matrix BF -1-1 and the principal component vector weight matrix of the bright-field image of the first single-cell region in the first field of view of the sample to be tested ; and so on, calculate the principal component matrix and the principal component vector weight matrix p of the bright-field image of the BF -1-p th single-cell region in the first field of view of the sample to be tested and the principal component vector weight matrix ; S10. For the FRET three-channel images of the first single-cell region in the first field of view of the sample to be tested DD -1 , AA -1 , DA -1 , calculate the gray values of the three channels pixel by pixel I DD 、I AA 、I DA , and calculate the average FRET efficiency of this single-cell region according to the E-FRET efficiency calculation formula ED1 ; And so on, calculate the average FRET efficiency of the p th single-cell region in the first field of view of the sample to be measured E Dp ; S11. Repeat steps 7 - 10 for the reference sample obtained in step 5, and calculate the mean of the principal component matrix and FRET efficiency of the single-cell regions of the obtained reference sample to obtain the principal component matrix and average FRET efficiency E D0 ; (5) Calculation of pharmacodynamic correction factor and pharmacodynamic parameters.

[0027] S12. According to the principal component matrix and average FRET efficiency data results of the first field of view of the sample to be measured obtained in steps 8 - 11, perform the following calculations: , , , …… ; , , …… ; wherein, S 1-1 and FERT 1-1 respectively represent the phenotypic characterization value and FRET characterization value of the first cell region in the first field of view of the sample; Calculate the S 1-1 , S 1-2 ,…, S 1-p and FERT 1-1 , FERT 1-2 ,…, FERT 1-p as the average values of the phenotypic characterization value and FRET characterization value of the first field of view of the sample to be measured , denoted as and ; And so on, calculate the average values M of the th field of view and ; Calculate the , as the average value of the sample to be measured The phenotypic characterization values and FRET characterization values are denoted as and .

[0028] S13. Repeat step 12 to obtain the drug to be tested D 1's t × c The phenotypic characterization values and FRET characterization values of the dish cells are denoted as , … , , … ; Then perform the following calculations: ; ; Substitute R ij expression into expression, and calculate the α + β = 1 when 's α , β values, and use them as the efficacy correction factor of the drug to be tested D 1.

[0029] S14. Substitute the α , β values calculated in step 13 into R ij expression to obtain the efficacy parameter of the sample to be tested R ij ; (6) Calculate D z The efficacy score of the drug S15. Repeat step 14 to obtain the concentration gradient of i , the time gradient j when D 1 drug's t × c efficacy parameters R 11 , R 12 ,…, R tc . Generate the efficacy parameter weight factor for the concentration gradient of i , the time gradient j according to the following formula: ; Among them, c i represents thei The value of a concentration gradient t j indicating the j value of the nth time gradient.

[0030] S16. Perform the following calculations based on the pharmacodynamic parameters obtained in step 14 and the weight factors of the pharmacodynamic parameters obtained in step 15: ; Obtain D the pharmacodynamic score of Drug 1 ; S17. Repeat steps 7 - 16 to obtain n the pharmacodynamic scores of multiple drugs ; Finally, based on obtain the drug with the highest pharmacodynamic score D max as the most effective targeted drug against this target.

[0031] The cell sample described in the present invention is an in vitro cancer cell sample cultured in the same container carrier; wherein, the container carrier is a container carrier capable of realizing cell culture and imaging in the art, including but not limited to cell culture dishes, well plates, glass slides, etc.; the in vitro cancer cell sample includes but not limited to breast cancer cells, lung cancer cells, liver cancer cells, etc.; the present invention selects a conventional cell culture dish for cell culture and imaging.

[0032] Preferably, the number of types of n test drugs described in step 1 is 2 - 10 types; preferably 4 types.

[0033] Preferably, the number of the time gradients i described in step 1 is 2 - 4; preferably 3.

[0034] Preferably, the number of the concentration gradients j described in step 1 is 2 - 4; preferably 3.

[0035] Preferably, the preferred time gradients described in step 1 are 6 h, 12 h, 18 h.

[0036] Preferably, the concentration gradient j described in step 1 should be the IC50 concentration of the drug, preferably within 20 μm above and below the IC50 concentration.

[0037] Preferably, the drug target FRET plasmid described in step 1 is a FRET plasmid linked with a fluorescent protein designed according to the drug target; preferably, one drug target molecule is linked with a CFP fluorescent protein and the other drug target molecule is linked with a YFP fluorescent protein; more preferably, the CFP-BCL-XL / YFP-BAK and CFP-EGFR / YFP-GRB2 drug target molecules verified in the early stage of the present invention. The ratio of the number (concentration) of the donor and the receptor in the transfected selected drug target FRET plasmid is 1: n or n :1 mixed FRET plasmid, n ≥1; preferably, the ratio of the number (concentration) of the donor and the receptor is 1:2 and 2:1 mixed FRET plasmids.

[0038] Preferably, the fluorescent proteins described in step 1 include but are not limited to CFP and YFP fluorescent proteins, and can be other pairs of donor-acceptor fluorescent probes and fluorescent proteins that can exhibit FRET phenomenon, etc.

[0039] Preferably, the sample container carrier described in step 1 is a container carrier capable of realizing cell culture and imaging, including but not limited to cell culture dishes, well plates, glass slides, etc.

[0040] Preferably, the donor fluorescent protein of the drug target FRET plasmid described in step 1 is ECFP; preferably, the donor plasmid ECFP purchased from the American addgene plasmid library.

[0041] Preferably, the receptor fluorescent protein of the drug target FRET plasmid described in step 1 is YFP; preferably, the receptor plasmid YFP purchased from the American addgene plasmid library.

[0042] Preferably, the cells described in steps 1 and 2 are preferably MCF7 and A549 cells.

[0043] Preferably, the selection of the M fields of view in step 3 is preferably achieved by the following method: placing the sample container carrier (such as a culture dish) of the cells transfected with the plasmid under a wide-field fluorescence microscope, and selecting M the number of cells within N ≥5 (and an integer) fields of view, and the number of cells in each field of view is more than 5; preferably 5-10.

[0044] Preferably, the three-channel and bright-field images DD, AA, DA, and BF images in step 3 are images with 2048×2048 pixels, 1024×1024 pixels, or 512×512 pixels; preferably, the image with 2048×2048 pixels.

[0045] Preferably, the Gaussian smoothing operation described in step 7 is implemented by the following calculation: randomly generate a Gaussian kernel with a size of n × n pixels that conforms to a two-dimensional Gaussian distribution, and then convolve the image with this Gaussian kernel to obtain the Gaussian-smoothed image; preferably, the size of the Gaussian kernel is 5×5.

[0046] Preferably, the downsampling operation described in step 7 is implemented in the following manner: change the image within the s × s window of the original image into one pixel, and the value of this pixel is the mean value of all pixels within the window, which is called downsampling the image by a factor of s; preferably, the downsampling ratio is 4.

[0047] Preferably, the grayscale value operation described in step 7 is implemented in the following manner: divide the grayscale value of each pixel of the original image by 257 simultaneously to compress the grayscale value range from 0 to 65535 to 0 to 255; preferably, the compression ratio is 257.

[0048] Preferably, the CNN segmentation model described in step 7 is a pre-trained model, and preferably, the pre-trained dataset is ImageNet.

[0049] Preferably, the masking of the original three-channel and bright-field images according to the mask image mask in step 7 is implemented in the following manner: the background pixel value of the mask image mask is 0, and the pixel value of each single-cell region is assigned according to the number of single-cell regions p values; multiply the mask image mask and the image to be masked element by element in the matrix to obtain the masked image.

[0050] Preferably, in step 8, whether the image is a bright-field image is judged according to the label of the input image.

[0051] Preferably, CellProfiler described in step 9 is a commercial image processing software, which can extract large categories of features such as regions, textures, and intensities from images; for bright-field images, preferably, the features are regions and textures; the q feature vectors, preferably q > 100, and more preferably q > 200.

[0052] Preferably, in step 9, the first 10 most important principal component vectors are taken as the representative features of the bright-field image, and preferably, the number of principal component vectors is 10.

[0053] The basic principle of the present invention is as follows: (a) Obtain the phenotypic characterization value of each dish of cells by using the principal component analysis method: The present invention extracts the pBright-field images of individual single-cell regions regarding texture and region q q feature vectors, denoted as ; For q q feature vectors, calculate the sample mean vector : , formula (1); Then centralize the q feature vectors, i.e., subtract the corresponding mean from each feature: , formula (2); , formula (3); Calculate the covariance matrix of the centralized feature matrix X p,centered : , formula (4); Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues and the corresponding eigenvectors ; Select the eigenvectors corresponding to the first k eigenvalues as the principal components to form a projection matrix W p (each column is an eigenvector); Project the dataset into the vector subspace formed by the first k eigenvectors: Y p = X p,centered * W p , formula (5); Obtain the principal component matrix Y after dimensionality reduction, with size N × k ; Among them, N represents the number of single-cell features extracted. Each principal component can be expressed as a linear combination of the original features: , where is the p th weight vector corresponding to the i’’ th principal component of the W p single-cell region; among them, the principal component vector weight matrix represents the contribution ratio of each principal component, usually sorted from largest to smallest; and .

[0054] Similarly, obtain the principal component matrix of the reference sampleY 0 and the principal component weight matrix W 0.

[0055] Said according to the sample to be measured and the reference sample C ij The principal component matrix, principal component vector weight matrix, and average FRET efficiency of the bright-field image of each single-cell region under each field of view are calculated for the sample to be measured of the phenotypic characterization value :[[]]END]] , formula (6); wherein, is the i’’ th principal component vector of the reference sample; is the p th single-cell region of the sample to be measured and the i’’ th principal component vector; is the p th single-cell region of the sample to be measured and the i’’ th weight factor of the principal component vector.

[0056] (b) Obtain the FRET characterization value FRET according to the E-FRET method: For the FRET sample imaged by the three-channel method, the fluorescence intensities of the three-channel donor and acceptor obtained by imaging I DD 、I AA 、I DA . It can be written as a function of the photophysical parameters, instrument parameters, the number of donors n D and the number of acceptors n A in the considered pixel, the fluorophore, and the FRET probability (efficiency) E D : , formula (7); , formula (8); , formula (9); wherein, L A is the excitation intensity of the selected wavelength for exciting the fluorescent molecule A ; L D is the excitation intensity of the selected wavelength for exciting the fluorescent molecule D ; is A at the A excitation wavelength; is D the absorption cross-section at the excitation wavelength; D is the absorption cross-section at the excitation wavelength; D is A the quantum yield of; is A the quantum yield of; is D the detection efficiency of photons emitted in the detection channel is A the detection efficiency of photons emitted in the detection channel A ; is the detection efficiency of photons emitted in the detection channel D is D the detection efficiency of photons emitted in the detection channel is A the detection efficiency of photons emitted in the detection channel D ;

[0057] a To simplify the calculation, four correction factors are defined by formula (10) to simplify formulas (7)-(9): the direct excitation correction factor is; the leakage correction factor is d ; the correction factor for different detection efficiencies in the two channels is G ; the correction factor for different excitation efficiencies in the two channels is β ;

[0058] , , , , formula (10); The fluorescence intensity sensitized to the acceptor in the FRET sample is obtained by subtracting the fluorescence of the acceptor excitation crosstalk and the fluorescence of the donor emission crosstalk in the FRET channel ( I DA ) F C : F C = I DA - a*I AA - d *I DD , formula (11); The FRET efficiency can be expressed as: , formula (12); According to the FRET efficiency E D the FRET characterization value of each dish of cells can be calculated: , formula (13); Among them, E D0 is the average FRET efficiency of the dish for the reference sample; E D is the average FRET efficiency of the dish for the sample to be tested.

[0059] After obtaining the phenotypic characterization values and FRET characterization values of each dish of cells, it is necessary to construct the corresponding pharmacodynamic function S and FRET for each dish of cells. The constructed pharmacodynamic function should be able to objectively reflect the changes in the independent variables 𝑆 and 𝐹𝑅𝐸𝑇. That is, the larger 𝑆 the larger; the larger 𝐹𝑅𝐸𝑇 R ij the larger, and the pharmacodynamic function can be adjusted by coefficients R ij for different sensitivities of the two independent variables, indicating that different drugs have different responses to phenotype and FRET. And there should be constraint conditions: when 𝑆 = 0, R ij = 0; when 𝐹𝑅𝐸𝑇 = 0, R ij = 0. It indicates that only when the drug simultaneously produces changes in FRET characterization values and phenotypic characterization values does the drug have the efficacy of a targeted drug. According to the above requirements and constraint conditions, the present invention constructs a pharmacodynamic function R ij related to the phenotypic characterization value S and the FRET characterization value, and the expression is: , formula (14); Among them, α is the FRET characterization value pharmacodynamic correction factor; β is the phenotypic characterization value pharmacodynamic correction factor. α and β are only related to the drug, the target, and the cell line.

[0060] (d) For formula (14), it is necessary to determine the values of the pharmacodynamic correction factors α and β . The present invention hopes to calculate relatively large R ij from the phenotypic characterization values and 𝐹𝑅𝐸𝑇 obtained by treating the same cell line with the same drug; since the variance is a measure of the degree of dispersion when measuring a random variable or a set of data, the variance formula for discrete data is: , formula (15); (e) Construct the drug efficacy variance formula according to the variance formula. Substitute formula (14) into formula (15) to obtain: , formula (16); Since are monotonically increasing functions of the drug efficacy correction factors α and β respectively; that is, α the larger is, β the larger is. Therefore, constraint conditions need to be added. Also, since the drug efficacy correction factors α and β represent the different sensitivities of the drug efficacy function R ij to the phenotypic characterization value and the FRET characterization value. Therefore, add the constraint condition: α + β = 1, formula (17); (f) Calculate under the constraint condition of formula (17) to obtain α , β as the drug efficacy correction factors.

[0061] (g) Substitute the drug efficacy correction factors α , β calculated by formula (16) into formula (14), and the R ij drug efficacy parameters can be obtained.

[0062] (h) The concentration gradient of the sample to be tested D z is i , and the time gradient j The drug efficacy parameters R ij can be written as t × c matrix, as shown below: , formula (18); (i) Each term in the matrix of formula (18) represents the drug efficacy parameters obtained from different time gradients t and concentration gradients c . According to the Hill function expression: , formula (19); Among them, H is the drug efficacy; a is the asymptote under the curve; b is the asymptote on the curve; γ is the x of the inflection point of the curveValue; δ δ is the steepness of the curve exponential control curve, generally taken as 1. The Hill function can reflect the relationship between the drug efficacy and the time gradient and concentration gradient. Therefore, the Hill function is used to weight-correct the efficacy parameters of the test sample to obtain the weight function It is: , formula (20); Since the weight function is related to the monotonicity of the independent variable t and the base and the size compared to 1. When , the weight function is a monotonically increasing function with respect to the independent variable t; when , the weight function is a monotonically decreasing function with respect to the independent variable t ; This application hopes that the weight function and the independent variable t should always maintain a monotonically decreasing relationship. Therefore, the weight function is corrected to: , formula (21); (j) For the test sample D z , its final efficacy score can be expressed as: , formula (22); (k) Substitute formulas (21) and (18) into formula (22), and the efficacy score D z of the test sample can be obtained.

[0063] Example 2: In another aspect of this example, by setting specific parameters for the solution in Example 1, those skilled in the art can better implement the solution of this application.

[0064] S1. Plasmid construction.

[0065] Given fluorescent protein donor-receptor pair: The donor is the gene-encoded fluorescent protein CFP, and the receptor is the gene-encoded fluorescent protein YFP; Construct plasmid YFP-BAK: PCR the BAK target gene using CFP-BAK as a template, and insert the BAK target gene between the double digestion sites (XhoI and BamHI) of the vector pEYFP-C1-Drp1 (Addgene, #45160). Sequence verification shows that the sequence of YFP-BAK is correct.

[0066] Plasmid CFP-BCL-XL: Gifted by Professor A.P. Gilmore.

[0067] Source: Plasmids ECFP-BAK (#31501), pcDNA3-CFP (CFP) (#13030), and pcDNA3-YFP (YFP) (#13033) were purchased from Addgene.

[0068] S2. Wide-field fluorescence microscopy imaging system.

[0069] The wide-field fluorescence microscope is produced by Olympus Corporation of Japan, model IX73. The light source is a mercury lamp of the Olympus HGLGPS series from Japan. The objective lens is an oil immersion lens with a magnification of 60 and a numerical aperture of 1.42 (60 × 1.42NA, UPLFLN40XO, Olympus, Japan). The wide-field fluorescence microscopy imaging system has an excitation wheel equipped with four excitation filters, a wheel equipped with eight cubes (each cube can install one excitation filter, dichroic mirror, and emission filter), an electric emission wheel equipped with six emission filters, and an external CCD camera. The excitation light wavelength is selected by rotating the excitation wheel.

[0070] S3. Cell culture and plasmid transfection.

[0071] In the experiment, MCF7 cells were obtained from the School of Medicine, Jinan University (Guangzhou, China). The culture method is as follows: The cells were cultured in DMEM medium containing 15% fetal bovine serum (Vistech) and 100 mg / ml streptomycin and penicillin (Gibco) in a 37 °C constant temperature incubator (containing 5% carbon dioxide). When transfecting, the cells were first digested with trypsin (Xinsaimai), resuspended by pipetting with a 1000 µl pipette, and transferred to a confocal dish for 18 h. When the cells grew to 70-90%, the plasmid was transiently transfected into the cells using a cationic polymer transfection reagent (ExFect Transfection Reagent, Novoprotein). The specific steps of transfection are as follows: The specific steps of transfection are as follows: (1) Take a sterilized EP tube, first add 100 μL of serum-free DMEM medium to it, then add 1-2 μL of the transfection reagent to it, and then add 1-2 μL (500-600 ng / μL) of the plasmid to it. Gently pipette 6-8 times and then let it stand for 20 minutes; (2) After 20 minutes, add another 100 μL of serum-free DMEM medium to the EP tube and gently mix it; (3)Wash the cells in the culture dish 2 - 3 times with serum-free DMEM medium or PBS, mainly to wash away dead cells and other contaminants, and then transfer the mixture in step (2) above to the culture dish, and put the culture dish back into the incubator for 6 h; When Z-VAD-FMK treatment is required, 20 µm of Z-VAD-FMK has been present in DMEM since the start of cell transfection.

[0072] S4, Drug induction.

[0073] Transfect CFP-BCL-XL and YFP-BAK into MCF-7 cells together according to the above plasmid transfection steps. A total of 10 dishes of transfected cells were cultured in the experiment, and the 10 dishes of cells were divided into 5 groups. The following treatments were carried out respectively after 24 h of culture: Group 1: Control group, only transfected and added 20 µm of apoptosis inhibitor Z-VAD-FMK; Group 2: Targeted drug treatment group, on the basis of Group 1, add 1 µm of apoptosis drug A1331852 targeting BCL-XL and BAK; Group 3: Non-targeted drug treatment group, on the basis of Group 1, add 1 µm of apoptosis drug ABT-199 non-targeting BCL-XL and BAK; Group 4: Non-lethal drug treatment group, on the basis of Group 1, add 100 µm of digestive juice Trypsin non-targeting BCL-XL and BAK; Group 5: Necrosis drug treatment group, on the basis of Group 1, add 600 mm of necrosis drug NaCl non-targeting BCL-XL and BAK; S5, Sample imaging process.

[0074] After inducing MCF-7 cells according to the above drug induction steps, perform E-FRET imaging and bright-field imaging in groups. The wide-field fluorescence microscope is configured as follows: Use light with wavelengths of 436 / 20 nm and 500 / 20 nm as the excitation light channels for CFP and YFP respectively, and use 480 / 40 nm and 535 / 30 nm as the detection light channels for CFP and YFP respectively. Under 436 nm excitation, the detector collects images of all detection channels. Under 500 nm excitation, the detector only collects images of the 535 nm detection channel. A total of 3 images are recorded as DD, DA, AA as the E-FRET data for one field of view; then turn off the fluorescence excitation in the same field of view and turn on the bright-field light source for 1 bright-field imaging, recorded as BF as the bright-field data for one field of view. The imaging times of different drug treatment groups are different: Group 1: Imaging was performed at 5 h and 11 h respectively. Group 2: Imaging was performed at 5 h and 11 h respectively. Group 3: Imaging was performed at 5 h and 11 h respectively. Group 4: Imaging was performed at 1 h and 2 h respectively. Group 5: Imaging was performed at 1 h and 2 h respectively.

[0075] S6. Data processing.

[0076] The flowchart of data processing is as Figure 1 shown.

[0077] S6.1. Image preprocessing and image segmentation were performed according to bright-field images and FRET three-channel images. The specific process is as follows: For each group in groups 1 - 5 obtained in steps 4 and 6, 10 fields of view were randomly selected at each time point (a total of 100 fields of view and 400 images, including 100 DD, DA, AA, and BF images each). Then the images were divided into bright-field images BF and FRET three-channel images DD, DA, AA. Image preprocessing and image segmentation were performed on the bright-field images and FRET three-channel images respectively, as follows: As Figure 2 shown in a, for control group 1, 1 field of view was selected for E-FRET and bright-field imaging 5 h after drug treatment to obtain DD, DA, AA, and BF images of this field of view. The size of all four images was 2048*2048 pixels. Then Gaussian smoothing was performed on the DD, DA, AA, and BF images in sequence to obtain the smoothed images DD1, DA1, AA1, and BF1; then downsampling compression and normalization were performed on the DD1, DA1, and AA1 images to convert the 16-bit images of 2048*2048 pixels into 8-bit grayscale images of 512*512 pixels DD2, DA2, AA2; then DD2, DA2, AA2 were segmented into single cells using a pre-trained CNN segmentation model and upsampled to obtain a single-cell segmentation image mask of 2048*2048 pixels ( Figure 2 b); finally, mask was used as a mask template to mask the smoothed images DD1, DA1, AA1, and BF1 to obtain FRET three-channel images and bright-field images containing only single-cell regions ( Figure 2 c, d).

[0078] S6.2. Determine the FRET characterization value according to the FRET three-channel images.

[0079] For the 5 groups obtained in steps 4 and 6, at two time points for each group, and 10 fields of view at each time point, calculate the average FRET characterization value of the 10 fields of view as the FRET characterization value of this group at this time point. Specifically: S6.2.1. Calculate the average FRET efficiency of each group E D。

[0080] For control group 1, one field of view was selected for E-FRET and bright-field imaging 5 h after drug treatment to obtain the DD, DA, AA, and BF images of this field of view. Then, background recognition and subtraction were performed on the three-channel images, including the following four steps: (1) Count the number of pixels with different gray values in each image separately to obtain the relationship diagram between the corresponding count and gray value; (2) Determine the first peak intensity between I BG (the minimum gray value of intensity) and I min (the median gray value of intensity); (3) Take the intersection of the three-channel images I middle as the final background template BG; (4) Subtract BG from the corresponding original image, and finally represent the corrected image as I BG (negative pixel intensity is represented as zero). After obtaining the background of each image and correcting the original image, calculate the per-pixel FRET efficiency of this field of view according to the per-pixel intensity value I corr and formula (12). Similarly, calculate the per-pixel FRET efficiency of 10 fields of view in turn, and obtain the average FRET efficiency of 10 fields of view according to the number of valid pixels I DD 、I AA 、I DA as the FRET efficiency after 5 h of drug treatment for group 1, that is E D = 0.3948; and so on, calculate the FRET efficiency after treatment at two time points for 5 groups (Trypsin is a non-lethal treatment, and the FRET efficiency does not change between 1 h and 2 h of treatment, so the same FRET efficiency is used for 1 h and 2 h of treatment), as shown in Table 1: E D As the FRET efficiency after 5 h of drug treatment for group 1, that is E D = 0.3948; and so on, calculate the FRET efficiency after treatment at two time points for 5 groups (Trypsin is a non-lethal treatment, and the FRET efficiency does not change between 1 h and 2 h of treatment, so the same FRET efficiency is used for 1 h and 2 h of treatment), as shown in Table 1: ; Table 1. FRET efficiency of 10 dishes of cells in five groups E D 。

[0081] S 6.2.2. Calculate the normalized FRET characterization value for each group.

[0082] Obtain the FRET efficiency for each group according to the above method E DAfter that, calculate the FRET characterization values for each group according to formula (13). Among them E D0 Calculate according to the mean value of the FRET efficiency ED of group 1, that is, the control group, that is E D0 = 0.4007; for group 2, the calculated FRET characterization value is FRET = 0.3431; and so on to obtain the FRET characterization values of 5 groups at two time points, as shown in Table 2: ; Table 2. FRET characterization values of 10 dishes of cells in five groups.

[0083] Finally, normalize the 10 FRET characterization values of 10 dishes of cells in 5 groups to obtain the normalized FRET characterization values as shown in Table 3: ; Table 3. Normalized FRET characterization values of 10 dishes of cells in five groups.

[0084] S6.3. Determine the phenotypic characterization values according to the bright-field images.

[0085] For the 5 groups obtained in steps 4 and 6, two time points for each group, and 10 fields of view for each time point, calculate the average phenotypic characterization value of all single-cell regions in the 10 fields of view as the phenotypic characterization value of the group at that time point. Specifically: S 6.3.1. Calculate the principal component matrix of the single-cell region.

[0086] After 5 h of drug treatment in group 1, select 1 field of view and obtain a bright-field image containing only single-cell regions according to 6.1 Image preprocessing and image segmentation ( Figure 2 b). Extract features from the bright-field image of the single-cell region for regional features and texture features. A total of 181 feature vectors are obtained, including 25 regional features and 156 texture features. Regional features include cell area, shape factor, and compactness, etc.; texture features include texture entropy, texture variance, and texture mean, etc. Then perform principal component analysis on the 181 feature vectors, and take the first 10 most important principal component vectors as the representative features of the single-cell region to obtain the principal component matrix as and the principal component vector weight matrix .

[0087] S 6.3.2. Calculate the phenotypic characterization value of the single-cell region.

[0088] Repeat step 6.3.1 to obtain the mean value of the principal component matrix of the single-cell region in all fields of view after 5 h and 11 h of drug treatment in group 1 as Similarly, repeat step 6.3.1. After obtaining the principal component matrix and the principal component vector weight matrix of all single-cell regions in one field of view selected by group 2 after 5 h of drug treatment, calculate the phenotypic characterization value of each single-cell region according to the following formula S : ; S 6.3.3. Calculate the normalized phenotypic characterization value of each group.

[0089] Repeat step 6.3.2 to calculate the phenotypic characterization values of 654 single-cell regions after treatment at two time points from group 2 to group 5 (where group 2: 179 single-cell regions; group 3: 223 single-cell regions; group 4: 72 single-cell regions; group 5: 180 single-cell regions). Draw a box plot according to the group and time point as Figure 3 shown. Obtain the average phenotypic characterization value of each group and time point S (Trypsin is a non-lethal treatment, and the phenotypic characterization values at 1 h and 2 h of treatment do not change, so the same FRET efficiency is used for 1 h and 2 h of treatment), see Table 4: ; Table 4. Phenotypic characterization values of cells in 10 dishes of five groups S .

[0090] Finally, normalize the 10 phenotypic characterization values at two time points of the 5 groups to obtain the normalized phenotypic characterization values, see Table 5: ; Table 5. Normalized phenotypic characterization values of cells in 10 dishes of five groups

[0091] S 6.4. Calculate the drug effect correction factor.

[0092] According to steps 6.2.2 and 6.3.3, obtain 20 normalized FRET characterization values and phenotypic characterization values of the 5 groups S , and then substitute the 20 normalized FRET characterization values and phenotypic characterization values into the following formula: ; ; Calculate the α + β = 1 of α , β values as the drug effect correction factors for each of the 5 groups. The specific drug effect correction factors are shown in Table 6: ; Table 6. Drug effect correction factors of five groups α, β .

[0093] Note: The drug efficacy correction factor is determined based on the FRET characterization values and phenotypic characterization values measured for each group. There may be a small amount of data error during measurement, and re-measurement can be performed.

[0094] S 6.5. Calculate the drug efficacy score for each drug according to the drug efficacy correction factor.

[0095] Obtain the drug efficacy correction factor for each drug according to Step 6.4 α , β , substitute the α , β values, the FRET characterization values and phenotypic characterization values of each group into the following formula: ; Obtain the drug efficacy parameters of 10 dishes of cells in 5 groups as shown in Table 7: ; Table 7. Drug efficacy parameters of 10 dishes of cells in five groups.

[0096] Next, remove the influence of concentration and time on the drug efficacy parameters according to the following formula. Among them, the drug concentrations of groups 2 and 3 are 1 µm, and the time gradients are 5 h and 11 h; the drug concentration of group 4 is 100 µm, and the time gradients are 1 h and 2 h; the drug concentration of group 5 is 600 mm, and the time gradients are 1 h and 2 h; substitute the data of drug concentration and time gradient into the following formula to calculate the weight factor of drug efficacy parameters related to concentration and time gradient: ; Obtain the weight factors of drug efficacy parameters of 10 dishes of cells in 5 groups as shown in Table 8: ; Table 8. Weight factors of drug efficacy parameters of 10 dishes of cells in five groups.

[0097] Note: No drug is added to the control group, and the drug concentration is 0. Therefore, the corrected weight factor of drug efficacy parameters is 0; group 5 is the NACL drug treatment group, and the drug concentration is 600 mm, which is much higher than other drug treatment groups, resulting in the exponent of the drug efficacy weight factor function being less than 1. The drug efficacy weight factor is a monotonically increasing function related to the independent variable t , which does not conform to the design of the drug efficacy parameter weight factor. Therefore, according to the constraint condition, when , the drug efficacy weight factor is corrected to: ; After correction according to the above formula, obtain the weight factors of drug efficacy parameters of 10 dishes of cells in 5 groups as shown in Table 9: ; Table 9, the weight factors of the pharmacodynamic parameters of 10 dishes of cells in 5 groups after correction.

[0098] Substitute the pharmacodynamic parameters and pharmacodynamic weight factors of 10 dishes of cells in each group into formula (22) to obtain the pharmacodynamic score of each drug. As shown in Table 10: ; Table 10, the pharmacodynamic scores of four drugs.

[0099] Among them, the pharmacodynamic score of A1331852 is 10.0848, which is much greater than the other three drugs. Therefore, A1331852 is the most effective targeted drug for targeting BCL-XL and BAK targets in breast cancer MCF7 cell line.

[0100] Example 3: In this example, a pharmacodynamic evaluation system combining the FRET information of drug target molecules and the bright-field phenotype information of living cells is provided. The system includes a sample preparation module, an imaging and segmentation module, a feature extraction and analysis module, a pharmacodynamic parameter calculation module, and a pharmacodynamic score module; The sample preparation module is used to culture cell samples and transfect the selected drug target molecule FRET plasmid, and add t a time gradient, c a concentration gradient, and n a solution of D n test drugs to obtain n × t × c test samples and t × c reference samples, where n , t, c ≥2 and is an integer; The imaging and segmentation module is used to perform E-FRET three-channel imaging and bright-field imaging on the z th test drug at the i th time gradient and the j th concentration gradient of the test sample and the reference sample at the i th time gradient and the j th concentration gradient C ij Select M fields of view for E-FRET three-channel imaging and bright-field imaging, and perform single-cell segmentation to obtain the three-channel images and bright-field images of each single-cell region in each field of view of the test sample and the reference sample C ij , where M ≥10 and is an integer; The feature extraction and analysis module is used for the sample to be measured and the reference sample C ij to extract regional and texture features from the bright-field images of single-cell regions, and perform PCA principal component analysis to obtain the sample to be measured after normalization and the reference sample C ij the principal component matrix and the principal component vector weight matrix of the bright-field images of each single-cell region in each field of view; calculate the sample to be measured respectively and the reference sample C ij the average FRET efficiency of the three-channel images of each single-cell region in each field of view; respectively according to the sample to be measured and the reference sample C ij the principal component matrix of the bright-field images of each single-cell region in each field of view and the average FRET efficiency in the sample to be measured to calculate the phenotypic characterization value and the FRET characterization value ; The pharmacodynamic parameter calculation module is used to calculate the pharmacodynamic correction factor of the th test drug z and the pharmacodynamic parameter D z of the sample to be measured according to the phenotypic characterization value and the FRET characterization value of the sample to be measured; calculate the pharmacodynamic parameter weight factor according to the pharmacodynamic parameter of the sample to be measured ; The pharmacodynamic scoring module is used to calculate the pharmacodynamic score z th D z test drug according to the pharmacodynamic parameter weight factor, and use the test drug corresponding to the maximum pharmacodynamic score as the most effective targeted drug for this target.

[0101] It should be noted here that the system provided in the above embodiments is only illustrated by the above division of each functional module. In actual applications, the above functions can be assigned to different functional modules according to needs, that is, the internal structure is divided into different functional modules to complete all or part of the functions described above. This system is a pharmacodynamic evaluation method that combines drug target FRET information and bright-field phenotypic information applied to the above embodiments.

[0102] Example 4: In this embodiment, a storage medium is provided, storing a program which, when executed by a processor, implements a method for evaluating drug efficacy by combining FRET information of a drug target and bright-field phenotype information.

[0103] It should be understood that each part of the present application can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiment, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0104] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be equivalent replacement methods and are all included in the protection scope of the present invention.

Claims

1. A method for evaluating drug efficacy by combining drug target FRET information and bright field phenotype information, characterized in that, Including the following steps: Cultivate cell samples and transfect them with the selected FRET plasmids of drug target molecules, add t time gradients, c concentration gradients and n test drugs D n solution to obtain n × t × c test samples and t × c reference samples, where n , t,c ≥2 and is an integer; For the z th test drug, the i th time gradient and the j th concentration gradient of the test sample and the i th time gradient and the j th concentration gradient of the reference sample C ij Select M fields of view for E-FRET three-channel imaging and bright-field imaging, and perform single-cell segmentation to obtain the test sample and the reference sample C ij three-channel images and bright-field images of each single-cell region in each field of view, where M ≥10 and is an integer; The sample to be tested and the reference sample C ij Extract regional and texture features from the bright-field images of single-cell regions, and perform PCA principal component analysis to obtain the normalized sample to be tested and the reference sample C ij The principal component matrix and the principal component vector weight matrix of the bright-field images of each single-cell region under each field of view; Calculate the test sample separately and the reference sample C ij the average FRET efficiency of the three-channel images of each single-cell region under each field of view; According to the sample to be measured and the reference sample C ij calculate the principal component matrix and the average FRET efficiency of the bright-field images of each single-cell region in each field of view, and calculate the phenotypic characterization value and the FRET characterization value of the sample to be measured ; According to the sample to be tested of the phenotypic characterization value and the FRET characterization value calculate the z th drug to be tested D z pharmacodynamic correction factor and the pharmacodynamic parameters of the sample to be tested ; ; According to the sample to be tested pharmacodynamic parameters calculate the weight factor of pharmacodynamic parameters ; According to the weight factor of the pharmacodynamic parameter Calculate the z th drug to be tested D z pharmacodynamic score , and use the drug to be tested corresponding to the maximum pharmacodynamic score as the most effective targeted drug for this target.

2. The pharmacodynamic evaluation method combining drug target FRET information and bright field phenotype information according to claim 1, characterized in that The single-cell segmentation is specifically as follows: The sample to be tested and the reference sample C ij Select M fields of view for E-FRET three-channel imaging and bright-field imaging, and perform Gaussian smoothing, downsampling compression, and gray value compression operations on the generated M three-channel images in each field of view in sequence, and then input them into a pre-trained CNN segmentation model to obtain single-cell segmentation images; upsample and restore the single-cell segmentation images as the mask images mask for the corresponding fields of view; mask the original three-channel images and bright-field images according to the mask images mask for this field of view to obtain the three-channel images and bright-field images of each cell region under the corresponding field of view; The Gaussian smoothing specifically includes: randomly generating a Gaussian kernel with a size of n × n pixels that conforms to a two-dimensional Gaussian distribution, and then convolving the three-channel image with the Gaussian kernel to obtain a three-channel image after Gaussian smoothing; The masking is specifically as follows: the background pixel value of the masking image mask is 0, and the pixel values of each single-cell region are assigned p values according to the number of single-cell regions; the masking image mask and the image to be masked are multiplied by corresponding elements of the matrix to obtain the masked image.

3. The pharmacodynamic evaluation method combining drug target FRET information and bright-field phenotype information according to claim 1, wherein The extraction of the sample to be measured and the reference sample C ij for the bright-field image region and texture features, and perform PCA principal component analysis on the obtained feature vectors to obtain the normalized sample to be measured and the reference sample C ij for the principal component matrix and principal component vector weight matrix of the bright-field image of each single-cell region, specifically: Extract the p bright-field images of the single-cell regions numbered q . Denote the feature vectors as ; and calculate the sample mean vector Center q feature vectors, that is, subtract the corresponding mean from each feature vector, as shown in the following formula: ； ； Calculate the feature matrix of the p th single-cell region after centering processing X p,centered covariance matrix, as follows: ； For the covariance matrix Cov p perform eigenvalue decomposition to obtain the eigenvalues and the corresponding eigenvectors ; select the eigenvectors corresponding to the first k eigenvalues as the principal components to form a projection matrix W p , where each column is an eigenvector; Project the dataset onto the k vector subspace formed by the first Y p = X p,centered * W p ； Obtain the principal component matrix after dimensionality reduction Y p , with size N × k ; Each principal component is expressed as a linear combination of the original features: , where is the p th weight vector corresponding to the i’’ th principal component of the W p represents the contribution ratio of each principal component, sorted from largest to smallest; The principal component matrix and the principal component weight matrix are respectively expressed as and ; Similarly, the principal component matrix of the reference sample is obtained Y 0 and the principal component weight matrix W 0.

4. The pharmacodynamic evaluation method combining drug target FRET information and bright field phenotype information according to claim 1, wherein The calculation of the FRET efficiency is specifically as follows: According to the donor and acceptor fluorescence intensities of the three channels of the three-channel image I DD 、I AA 、I DA , we obtain: F C = I DA - a*I AA - d *I DD ； ； wherein F C is the fluorescence intensity sensitized to the acceptor in the FRET sample, a is the direct excitation correction factor; d is the leakage correction factor; G is the correction factor for different detection efficiencies in the two channels; E D is the FRET efficiency.

5. The pharmacodynamic evaluation method combining drug target FRET information and bright field phenotype information according to claim 1, characterized in that The said according to the sample to be tested and the reference sample C ij For each single-cell region in each field of view, calculate the principal component matrix, principal component vector weight matrix, and average FRET efficiency of the bright-field image, and calculate the phenotypic characterization value of the sample to be tested and the FRET characterization value , specifically: ； ； Wherein, S m-p and FERT m-p are the phenotypic characterization value and the FRET characterization value of the th m single-cell region in the p th field of view of the sample to be measured, m ∈ [1, M and is an integer; is the principal component matrix of the reference sample, is the i’’ th principal component vector of the reference sample; is the principal component matrix of the p th single-cell region of the sample to be measured, is the p th i’’ th principal component vector of the th single-cell region of the sample to be measured; p is the weight factor of the i’’ th principal component vector of the E Dp and E D0 are the average FRET efficiencies of the th single-cell region of the sample to be measured and the reference sample p C ij ij ; Calculate separately S m-1 , S m-2 ,…, S m-p and FERT m-1 , FERT m-2 ,…, FERT m-p The average value of... is used as the average phenotypic characterization value and the average FRET characterization value of the m th visual field, denoted as and ; Calculate separately , The mean value of... is used as the phenotypic characterization value and the FRET characterization value of the sample to be tested .

6. The pharmacodynamic evaluation method combining drug target FRET information and bright-field phenotype information according to claim 1, characterized in that The phenotypic characterization value of the sample to be tested and the FRET characterization value are used to calculate the efficacy correction factor of the nth z drug to be tested D z and the efficacy parameter of the sample to be tested , specifically as follows: ​ Formula One: ; Formula Two: ; Substitute Equation 1 into Equation 2 and calculate when the constraint condition α + β = 1 of the α 、 β value is used as the efficacy correction factor of the drug to be tested D z ; The drug to be tested D z of the pharmacodynamic correction factor α , β Substitute into Equation (1) to obtain the pharmacodynamic parameter of the test sample .

7. The pharmacodynamic evaluation method combining drug target FRET information and bright field phenotype information according to claim 1, characterized in that The pharmacodynamic parameters according to the sample to be measured of calculate the weight factor of the pharmacodynamic parameter , specifically as follows: ； Among them, c i represents the value of the i th concentration gradient, t j represents the value of the j th time gradient.

8. The pharmacodynamic evaluation method combining drug target FRET information and bright field phenotype information according to claim 1, wherein, Said according to the efficacy parameter weight factor Calculate the z kind of drug to be tested D z efficacy score , specifically: ； Among them, represents c i the weight factor of the pharmacodynamic parameter at t j concentration and represents c i the pharmacodynamic parameter at t j concentration and time.

9. A pharmacodynamic evaluation system that combines drug target FRET information and bright field phenotype information, characterized in that, Applied to a drug efficacy evaluation method combining FRET information of a drug target and bright-field phenotype information according to any one of claims 1-8, including a sample preparation module, an imaging and segmentation module, a feature extraction and analysis module, a drug efficacy parameter calculation module, and a drug efficacy scoring module; The sample preparation module is used to culture cell samples, transfect the selected drug target molecule FRET plasmid, and add t time gradients, c concentration gradients, and n test drugs D n solution to obtain n × t × c test samples and t × c reference samples, where n , t,c ≥2 and is an integer; The imaging and segmentation module is used to z for the i th time gradient and the j th concentration gradient of the test sample i and the j th time gradient and the C ij select M fields of view for E-FRET three-channel imaging and bright-field imaging, and perform single-cell segmentation to obtain the test sample C ij and the three-channel images and bright-field images of each single-cell region in each field of view of the reference sample, where M ≥10 and is an integer; The feature extraction and analysis module is used for the sample to be measured and the reference sample C ij to extract the regional and texture features of the bright-field images of single-cell regions, and perform PCA principal component analysis to obtain the sample to be measured after normalization and the reference sample C ij the principal component matrix and the principal component vector weight matrix of the bright-field images of each single-cell region in each field of view; calculate the sample to be measured respectively and the reference sample C ij the average FRET efficiency of the three-channel images of each single-cell region in each field of view; respectively according to the sample to be measured and the reference sample C ij the principal component matrix of the bright-field images of each single-cell region in each field of view and the average FRET efficiency, calculate the sample to be measured of the phenotypic characterization value and the FRET characterization value ; The pharmacodynamic parameter calculation module is used to calculate the pharmacodynamic correction factor of the nth test drug and the pharmacodynamic parameters of the test sample according to the phenotypic characterization value and the FRET characterization value of the test sample; calculate the pharmacodynamic parameter weight factor according to the pharmacodynamic parameters z of the test sample D z ; The pharmacodynamic parameter calculation module is used to calculate the pharmacodynamic correction factor of the nth test drug and the pharmacodynamic parameters of the test sample according to the phenotypic characterization value and the FRET characterization value of the test sample; calculate the pharmacodynamic parameter weight factor according to the pharmacodynamic parameters of the test sample; The pharmacodynamic score module is used to calculate the pharmacodynamic score of the nth z drug to be tested D z according to the pharmacodynamic parameter weight factor , and take the drug to be tested corresponding to the maximum pharmacodynamic score as the most effective targeted drug for this target.

10. A storage medium stores a program, characterized in that: When the program is executed by a processor, it implements a drug efficacy evaluation method combining FRET information of a drug target and bright-field phenotype information according to any one of claims 1-8.

Images