A method and system for evaluating anticancer drug sensitivity based on semi-supervised learning

Abstract

The present invention relates to the technical field of drug sensitivity assessment, and specifically discloses an anticancer drug sensitivity assessment method based on semi-supervised learning, the method comprising: obtaining basic data, and preprocessing the basic data to obtain initial features; constructing an anticancer drug sensitivity assessment model based on a semi-supervised multi-task deep neural network framework and the initial features; based on the anticancer drug sensitivity assessment model, combining Monte Carlo and integral gradient algorithms, obtaining the contribution of the initial features to drug sensitivity. The present invention uses a comprehensive multi-task learning framework to share feature representations between different tasks, optimize neural networks from different angles, thereby improving the accuracy and robustness of the model, and introduces Monte Carlo and integral gradient algorithms to solve the black box problem of deep learning models, provide feature interpretability, and help researchers understand the calculation process of the model, so as to discover the advantages of key features that affect the final results.

Description

Technical Field

[0001] The present invention relates to the technical field of drug sensitivity assessment, and in particular to an anticancer drug sensitivity assessment method and system based on semi-supervised learning. Background Art

[0002] Gynecological cancers are mainly composed of breast cancer (BRCA), ovarian cancer (OV), cervical cancer (CESC), and endometrial cancer (UCEC). However, gynecological cancer is a highly heterogeneous disease, that is, different patients may show very different drug sensitivity and resistance in the same type of tumor. With the rise of artificial intelligence technology, more and more machine learning technologies are used to evaluate the effectiveness of drugs for cancer patients. At the same time, in the biomedical field, the rapid development of high-channel sequencing technology has generated a large amount of gene expression data. This enables machine learning methods to analyze the patient's genetic information, extract individual differences from it, and comprehensively consider the differences between different drugs, thereby establishing an individualized drug sensitivity assessment model, which is expected to predict the sensitivity of each patient to a specific drug based on their specific characteristics, thereby providing decision support for the formulation of personalized clinical treatment plans.

[0003] However, there is still much room for improvement in the various mainstream machine learning methods currently used to evaluate the effectiveness of cancer drugs. The biggest adjustment that affects its prediction accuracy is the high feature dimension: a large amount of high-dimensional feature data will be generated during the experimental data collection process, including gene expression data and drug structural data. The sum of the two often makes the feature dimension reach more than 10,000, which increases the computational complexity. Traditional machine learning methods are not good at dealing with such high-dimensional and complex nonlinear features. In recent years, with the development of bioinformatics technology, although deep learning methods have achieved certain advantages in dealing with such problems, considering the heterogeneity and complexity of gynecological tumors, the accuracy of existing methods is still not satisfactory. In addition, another important challenge is the feature interpretability of deep learning networks. Deep learning models are usually presented in the form of black boxes, which means that it is difficult for researchers to understand the calculation process inside the neural network, explain how the model comes up with a specific prediction result, and find the key features that affect the final result, which poses a serious limitation to medical decision-making and targeted drug development. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide an anticancer drug sensitivity assessment method based on semi-supervised learning, which can share feature representations between different tasks through a comprehensive multi-task learning framework, optimize neural networks from different angles, thereby improving the accuracy and robustness of the model, and introduce Monte Carlo and integral gradient algorithms to solve the black box problem of deep learning models, provide feature interpretability, and help researchers understand the calculation process of the model, so as to discover the advantages of key features that affect the final results.

[0005] In order to solve the above technical problems, the first aspect of the present invention discloses a method for evaluating anticancer drug sensitivity based on semi-supervised learning, the method comprising:

[0006] Acquire basic data, and preprocess the basic data to obtain initial features;

[0007] Based on the semi-supervised multi-task deep neural network framework and the initial features, an anticancer drug sensitivity assessment model is constructed;

[0008] Based on the anticancer drug sensitivity evaluation model, the contribution of the initial features to drug sensitivity is obtained by combining the Monte Carlo and integral gradient algorithms.

[0009] As an optional implementation manner, in the first aspect of the present invention, the step of obtaining basic data and preprocessing the basic data to obtain initial features includes:

[0010] Deleting samples and features that are missing more than 20% of the basic data; the basic data includes patient cancer cell gene expression data, drug structure feature data, and drug effectiveness data;

[0011] The remaining missing items were imputed using the median;

[0012] The gene expression data of the patient's cancer cells are normalized to obtain initial features.

[0013] As an optional embodiment, in the first aspect of the present invention, the anticancer drug sensitivity assessment model is constructed based on the semi-supervised multi-task deep neural network framework and the initial features, comprising:

[0014] Based on a semi-supervised multi-task deep neural network, the initial features are reduced in dimension and decoded to obtain decoding features;

[0015] Performing clustering processing on the decoding features to obtain drug similarity features;

[0016] The decoded features were used for the drug sensitivity prediction task using a binary cross entropy loss function.

[0017] As an optional implementation, in the first aspect of the present invention, the dimension reduction and decoding of the initial features based on the semi-supervised multi-task deep neural network to obtain the decoded features includes:

[0018] Creating a neural network with three fully connected layers, and obtaining reduced-dimensionality features by performing dimensionality reduction processing on the initial features;

[0019] The dimension reduction features are decoded to obtain decoded features.

[0020] As an optional implementation, in the first aspect of the present invention, clustering the decoding features to obtain drug similarity features includes:

[0021] The reduced-dimensional features are clustered in a neural network through the medication similarity loss function, and the reduced-dimensional features are divided into K clusters, so that the samples in each cluster have high similarity, while the features between different clusters are quite different.

[0022] As an optional embodiment, in the first aspect of the present invention, the contribution of the initial features to drug sensitivity is obtained based on the anticancer drug sensitivity assessment model in combination with Monte Carlo and integral gradient algorithms, including:

[0023] Assume that lgx i represents the comprehensive gradient importance of the i-th node, then the uncertainty of the i-th feature is designed as: Where std(*) and ave(*) are standard deviation and mean respectively;

[0024] The importance weight of the initial feature in the network is expressed as: V(x i )=(1-U'(x i ))*IG(x i ),in To use Gauss-legendre quadrature to approximate the integral of the gradient, thereby avoiding the gradient saturation problem;

[0025] Among them, U'(x i ) is U(x i ) is the integrated gradient value after logarithmic transformation and minimum and maximum normalization; thereby obtaining the contribution of the initial features to drug sensitivity.

[0026] The second aspect of the present invention discloses an anticancer drug sensitivity assessment system based on semi-supervised learning, characterized in that the system comprises:

[0027] An initial feature acquisition module is used to acquire basic data and preprocess the basic data to obtain initial features;

[0028] A model building module, used to build an anticancer drug sensitivity assessment model based on a semi-supervised multi-task deep neural network framework and the initial features;

[0029] The contribution acquisition module is used to acquire the contribution of the initial features to the drug sensitivity based on the anticancer drug sensitivity assessment model in combination with the Monte Carlo and integral gradient algorithms.

[0030] The third aspect of the present invention discloses another anticancer drug sensitivity assessment system based on semi-supervised learning, characterized in that the device comprises:

[0031] A memory storing executable program code;

[0032] a processor coupled to the memory;

[0033] The processor calls the executable program code stored in the memory to execute the above-mentioned anticancer drug sensitivity assessment method based on semi-supervised learning.

[0034] A fourth aspect of an embodiment of the present invention discloses a computer storage medium, characterized in that the computer storage medium stores computer instructions, and when the computer instructions are called, they are used to execute the above-mentioned anticancer drug sensitivity assessment method based on semi-supervised learning.

[0035] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: first, by introducing the encoding and decoding loss function and drug similarity information, the learning ability of the neural network to process high-dimensional nonlinear features and distinguish drug differences is improved; second, through a comprehensive multi-task learning framework, feature representations can be shared between different tasks, and the neural network can be optimized from different angles, thereby improving the accuracy and robustness of the model; third, the Monte Carlo and integral gradient algorithms are introduced to solve the black box problem of the deep learning model, provide feature interpretability, and help researchers understand the calculation process of the model, so as to discover the key features that affect the final results, and then discover potential gene targets that can inhibit the development of cancer tumors. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0037] Figure 1 A schematic diagram of a process for evaluating anticancer drug sensitivity based on semi-supervised learning disclosed in an embodiment of the present invention;

[0038] Figure 2 A schematic diagram of the structure of an anticancer drug sensitivity assessment system based on semi-supervised learning disclosed in an embodiment of the present invention;

[0039] Figure 3 A schematic diagram of the structure of an anticancer drug sensitivity assessment device based on semi-supervised learning disclosed in an embodiment of the present invention;

[0040] Figure 4 is the AUC score of the present invention applied to 8 drug data sets;

[0041] Figure 5 is the ACC score of the present invention applied to 8 drug data sets;

[0042] Figure 6 is the F1 score of the present invention applied to 8 drug data sets; DETAILED DESCRIPTION

[0043] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, product or end including a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products or ends.

[0045] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0046] The present invention discloses a method for evaluating the sensitivity of anticancer drugs based on semi-supervised learning, which can share feature representations between different tasks through a comprehensive multi-task learning framework, optimize neural networks from different angles, thereby improving the accuracy and robustness of the model, and introduces Monte Carlo and integral gradient algorithms to solve the black box problem of deep learning models, provide feature interpretability, and help researchers understand the calculation process of the model, so as to discover the advantages of key features that affect the final results. The following are detailed descriptions.

[0047] Embodiment 1

[0048] See also Figure 1 , Figure 1 Schematic diagram of a method for evaluating anticancer drug sensitivity based on semi-supervised learning disclosed in an embodiment of the present invention. Figure 1 As shown, the method may include the following operations:

[0049] 101. Obtain basic data, and preprocess the basic data to obtain initial features.

[0050] Specifically, samples and features with more than 20% missing in the basic data are deleted; the basic data includes patient cancer cell gene expression data, drug structure feature data, and drug effectiveness data; the remaining missing items are filled with the median; the patient cancer cell gene expression data is standardized to obtain initial features.

[0051] 102. Based on the semi-supervised multi-task deep neural network framework and initial features, an anticancer drug sensitivity assessment model is constructed.

[0052] Specifically, based on a semi-supervised multi-task deep neural network, the initial features are reduced in dimension and decoded to obtain decoded features; the decoded features are clustered to obtain drug similarity features; the decoded features are used for the drug sensitivity prediction task, which uses a binary cross entropy loss function.

[0053] Specifically, a three-layer fully connected neural network is created to encode high-dimensional features to obtain their low-dimensional representation Z. The output of the fully connected network is used as a shared feature representation, which is optimized and back-propagated by multiple loss functions (i.e., multi-task modules). The neural network uses Relu as a nonlinear activation function, h l+1 =Relu(W*h l +b), W is the weight matrix and b is the bias.

[0054] The low-dimensional representation Z obtained by the fully connected layer is optimized by the data decoding loss function (i.e., the data encoding and decoding task). The process of constructing a low-dimensional representation of a high-dimensional feature by the fully connected layer is defined as: Z = E(X), where X is the initial high-dimensional feature of the input, and the decoded feature representation is X' = D(Z); the data decoding loss function can be expressed as:

[0055] The decoded features are clustered in the neural network through the medication similarity loss function (i.e., feature unsupervised clustering task), and the samples are divided into K clusters, so that the samples in each cluster are highly similar, while the features between different clusters are quite different. The number of clusters K is determined by the highest silhouette coefficient obtained by K-MEANS calculation of different values ​​of K. In each iteration, the samples are assigned to the closest cluster and the center point of the cluster is updated, and this process is repeated until the cluster center no longer changes significantly. The similarity loss function in the network layer can be expressed as the KL divergence between the two distributions S and T, that is: where t ij Used to measure the cluster center μ j and the embedding point z j The similarity between them is recorded as: s ij The target distribution can be expressed as:

[0056] The decoded features are also used in the drug sensitivity prediction task, which uses a binary cross entropy loss function, denoted as: L P =-[y*log(p)+(1-y)*log(1-p)], where y is the true label of drug effectiveness and p is the predicted value. This task uses the reduced feature Z as input and combines it with the binary cross entropy loss function L for model training. During the training process, the model will minimize the loss function by continuously adjusting parameters so that the predicted result is as close to the true label p as possible.

[0057] According to the above multi-task optimization strategy, the overall loss function of the entire neural network can be expressed as L = γL D +βL C +L P Among them, γ and β are regarded as hyperparameters selected in the cross-validation step to balance the importance of these tasks.

[0058] 103. Based on the anticancer drug sensitivity evaluation model, the Monte Carlo and integral gradient algorithms are combined to obtain the contribution of the initial features to drug sensitivity.

[0059] Specifically, assuming lgx i represents the comprehensive gradient importance of the i-th node, then the uncertainty of the i-th feature is designed as: Where std(*) and ave(*) are standard deviation and mean respectively;

[0060] The importance weight of the initial feature in the network is expressed as: V(x i )=(1-U'(x i ))*IG(x i ),in To use Gauss-legendre quadrature to approximate the integral of the gradient, thereby avoiding the gradient saturation problem;

[0061] Among them, U'(x i ) is U(x i ) is the integrated gradient value after logarithmic transformation and minimum and maximum normalization; thereby obtaining the contribution of gene expression or drug structure in the initial features to drug sensitivity.

[0062] The results of tests on different data sets show that, compared with the existing commonly used methods, the method adopted by the present invention has an AUC improvement of about 3%, an ACC improvement of 3%, and an F1 index improvement of 6.7%.

[0063] In order to evaluate the performance of the anticancer drug effectiveness evaluation method based on semi-supervised multi-task neural network proposed in the present invention, the present invention applies this evaluation method to 8 drug data sets (Docetaxel, Fulvestrant, Uprosertib, Ulixertinib, BRCA, OV, UCEC, CESC), and uses AUC, ACC and F1 scores as evaluation indicators for experimental evaluation. The AUC scores of the effectiveness of the 8 anticancer drugs are as follows: Figure 4 As shown, the ACC score is Figure 5 As shown, the F1 score is Figure 6 As shown. According to the experimental results, the ACC index of these 8 drugs ranged from 0.597 to 0.761, with the highest in Uprosertib and the lowest in UCEC. The ACC index was concentrated between 0.600 and 0.700, indicating that the performance of the method of the present invention was relatively close on different data sets. The F1 score ranged basically between 0.670 and 0.700, with a higher F1 score being achieved in Uprosertib.

[0064] Referring to Tables 1 to 3, the method of the present invention and four other existing commonly used machine learning algorithms SVM, RandomForest, LogistcRegression, and DNN deep neural network were applied to 8 drug data sets for accuracy comparison, and the evaluation index AUC value (Table 1), ACC value (Table 2) and F1 score (Table 3) of each algorithm under each drug data set were calculated for comparison. Experiments show that the AUC obtained by the method of the present invention is between 0.606 (UCEC) and 0.815 (Uprosertib), with an average value of 0.699. Among other methods, Logistic and DNN achieved the highest AUC index of 0.799 (Uprosertib), and DNN achieved the lowest AUC index of 0.505 (Fulvestrant). The average AUC of the comparison method is 0.669. The experimental results show that the method of the present invention is superior to the comparison method and improves the AUC index by 3% on average.

[0065] Table 1 AUC scores of different methods on 8 datasets

[0066]

[0067] Table 2 ACC scores of different methods on 8 datasets

[0068]

[0069] Table 3 F1 scores of different methods on 8 datasets

[0070]

[0071]

[0072] Embodiment 2

[0073] See also Figure 2 , Figure 2 Schematic diagram of the structure of an anticancer drug sensitivity assessment system based on semi-supervised learning disclosed in an embodiment of the present invention. Figure 2 As shown, the system may include:

[0074] The initial feature acquisition module 100 is used to acquire basic data and pre-process the basic data to obtain initial features;

[0075] A model building module 200 is used to build an anticancer drug sensitivity assessment model based on a semi-supervised multi-task deep neural network framework and initial features;

[0076] The contribution acquisition module 300 is used to acquire the contribution of the initial features to the drug sensitivity based on the anticancer drug sensitivity evaluation model in combination with the Monte Carlo and integral gradient algorithms.

[0077] Embodiment 3

[0078] See also Figure 3 , Figure 3 FIG. 1 is a schematic diagram of another anticancer drug sensitivity assessment device based on semi-supervised learning disclosed in an embodiment of the present invention. Figure 3 As shown, the device may include:

[0079] A memory 301 storing executable program codes;

[0080] a processor 302 coupled to the memory 301;

[0081] The processor 302 calls the executable program code stored in the memory 301 to execute part or all of the steps in the anticancer drug sensitivity assessment method based on semi-supervised learning disclosed in the first or second embodiment of the present invention.

[0082] Embodiment 4

[0083] An embodiment of the present invention discloses a computer storage medium, which stores computer instructions. When the computer instructions are called, they are used to execute some or all steps in the anticancer drug sensitivity assessment method based on semi-supervised learning disclosed in Example 1 of the present invention.

[0084] The system embodiment described above is only illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without paying creative labor.

[0085] Through the specific description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the above technical solution can be essentially or partly contributed to the prior art in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0086] Finally, it should be noted that the anticancer drug sensitivity assessment method and system based on semi-supervised learning disclosed in the embodiment of the present invention discloses only the preferred embodiment of the present invention, which is only used to illustrate the technical scheme of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical schemes described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical schemes from the spirit and scope of the technical schemes of the various embodiments of the present invention.

Claims

1. A method for evaluating anticancer drug sensitivity based on semi-supervised learning, characterized in that: The method includes: Acquire basic data, and preprocess the basic data to obtain initial features; Based on the semi-supervised multi-task deep neural network framework and the initial features, an anticancer drug sensitivity assessment model is constructed; Based on the anticancer drug sensitivity evaluation model, the contribution of the initial features to drug sensitivity is obtained by combining Monte Carlo and integral gradient algorithms; Wherein, the anticancer drug sensitivity assessment model is constructed based on the semi-supervised multi-task deep neural network framework and the initial features, including: Based on a semi-supervised multi-task deep neural network, the initial features are reduced in dimension and decoded to obtain decoding features; Performing clustering processing on the decoding features to obtain drug similarity features; Using the decoded features for a drug sensitivity prediction task using a binary cross entropy loss function; The step of reducing the dimension and decoding the initial features based on a semi-supervised multi-task deep neural network to obtain decoded features includes: Creating a neural network with three fully connected layers, and obtaining reduced-dimensionality features by performing dimensionality reduction processing on the initial features; Decoding the dimension reduction features to obtain decoding features; The clustering process of the decoded features to obtain drug similarity features includes: The reduced-dimensional features are clustered in a neural network through a medication similarity loss function, and the reduced-dimensional features are divided into K clusters, so that the samples in each cluster have a high similarity, while the features between different clusters have a large difference; The method of obtaining the contribution of the initial features to drug sensitivity based on the anticancer drug sensitivity evaluation model and combining Monte Carlo and integral gradient algorithms includes: Assumptions Represents the comprehensive gradient importance of the i-th node, then the uncertainty of the i-th feature is designed as: , where std(*) and ave(*) are the standard deviation and mean respectively; The importance weight of the initial feature in the network is expressed as: ,in To use Gauss-legendre quadrature to approximate the integral of the gradient, thereby avoiding the gradient saturation problem; in, yes The integrated gradient value after logarithmic transformation and minimum and maximum normalization; thus, the contribution of the initial feature to drug sensitivity is obtained.

2. The anticancer drug sensitivity assessment method based on semi-supervised learning according to claim 1, characterized in that: The acquiring of basic data and preprocessing of the basic data to obtain initial features includes: Delete samples and features that are missing more than 20% in the basic data; the basic data includes patient cancer cell gene expression data, drug structure feature data, and drug effectiveness data; The remaining missing items were imputed using the median; The gene expression data of the patient's cancer cells are normalized to obtain initial features.

3. An anticancer drug sensitivity assessment system based on semi-supervised learning, characterized in that: The system comprises: An initial feature acquisition module is used to acquire basic data and preprocess the basic data to obtain initial features; A model building module, used to build an anticancer drug sensitivity assessment model based on a semi-supervised multi-task deep neural network framework and the initial features; A contribution acquisition module, used to acquire the contribution of the initial features to drug sensitivity based on the anticancer drug sensitivity assessment model in combination with Monte Carlo and integral gradient algorithms; Wherein, the anticancer drug sensitivity assessment model is constructed based on the semi-supervised multi-task deep neural network framework and the initial features, including: Based on a semi-supervised multi-task deep neural network, the initial features are reduced in dimension and decoded to obtain decoding features; Performing clustering processing on the decoding features to obtain drug similarity features; Using the decoded features for a drug sensitivity prediction task using a binary cross entropy loss function; The step of reducing the dimension and decoding the initial features based on a semi-supervised multi-task deep neural network to obtain decoded features includes: Creating a neural network with three fully connected layers, and obtaining reduced-dimensionality features by performing dimensionality reduction processing on the initial features; Decoding the dimension reduction features to obtain decoding features; The clustering process of the decoded features to obtain drug similarity features includes: The reduced-dimensional features are clustered in a neural network through a medication similarity loss function, and the reduced-dimensional features are divided into K clusters, so that the samples in each cluster have a high similarity, while the features between different clusters have a large difference; The method of obtaining the contribution of the initial features to drug sensitivity based on the anticancer drug sensitivity evaluation model and combining Monte Carlo and integral gradient algorithms includes: Assumptions represents the comprehensive gradient importance of the i-th node, then the uncertainty of the i-th feature is designed as: , where std(*) and ave(*) are the standard deviation and mean respectively; The importance weight of the initial feature in the network is expressed as: ,in To use Gauss-legendre quadrature to approximate the integral of the gradient, thereby avoiding the gradient saturation problem; in, yes The integrated gradient value after logarithmic transformation and minimum and maximum normalization; thus, the contribution of the initial feature to drug sensitivity is obtained.

4. An anticancer drug sensitivity assessment system based on semi-supervised learning, characterized in that: The system comprises: A memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the anticancer drug sensitivity assessment method based on semi-supervised learning as described in any one of claims 1-2.

5. A computer storage medium, characterized in that: The computer storage medium stores computer instructions, which, when called, are used to execute the anticancer drug sensitivity assessment method based on semi-supervised learning as described in any one of claims 1-2.

Images