Drug sensitivity prediction method and device, computer equipment and readable storage medium

Abstract

The invention relates to a drug sensitivity prediction method and device, computer equipment and a readable storage medium. The method comprises the following steps: determining a cross-dimension core pathway set jointly regulated and controlled by a target drug in a plurality of tissue samples; according to the incidence relation between the multiple pairs of cross-dimensional core pathways, constructing a cross-dimensional pathway network for co-regulation and control of the target drug in the multiple tissue samples; in the cross-dimensional pathway network, generating a plurality of pairs of initial pathway groups with interpretability according to regulation and control association characteristics of the target drug to each cross-dimensional core pathway; setting a first to-be-disturbed path group on the basis of the plurality of pairs of initial path groups, and performing probabilistic cascade iterative disturbance on the path response network under the action of the target drug on the basis of the path characteristic data of the first to-be-disturbed path group until a disturbance change value of the target drug under the path response network is obtained; and performing sensitivity prediction on the target drug according to the disturbance change value. By adopting the method, the interpretability of drug sensitivity prediction is improved.

Description

Technical Field

[0001] The present application relates to the interdisciplinary technical field of bioinformatics and medical diagnosis assistance, and in particular to a method, apparatus, computer device, and computer-readable storage medium for predicting drug sensitivity. Background Art

[0002] With the continuous development of bioinformatics, drug sensitivity prediction has become increasingly important. DSP (Drug Sensitivity Prediction) refers to the quantitative assessment of the responsiveness of individual cells, tissues, and patients to specific drugs by integrating multi-omics data (genome, transcriptome, proteome, etc.) with machine learning algorithms. By quantitatively assessing the differences in individual or cellular responses to drugs, it can play an important role in optimizing treatment decisions, reducing drug resistance, and accelerating the development of new drugs.

[0003] Currently, drug sensitivity prediction is usually performed by constructing a DNN (Deep Neural Network) model. Precily is a highly representative DNN model. It uses multimodal fusion of GSVA (Gene Set Variation Analysis) and SMILES (Simplified Molecular Input Line Entry System) to construct a DNN model to predict IC50 (50% Inhibitory Concentration) values. This can alleviate the batch effect of gene expression data and support new drug prediction. However, due to the complex synergistic and antagonistic relationships between pathways, it is difficult to quantify the specific contribution of pathways to the prediction results of the DNN model, which in turn makes it prone to clinical and scientific research risks. Therefore, the current interpretability of drug sensitivity prediction is low. Summary of the Invention

[0004] Based on this, it is necessary to provide a drug sensitivity prediction method, apparatus, computer device and computer-readable storage medium to improve the interpretability of drug sensitivity prediction in order to address the above technical problems.

[0005] In a first aspect, the present application provides a method for predicting drug sensitivity, comprising:

[0006] Determining a cross-dimensional core pathway set that is co-regulated by a target drug in multiple tissue samples, wherein the dimension spanned by the cross-dimensional core pathway set includes at least one of a sample dimension and a drug dimension, and the cross-dimensional core pathway set includes multiple pairs of cross-dimensional core pathways;

[0007] Constructing a cross-dimensional pathway network co-regulated by the target drug in the multiple tissue samples based on the association relationships between the multiple pairs of cross-dimensional core pathways;

[0008] In the cross-dimensional pathway network, multiple pairs of initial pathway groups with interpretability are generated based on the regulatory association characteristics of the target drug on each cross-dimensional core pathway;

[0009] A first pathway group to be disturbed is set based on the multiple pairs of initial pathway groups, and based on the pathway characteristic data of the first pathway group to be disturbed, a probabilistic cascade iterative perturbation is performed on the pathway response network under the action of the target drug until a perturbation change value of the target drug under the pathway response network is obtained;

[0010] The sensitivity of the target drug is predicted according to the disturbance change value.

[0011] In one embodiment, determining a cross-dimensional core pathway set that is co-regulated by a target drug in multiple tissue samples comprises:

[0012] predicting a drug sensitivity baseline value of the target drug in the multiple tissue samples based on the drug characteristic information of the target drug and the sample characteristic information of the multiple tissue samples;

[0013] Setting a second group of pathways to be disturbed that are commonly regulated in multiple tissue samples for the target drug, wherein the second group of pathways to be disturbed includes multiple monomer core pathways;

[0014] According to the drug sensitivity benchmark value, performing perturbation analysis on the multiple monomer core pathways respectively using a preset drug sensitivity prediction model to obtain multiple drug sensitivity prediction values;

[0015] Among the multiple monomer core pathways, the cross-dimensional core pathway set is constructed according to the multiple drug sensitivity prediction values.

[0016] In one embodiment, the cross-dimensional core pathway set is constructed based on the multiple drug sensitivity prediction values ​​in the multiple monomer core pathways, including:

[0017] Determining a mean pathway contribution weight corresponding to the preset drug sensitivity prediction model and a pathway contribution weight value corresponding to each of the multiple monomer core pathways;

[0018] Selecting a plurality of first monomer core pathways from all monomer core pathways according to a relationship between the pathway contribution weight mean and the plurality of pathway contribution weights, and constructing a first monomer core pathway set based on the plurality of first monomer core pathways;

[0019] selecting a plurality of second monomer core pathways from all the monomer core pathways according to the plurality of drug sensitivity prediction values, and constructing a second monomer core pathway set based on the plurality of second monomer core pathways;

[0020] The cross-dimensional core pathway set is constructed based on the first monomer core pathway set and the second monomer core pathway set.

[0021] In one embodiment, constructing the cross-dimensional core pathway set based on the first monomer core pathway set and the second monomer core pathway set includes:

[0022] Determine a plurality of candidate monomer core pathways having intersections between the first monomer core pathway set and the second monomer core pathway set;

[0023] Normalizing the drug sensitivity prediction values ​​and pathway contribution weight values ​​of the multiple candidate monomer core pathways to obtain prediction standard values ​​and weight standard values ​​corresponding to each of the multiple candidate monomer core pathways;

[0024] By fusing the predicted standard value and weight standard value of each candidate monomer core pathway, the pathway importance of each candidate monomer core pathway is obtained;

[0025] According to the importance of multiple pathways, multiple target monomer core pathways are selected from all candidate monomer core pathways, and the multiple target monomer core pathways are integrated into the cross-dimensional core pathway set.

[0026] In one embodiment, the association relationship includes association similarity; and constructing a cross-dimensional pathway network co-regulated by the target drug in the multiple tissue samples based on the association relationship between the multiple pairs of cross-dimensional core pathways includes:

[0027] Determining, based on a magnitude relationship between the correlation similarity and a preset correlation similarity threshold, a plurality of pairs of cross-dimensional core pathways having a correlation relationship among the plurality of pairs of cross-dimensional core pathways;

[0028] By constructing cross-dimensional regulatory edges of the multiple pairs of cross-dimensional core pathways and setting edge weights for the cross-dimensional regulatory edges, a cross-dimensional pathway network that is co-regulated by the target drug in the multiple tissue samples is constructed.

[0029] In one embodiment, the regulatory association feature includes betweenness centrality; in the cross-dimensional pathway network, based on the regulatory association feature of the target drug on each cross-dimensional core pathway, multiple pairs of interpretable initial pathway groups are generated, including:

[0030] In the cross-dimensional pathway network, all cross-dimensional core pathways are divided into multiple pairs of first cross-dimensional core pathways and multiple pairs of second cross-dimensional core pathways according to the betweenness centrality of the target drug to each cross-dimensional core pathway, wherein the betweenness centralities of the multiple pairs of first cross-dimensional core pathways are all greater than a preset betweenness centrality threshold, and the betweenness centralities of the multiple pairs of second cross-dimensional core pathways are all less than the preset betweenness centrality threshold;

[0031] Selecting a first target inter-dimensional core path from all first inter-dimensional core paths according to the edge weights of the multiple pairs of first inter-dimensional core paths;

[0032] Selecting a second target inter-dimensional core pathway from all second inter-dimensional core pathways according to the edge weights of the multiple pairs of second inter-dimensional core pathways;

[0033] Selecting a third target interdimensional core pathway from all of the first interdimensional core pathways and all of the second interdimensional core pathways;

[0034] The first target cross-dimensional core pathway, the second target cross-dimensional core pathway and the third target cross-dimensional core pathway are used to generate the multiple pairs of initial pathway groups.

[0035] In one embodiment, after predicting the sensitivity of the target drug according to the disturbance change value, the method further includes:

[0036] Obtaining a pathway network characteristic value of the cross-dimensional pathway network;

[0037] Generating cascade regulation breadth information and perturbation effect intensity information corresponding to the target drug according to the pathway network characteristic value and the perturbation change value;

[0038] The cascade control breadth information and the disturbance effect intensity information are displayed on a preset display interface.

[0039] In a second aspect, the present application also provides a drug sensitivity prediction device, comprising:

[0040] a determination module, configured to determine a set of cross-dimensional core pathways co-regulated by a target drug in multiple tissue samples, wherein the dimensions spanned by the cross-dimensional core pathway set include at least one of a sample dimension and a drug dimension, and the cross-dimensional core pathway set includes multiple pairs of cross-dimensional core pathways;

[0041] A construction module, configured to construct a cross-dimensional pathway network co-regulated by the target drug in the multiple tissue samples based on the association relationships between the multiple pairs of cross-dimensional core pathways;

[0042] A generation module is used to generate a plurality of pairs of initial pathway groups with interpretability in the cross-dimensional pathway network according to the regulatory association characteristics of the target drug on each cross-dimensional core pathway;

[0043] a perturbation module, configured to set a first pathway group to be perturbed based on the multiple pairs of initial pathway groups, and perform a probabilistic cascade iterative perturbation on the pathway response network under the action of the target drug based on the pathway characteristic data of the first pathway group to be perturbed, until a perturbation change value of the target drug under the pathway response network is obtained;

[0044] The prediction module is used to predict the sensitivity of the target drug according to the disturbance change value.

[0045] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0046] Determine a set of cross-dimensional core pathways that are jointly regulated by a target drug in multiple tissue samples, wherein the dimensions spanned by the cross-dimensional core pathway set include at least one of the sample dimension and the drug dimension, and the cross-dimensional core pathway set includes multiple pairs of cross-dimensional core pathways; construct a cross-dimensional pathway network that is jointly regulated by the target drug in the multiple tissue samples based on the association relationship between the multiple pairs of cross-dimensional core pathways; in the cross-dimensional pathway network, generate multiple pairs of initial pathway groups with interpretability based on the regulatory association characteristics of the target drug on each cross-dimensional core pathway; set a first pathway group to be disturbed based on the multiple pairs of initial pathway groups, and perform probabilistic cascade iterative perturbation on the pathway response network under the action of the target drug based on the pathway feature data of the first pathway group to be disturbed, until a disturbance change value of the target drug under the pathway response network is obtained; and perform sensitivity prediction on the target drug based on the disturbance change value.

[0047] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0048] Determine a set of cross-dimensional core pathways that are jointly regulated by a target drug in multiple tissue samples, wherein the dimensions spanned by the cross-dimensional core pathway set include at least one of the sample dimension and the drug dimension, and the cross-dimensional core pathway set includes multiple pairs of cross-dimensional core pathways; construct a cross-dimensional pathway network that is jointly regulated by the target drug in the multiple tissue samples based on the association relationship between the multiple pairs of cross-dimensional core pathways; in the cross-dimensional pathway network, generate multiple pairs of initial pathway groups with interpretability based on the regulatory association characteristics of the target drug on each cross-dimensional core pathway; set a first pathway group to be disturbed based on the multiple pairs of initial pathway groups, and perform probabilistic cascade iterative perturbation on the pathway response network under the action of the target drug based on the pathway feature data of the first pathway group to be disturbed, until a disturbance change value of the target drug under the pathway response network is obtained; and perform sensitivity prediction on the target drug based on the disturbance change value.

[0049] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the following steps:

[0050] Determine a set of cross-dimensional core pathways that are jointly regulated by a target drug in multiple tissue samples, wherein the dimensions spanned by the cross-dimensional core pathway set include at least one of the sample dimension and the drug dimension, and the cross-dimensional core pathway set includes multiple pairs of cross-dimensional core pathways; construct a cross-dimensional pathway network that is jointly regulated by the target drug in the multiple tissue samples based on the association relationship between the multiple pairs of cross-dimensional core pathways; in the cross-dimensional pathway network, generate multiple pairs of initial pathway groups with interpretability based on the regulatory association characteristics of the target drug on each cross-dimensional core pathway; set a first pathway group to be disturbed based on the multiple pairs of initial pathway groups, and perform probabilistic cascade iterative perturbation on the pathway response network under the action of the target drug based on the pathway feature data of the first pathway group to be disturbed, until a disturbance change value of the target drug under the pathway response network is obtained; and perform sensitivity prediction on the target drug based on the disturbance change value.

[0051] The above-mentioned drug sensitivity prediction method, device, computer equipment and computer-readable storage medium first identify the cross-dimensional core pathway set that is jointly regulated by the target drug in different tissue samples, wherein the dimensions spanned by the cross-dimensional core pathway set include at least one of the sample dimension and the drug dimension, and the cross-dimensional pathway set includes multiple pairs of cross-dimensional core pathways, so that multiple pairs of cross-dimensional core pathways under the joint constraints of drugs and tissue samples can be determined; then, through the association relationship between multiple pairs of cross-dimensional core pathways, a cross-dimensional pathway network that is jointly regulated by the target drug in multiple tissue samples is constructed, and the regulatory dependency relationship and drug response topology between different pathways can be displayed through the cross-dimensional pathway network; then, in the cross-dimensional pathway network, according to the regulatory association characteristics of the target drug on each cross-dimensional core pathway, multiple pairs of initial pathway groups with biological significance are generated; then, based on the multiple pairs of initial pathway groups, a first pathway group to be disturbed is set, and based on the pathway feature data of the first pathway group to be disturbed, the target drug is disturbed. The pathway response network under the action of the target drug is subjected to probabilistic cascade iterative perturbation until the perturbation change value of the target drug under the pathway response network is obtained, thereby realizing the dynamic impact of the simulated pathway perturbation on the sensitivity of the target drug and quantifying the specific contribution of the pathway. Finally, the sensitivity of the target drug is predicted by the perturbation change value. Since in the process of sensitivity to the target drug, the regulatory dependency relationship between pathways and the cascade effect of the pathway network under the action of simulated drugs can be portrayed through cross-dimensional pathway network modeling and probabilistic cascade perturbation analysis, the purpose of converting the traditional black box prediction method into a traceable and verifiable prediction process is achieved, rather than only outputting quantitative prediction results. Therefore, the technical defect that it is difficult to quantify the specific contribution of the pathway to the prediction results of the deep neural network model due to the complex synergistic and antagonistic relationship between pathways, which makes it easy to cause clinical and scientific research risks, is overcome. Therefore, the interpretability of drug sensitivity prediction is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0053] Figure 1 FIG1 is a flow chart of a method for predicting drug sensitivity in one embodiment;

[0054] Figure 2 FIG1 is a flow chart of a method for predicting drug sensitivity in another embodiment;

[0055] Figure 3This is a schematic diagram of the overall process of predicting drug sensitivity in a method for predicting drug sensitivity in another embodiment;

[0056] Figure 4 is a structural block diagram of a drug sensitivity prediction device in one embodiment;

[0057] Figure 5 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0059] First, with the increasing incidence of cancer, the limitations of traditional "one-size-fits-all" treatments are becoming increasingly apparent. Studies have shown that patients with histologically similar tumors can respond significantly differently to the same chemotherapy regimen, partly due to tumor genetic heterogeneity. For example, HER2-positive and BRCA-mutated breast cancers require different targeted therapies. Therefore, customized medical plans based on molecular diagnostics (such as genomic and proteomic analysis) have become mainstream. The core of this approach is to identify predictive biomarkers such as HER2 and EGFR to screen for drug-sensitive patients, thereby reducing ineffective treatment and side effects. On the other hand, breakthroughs in high-throughput technology and molecular medicine have also provided a data foundation for drug sensitivity research. For example, projects such as the Cancer Genome Atlas and the Cancer Cell Line Encyclopedia integrate genomic and pharmacological data from thousands of cell lines. At the same time, large-scale in vitro drug screening has revealed differential response patterns among different cancer types and subtypes, and technologies such as single-cell mass accumulation rate can also assess drug sensitivity in real time. Furthermore, deep learning models use this data to predict drug responses, breaking through the limitations of traditional statistics.

[0060] At present, the mainstream way to predict drug sensitivity is to complete the prediction by constructing a deep neural network model. However, although the interpretability of the deep neural network model can be indirectly obtained by analyzing the relationship between the model prediction results and the known drug mechanism of action, due to the complex synergistic and antagonistic relationships between pathways and the black box characteristics of the deep neural network model, it is impossible to quantify the specific contribution of the pathway to the prediction results, and thus it is difficult to meet the clinical requirements for transparency of decision-making basis. In addition, XGraphCDS attempts to explain the model prediction based on the statistical method gradient integral, but it focuses on the marginal contribution of a single feature to the model output and still cannot explain the complex relationship between pathways. Therefore, the traditional technology of predicting drug sensitivity based on deep neural network models always has the problem of insufficient interpretability, which makes clinical conversion difficult and prone to clinical and scientific research risks. Therefore, there is an urgent need for a drug sensitivity prediction method that improves the interpretability of drug sensitivity prediction.

[0061] In one embodiment, Figure 1As shown, a drug sensitivity prediction method is provided. This embodiment takes the method applied to a drug sensitivity prediction system as an example. The drug sensitivity prediction system is deployed on a terminal, and the terminal includes but is not limited to a personal computer and a laptop computer. The drug sensitivity prediction system includes a determination module, a construction module, a generation module, a perturbation module and a prediction module, wherein the determination module is used to determine the cross-dimensional core pathway set that is jointly regulated by the target drug in multiple tissue samples, wherein the dimension spanned by the cross-dimensional core pathway set includes at least one of the sample dimension and the drug dimension, and the cross-dimensional core pathway set includes multiple pairs of cross-dimensional core pathways; the construction module is used to construct a cross-dimensional pathway network that is jointly regulated by the target drug in multiple tissue samples according to the correlation relationship between the multiple pairs of cross-dimensional core pathways; the generation module is used to generate multiple pairs of initial pathway groups with interpretability in the cross-dimensional pathway network according to the regulatory correlation characteristics of the target drug on each cross-dimensional core pathway; the perturbation module is used to generate multiple pairs of initial pathway groups with interpretability based on the regulatory correlation characteristics of the target drug on each cross-dimensional core pathway. A first pathway group to be disturbed is set in multiple pairs of initial pathway groups, and based on the pathway characteristic data of the first pathway group to be disturbed, a probabilistic cascade iterative perturbation is performed on the pathway response network under the action of the target drug until the perturbation change value of the target drug under the pathway response network is obtained; the prediction module is used to predict the sensitivity of the target drug according to the perturbation change value; and then by determining the information interaction between the module, the construction module, the generation module, the perturbation module and the prediction module, the regulatory dependency between pathways and the cascade effect of the pathway network under the action of the simulated drug can be portrayed through cross-dimensional pathway network modeling and probabilistic cascade perturbation analysis, so that the purpose of converting the traditional black box prediction method into a traceable and verifiable prediction process is achieved. Therefore, the effect of improving the interpretability of drug sensitivity prediction can be achieved. It can be understood that the drug sensitivity prediction system can also be deployed on a server, and can also be deployed on a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0062] Step 102: determine a cross-dimensional core pathway set that is commonly regulated by the target drug in multiple tissue samples, wherein the dimensions spanned by the cross-dimensional core pathway set include at least one of the sample dimension and the drug dimension, and the cross-dimensional core pathway set includes multiple pairs of cross-dimensional core pathways.

[0063] It should be noted that the cross-dimensional core pathway set refers to a core pathway set that spans at least one of the sample dimension and the drug dimension. Specifically, it can be a cross-dimensional core pathway set that is co-regulated by the same drug in different tissue samples, a cross-dimensional core pathway set that is co-regulated by different drugs in the same tissue sample, and a cross-dimensional core pathway set that is co-regulated by different drugs in different tissue samples. The sample dimension refers to a dimension constructed based on tissue samples as the distinguishing object. It is understandable that some pathways will show consistent regulation patterns in multiple tissue samples. For example, in tumor tissues of multiple different objects, PI3K- The AKT pathway is in an activated state, affecting cell growth and survival. The drug dimension refers to a dimension constructed with drugs as the distinguishing objects. It is understandable that in the drug dimension, different drugs have different effects on the same pathway. For example, for the PARP pathway, platinum drugs can inhibit it and thus exert an anti-cancer effect, while TKI drugs have no obvious effect; target drugs refer to drugs for sensitivity prediction, which can be one or more. For example, target drugs include small molecule targeted drugs, traditional chemical synthetic drugs and recombinant protein drugs; tissue samples refer to a collection of cell populations with specific structures and functions obtained from an organism, which can be one or more. For example, tissue samples include lung cancer tissue samples and liver cancer tissue samples; in some feasible embodiments, in tumor drug research, the pathway pair consisting of the PI3K-AKT pathway and the mTOR pathway often jointly affect the sensitivity of tumor cells to drugs, so the PI3K-AKT pathway and the mTOR pathway can be called a cross-dimensional core pathway set.

[0064] It should be noted that the cross-dimensional core pathway set includes multiple pairs of cross-dimensional core pathways, where each pair of cross-dimensional core pathways can be obtained by querying in an external pathway database. For example, in one feasible method, assuming that the external pathway database is MSigDB, multiple pairs of cross-dimensional core pathways can be queried in MSigDB, and a cross-dimensional core pathway set that is jointly regulated by the target sample in multiple tissue samples is constructed by multiple pairs of cross-dimensional core pathways.

[0065] As an example, step 102 includes: querying a preset pathway database to obtain multiple pairs of cross-dimensional core pathways that are co-regulated by the target drug in multiple tissue samples, and constructing a cross-dimensional core pathway set based on the multiple pairs of cross-dimensional core pathways.

[0066] Step 104 : constructing a cross-dimensional pathway network that is co-regulated by the target drug in multiple tissue samples based on the association relationships between multiple pairs of cross-dimensional core pathways.

[0067] It should be noted that the correlation between multiple pairs of cross-dimensional core pathways can be set in advance, specifically referring to the correlation between each pair of cross-dimensional core pathways and any other pair of cross-dimensional core pathways; the cross-dimensional pathway network is used to characterize the mutual connection and mutual regulation relationship between multiple pairs of cross-dimensional core pathways; it can be understood that the pathway refers to a series of interrelated and orderly biochemical reactions or signal transmission processes in the cell. For example, the MAPK pathway plays a key role in cell proliferation, differentiation and apoptosis, and is responsible for accurately transmitting extracellular signals to the cell nucleus and regulating gene expression; in some feasible implementation methods, the degree of correlation between multiple pairs of cross-dimensional core pathways can be quantified, and after quantification, multiple pairs of cross-dimensional core pathways can be connected to construct a cross-dimensional pathway network, wherein the construction of the cross-dimensional pathway network can specifically rely on mathematical and computational methods such as graph theory or network analysis.

[0068] As an example, step 104 includes: by analyzing the association relationship between multiple pairs of cross-dimensional core pathways, constructing a cross-dimensional pathway network that is co-regulated by the target drug in multiple tissue samples on a preset map.

[0069] Step 106: In the cross-dimensional pathway network, based on the regulatory association characteristics of the target drug on each cross-dimensional core pathway, generate multiple pairs of initial pathway groups with interpretability.

[0070] It should be noted that the regulatory association features are used to characterize the key characteristics of the target drug's regulatory effect on the cross-dimensional core pathway, such as degree centrality, proximity centrality, or eigenvector centrality. Among them, degree centrality is used to characterize the number of other cross-dimensional core pathways directly connected to each cross-dimensional core pathway, proximity centrality is used to characterize the average shortest path length from each cross-dimensional core pathway to all other cross-dimensional core pathways in the cross-dimensional pathway network, and eigenvector centrality is used to characterize the importance of each cross-dimensional core pathway relative to any other cross-dimensional core pathway in the cross-dimensional pathway network. By measuring the size relationship between the regulatory association features and the preset regulatory association feature threshold, multiple initial cross-dimensional core pathways with biological significance can be screened out in the cross-dimensional pathway network, thereby using multiple initial cross-dimensional core pathways to complete the construction of multiple pairs of initial pathway groups with interpretability. The initial pathway group refers to a collection of cross-dimensional core pathways with similar regulatory association features or functionally related functions. It can be understood that multiple pairs of initial pathway groups can be used as multiple functional modules under the regulatory effect of the target drug. The initial cross-dimensional core pathways in any functional module can work together during the drug action process to jointly achieve specific biological functions or regulatory goals.

[0071] As an example, step 106 includes: screening all cross-dimensional core pathways in the cross-dimensional pathway network based on the size relationship between the regulatory association characteristics of each cross-dimensional core pathway and a preset regulatory association characteristic threshold to obtain multiple pairs of initial cross-dimensional core pathways, and generating multiple pairs of initial pathway groups with interpretability based on the multiple pairs of initial cross-dimensional core pathways.

[0072] Step 108 : setting a first pathway group to be disturbed based on multiple pairs of initial pathway groups, and performing a probabilistic cascade iterative perturbation on the pathway response network under the action of the target drug based on the pathway characteristic data of the first pathway group to be disturbed, until a perturbation change value of the target drug under the pathway response network is obtained.

[0073] It should be noted that the first pathway group to be disturbed refers to a specific pathway group set based on multiple pairs of initial pathway groups for conducting simulated perturbation experiments. It can be understood that the first pathway group to be disturbed can be either a subset of multiple pairs of initial pathway groups or a pathway group generated based on multiple pairs of initial pathway groups according to preset generation rules; pathway characteristic data is used to characterize the data indicators presented by the pathway based on its own characteristics. For example, in one feasible method, the pathway characteristic data can be a GSVA enrichment score matrix. Specifically, the GSVA enrichment score matrix includes pathway name, tissue sample name, and pathway GSVA enrichment score, etc.; probabilistic cascade iterative perturbation is a computational method for simulating the dynamic changes of pathways in biological systems. Specifically, the first pathway group to be disturbed can be initially disturbed based on a probability model, and Then, based on the interaction relationship between each pathway in the first pathway group to be disturbed and the preset probability rules, the disturbance effect is cascaded to the entire pathway response network, and the dynamic response process of the cross-dimensional pathway network is simulated through a preset number of iterative calculations; the disturbance change value refers to a quantitative indicator of the state difference between the cross-dimensional pathway network before and after the disturbance under the action of the target drug. It can be understood that if the disturbance value is large, it means that the cross-dimensional pathway network is highly sensitive to the disturbance of the first pathway group to be disturbed. Accordingly, the target drug can significantly regulate the cross-dimensional pathway network through the first pathway group to be disturbed. Conversely, if the disturbance value is small, it means that the cross-dimensional pathway network is not sensitive to the disturbance of the first pathway group to be disturbed. Accordingly, the target drug will not significantly regulate the cross-dimensional pathway network through the first pathway group to be disturbed.

[0074] As an example, step 108 includes: selecting a first pathway group to be disturbed from multiple pairs of initial pathway groups, and based on the pathway characteristic data of the first pathway group to be disturbed, performing a probabilistic cascade iterative perturbation on the pathway response network under the action of the target drug until a perturbation change value of the target drug under the pathway response network is obtained.

[0075] In one practicable manner, the first path group to be disturbed, which is set based on multiple pairs of initial path groups, may be as follows:

[0076] a) (+50%, +50%): ;

[0077] a) (-50%, -50%): ;

[0078] c) (+50%, -50%): ;

[0079] d) (-50%, -50%): ;

[0080] in, Indicates different first paths to be disturbed in the first path group to be disturbed The raw values ​​in the GSVA enrichment score matrix, Indicates different first paths to be disturbed in the first path group to be disturbed Raw values ​​in the GSVA enrichment score matrix; Different first paths to be disturbed in the first path group to be disturbed The value after the disturbance is Different first paths to be disturbed in the first path group to be disturbed The value after disturbance; it can be understood that the above value is used for the subsequent calculation of the relative disturbance amplitude of the probabilistic cascade iterative disturbance.

[0081] In another practicable manner, the specific process of performing probabilistic cascade iterative perturbation on the first set of paths to be perturbed is as follows: 1) Select a path in the first set of paths to be perturbed , and calculate the path The relative disturbance amplitude is as follows:

[0082]

[0083] in, For access or The original value in the pathway feature data, path or After the probabilistic cascade iterative perturbation value; it can be understood that when the calculation path When the relative disturbance amplitude (path The processing is the same as that of The disturbance intensity When the disturbance intensity exceeds the preset threshold of 0.3, the detection path Neighbor paths that form a path pair Whether there is a connection, that is, When, and the passage Before being disturbed, the disturbance will propagate to the connected paths with a certain probability (default 0.7), and the propagation probability is proportional to the connection strength. Its dynamic propagation model is a designed probabilistic disturbance propagation rule, as shown below:

[0084]

[0085] in, is the raw activity value of the pathway; is the disturbance amplitude of the upstream pathway (e.g. +50% corresponds to = 0.5); is the connection strength of the pathway; is a random direction variable (simulating inhibition / activation effects), α ∈ [ − 1 , + 1 ] ; is a noise factor that introduces randomness to simulate real organisms. ϕ ∈ [ 0 . 7 , 1 . 3 ] The maximum amplitude of the cascade perturbation is 80% of the original value. Each propagation generates a new perturbation. With multiple rounds of iteration, a cascade reaction network is formed. The default maximum number of cascade iterations is 3. The calculation expression of the perturbation change value is as follows:

[0086]

[0087] Where N is the number of models in the ensemble (precily model provides 5 models for prediction at the same time); is the prediction model of the k-th model; It represents the prediction value of the kth model for the original sample X; it represents the prediction value of the kth model for the sample X' after cascade perturbation. The overall impact of the initial pathway perturbation and the cascade it triggers on drug sensitivity is measured: > 0 means that the perturbation leads to increased drug sensitivity. < 0 indicates that the perturbation leads to increased drug resistance.

[0088] Step 110: predict the sensitivity of the target drug according to the disturbance change value.

[0089] It should be noted that based on existing data and models, the difficulty or intensity of the target drug producing the expected pharmacological effect in different tissue samples (such as different patients or different cell lines, etc.) can be estimated and judged; sensitivity prediction can effectively screen out patient groups that respond well to the target drug, optimize treatment plans, and improve the effectiveness and safety of drug treatment.

[0090] As an example, step 110 includes: inputting the disturbance change value into a preset drug sensitivity prediction model to perform sensitivity prediction on the target drug.

[0091] The above-mentioned drug sensitivity prediction method first identifies the cross-dimensional core pathway set that is jointly regulated by the target drug in different tissue samples, wherein the dimensions spanned by the cross-dimensional core pathway set include at least one of the sample dimension and the drug dimension, and the cross-dimensional pathway set includes multiple pairs of cross-dimensional core pathways, so that multiple pairs of cross-dimensional core pathways under the joint constraints of drugs and tissue samples can be determined; and then, through the association relationship between multiple pairs of cross-dimensional core pathways, a cross-dimensional pathway network that is jointly regulated by the target drug in multiple tissue samples is constructed, and the regulatory dependency relationship and drug response topology between different pathways can be displayed through the cross-dimensional pathway network; and then, in the cross-dimensional pathway network, according to the regulatory association characteristics of the target drug on each cross-dimensional core pathway, multiple pairs of initial pathway groups with biological significance are generated; and then, based on the multiple pairs of initial pathway groups, a first pathway group to be disturbed is set, and based on the pathway feature data of the first pathway group to be disturbed, the pathways under the action of the target drug are disturbed. The response network is subjected to probabilistic cascade iterative perturbations until the perturbation change value of the target drug under the pathway response network is obtained, thereby realizing the dynamic impact of simulated pathway perturbations on the sensitivity of the target drug and quantifying the specific contribution of the pathway. Finally, the sensitivity of the target drug is predicted by the perturbation change value. In the process of sensitivity to the target drug, cross-dimensional pathway network modeling and probabilistic cascade perturbation analysis can characterize the regulatory dependency relationship between pathways and the cascade effect of the pathway network under the action of simulated drugs, thereby achieving the purpose of converting the traditional black box prediction method into a traceable and verifiable prediction process, rather than only outputting quantitative prediction results. Therefore, the technical defect of being difficult to quantify the specific contribution of the pathway to the prediction results of the deep neural network model due to the complex synergistic and antagonistic relationship between pathways is overcome, which makes it easy to cause clinical and scientific research risks. Therefore, the interpretability of drug sensitivity prediction is improved.

[0092] In one embodiment, Figure 2 As shown, a set of cross-dimensional core pathways that are co-regulated by target drugs in multiple tissue samples is identified, including:

[0093] Step 202 , predicting drug sensitivity baseline values ​​of the target drug in the multiple tissue samples based on the drug characteristic information of the target drug and the sample characteristic information of the multiple tissue samples;

[0094] It should be noted that the drug characteristic information is used to characterize the inherent characteristics of the target drug, which can be specifically SMILES structure vector quantized feature data. Specifically, the SMILES structure vector quantized feature data includes the drug name of the target drug and a preset number of feature vector values ​​possessed by the target drug; the sample characteristic information is used to characterize the characteristics of the tissue sample at the biological level, which can be specifically the pathway GSVA enrichment score matrix of the tissue sample; by inputting the drug characteristic information and the sample characteristic information into the existing drug sensitivity prediction model, the drug sensitivity benchmark value can be predicted, wherein the drug sensitivity benchmark value is a reference standard value used to measure the difficulty of the target drug to produce pharmacological effects in multiple tissue samples, which can be specifically the IC50 benchmark value.

[0095] As an example, step 202 includes: fusing the drug characteristic information of the target drug and the sample characteristic information of multiple tissue samples to obtain multimodal fusion information, and predicting the IC50 benchmark value of the target drug in multiple tissue samples by inputting the multimodal fusion information into a preset drug sensitivity prediction model.

[0096] Step 204 : setting a second pathway group to be perturbed that is commonly regulated in multiple tissue samples for the target drug, wherein the second pathway group to be perturbed includes multiple monomer core pathways.

[0097] It should be noted that when there is a lack of an external pathway database, a second group of pathways to be perturbed is constructed through single perturbation analysis to provide a basis for the construction of a cross-dimensional core pathway set. For example, in one feasible method, three monomer core pathways A, B and C are set based on single perturbation experimental analysis, where monomer core pathway A + 50%, monomer core pathway B - 50%, and monomer core pathway C is set to zero, thereby avoiding multi-pathway synergistic interference; it can be understood that single pathway perturbation analysis is equivalent to the computational simulation of in vitro single gene knockout / overexpression experiments, which systematically intervenes in each pathway feature to evaluate its importance and dynamically and quantitatively evaluate the sensitivity and regulatory direction of different pathways.

[0098] As an example, step 204 includes: setting a second group of pathways to be perturbed that are commonly regulated in multiple tissue samples for the target drug, wherein the second group of pathways to be perturbed includes multiple monomer core pathways.

[0099] Step 206 : Based on the drug sensitivity benchmark values, a preset drug sensitivity prediction model is used to perform perturbation analysis on multiple monomer core pathways to obtain multiple drug sensitivity prediction values.

[0100] It should be noted that, by setting up multiple monomer core pathways and performing perturbation analysis on each monomer core pathway, the drug sensitivity prediction values ​​of different monomer core pathways can be obtained. For example, in one feasible method, These are the IC50 predicted values ​​of the three monomer core pathways A, B and C respectively, predicted by the original pathway activity input model.

[0101] As an example, step 206 includes: inputting the drug sensitivity baseline value into a preset drug sensitivity prediction model, performing perturbation analysis on multiple monomer core pathways using the preset drug sensitivity prediction model, and obtaining the drug sensitivity prediction value corresponding to each monomer core pathway.

[0102] Step 208 : constructing a cross-dimensional core pathway set in the multiple monomer core pathways according to the multiple drug sensitivity prediction values.

[0103] It should be noted that after obtaining the drug sensitivity baseline value and drug sensitivity prediction value of each monomer core pathway, quantitative indicators are set for all monomer core pathways in the second group of pathways to be disturbed, and all monomer core pathways are evaluated by comparing the quantitative indicators to obtain the monomer core pathways required for constructing the cross-dimensional core pathway set. For example, in one feasible method, the quantitative indicator is the drug sensitivity change, where the expression of the drug sensitivity change is as follows:

[0104]

[0105] in, The IC50 predicted values ​​of the three monomer core pathways A, B and C are predicted by the original pathway activity input model. Assuming that the IC50 benchmark value is used, the monomer core pathway with the largest change is selected from the three monomer core pathways to construct a cross-dimensional core pathway set.

[0106] As an example, multiple candidate monomer core pathways whose drug sensitivity prediction values ​​are greater than a preset drug sensitivity threshold are determined among all monomer core pathways, and a cross-dimensional core pathway set is constructed based on the multiple candidate monomer core pathways.

[0107] In this embodiment, in the process of determining the cross-dimensional core pathway set, if there is a lack of an external pathway database, perturbation analysis is performed on all existing monomer core pathways, and based on the changes in different monomer core pathways before and after the perturbation, multiple candidate monomer core pathways are selected from all monomer core pathways to construct a cross-dimensional core pathway set. Therefore, while laying the foundation for improving the interpretability of drug sensitivity prediction, the limitations of drug sensitivity prediction are reduced.

[0108] In one embodiment, among multiple monomer core pathways, a cross-dimensional core pathway set is constructed based on multiple drug sensitivity prediction values, including:

[0109] Determine the pathway contribution weight mean corresponding to the preset drug sensitivity prediction model and the pathway contribution weight values ​​corresponding to each of the multiple monomer core pathways; based on the size relationship between the pathway contribution weight mean and the multiple pathway contribution weight values, select multiple first monomer core pathways from all monomer core pathways, and construct a first monomer core pathway set based on the multiple first monomer core pathways; based on the multiple drug sensitivity prediction values, select multiple second monomer core pathways from all monomer core pathways, and construct a second monomer core pathway set based on the multiple second monomer core pathways; construct a cross-dimensional core pathway set based on the first monomer core pathway set and the second monomer core pathway set.

[0110] It should be noted that in order to further improve the interpretability of drug sensitivity prediction, when constructing a cross-dimensional core pathway set, the synergistic and antagonistic relationships between pathways and drug effects are taken into consideration, so that monomer core pathways with intersections are selected based on the single pathway perturbation analysis results and model weight analysis. For example, in one feasible method, the common pathway C is determined for the pathways obtained in the single pathway perturbation analysis and the pathways obtained in the model weight analysis, and a cross-dimensional core pathway set is constructed based on pathway C, wherein the pathway contribution weight value is used to characterize the degree of influence of a single core pathway on the prediction result, and the pathway contribution weight mean is used to characterize the average contribution of all monomer core pathways to the prediction result; the first monomer core pathway set refers to a set of monomer core pathways that have a high contribution to the prediction result, and the second monomer core pathway is used to characterize the set of monomer core pathways that have a key influence on the drug effect.

[0111] As an example, the layer weight matrix of the preset drug sensitivity prediction model is used to calculate the mean pathway contribution weight and the pathway contribution weight values ​​corresponding to all monomer core pathways; the monomer core pathway whose pathway contribution weight value is greater than the mean pathway contribution weight is taken as the first monomer core pathway, and multiple first monomer core pathways are integrated into a first monomer core pathway set; the monomer core pathway whose drug sensitivity prediction value is greater than the preset drug sensitivity prediction value threshold is taken as the second monomer core pathway, and multiple second monomer core pathways are integrated into a second monomer core pathway set; candidate monomer core pathways with intersections are selected from the first monomer core pathway set and the second monomer core pathway set, and multiple candidate monomer core pathways are integrated into a cross-dimensional core pathway set.

[0112] In one possible implementation, the weight matrix of the first layer of the DNN model is extracted , and calculate the mean of the pathway contribution weights ,in, The calculation formula is as follows:

[0113]

[0114] in, is the mean value of the path contribution weight, is the connection weight from the jth input feature to the kth hidden unit, d is the input dimension, and h is the number of hidden units, thereby quantifying the feature contribution. This analysis, without relying on specific sample inputs, directly and statically reveals the contribution weights of each pathway in the decision-making process. Compared to single-pathway perturbation analysis, it provides a global perspective, understanding the parameter patterns learned by the model during training, which can be used to evaluate which pathways the model focuses on when making decisions.

[0115] In this embodiment, when constructing a cross-dimensional core pathway set, the first monomer core pathway set and the second monomer core pathway set are determined by weight quantification indicators and drug sensitivity quantification indicators, so that candidate monomer core pathways that take into account both basic importance and dynamic sensitivity can be screened, making the constructed cross-dimensional core pathway set more consistent with biological logic, thus laying the foundation for further improving the interpretability of drug sensitivity prediction.

[0116] In one embodiment, a cross-dimensional core pathway set is constructed based on the first monomer core pathway set and the second monomer core pathway set, including:

[0117] Determine multiple candidate monomer core pathways that have intersections in the first monomer core pathway set and the second monomer core pathway set; standardize the drug sensitivity prediction values ​​and pathway contribution weight values ​​of the multiple candidate monomer core pathways to obtain the corresponding prediction standard values ​​and weight standard values ​​of the multiple candidate monomer core pathways; obtain the pathway importance of each candidate monomer core pathway by fusing the prediction standard value and weight standard value of each candidate monomer core pathway; select multiple target monomer core pathways from all candidate monomer core pathways based on multiple pathway importances, and integrate the multiple target monomer core pathways into a cross-dimensional core pathway set.

[0118] It should be noted that the monomer core pathways that belong to both the first monomer core pathway set and the second monomer core pathway set have both high weight and high sensitivity characteristics, and are key pathways in the drug action mechanism. The standardization process is to convert the drug sensitivity prediction value and the pathway contribution weight value into dimensionless standard values, respectively, to eliminate the influence of different data dimensions. The target monomer core pathway is a core pathway screened based on the pathway importance, which has both high weight and high sensitivity characteristics. It integrates the set of all target pathways and can reflect the multi-level regulatory network of drug action.

[0119] As an example, multiple candidate monomer core pathways with intersections are screened in the first monomer core pathway set and the second monomer core pathway set; the drug sensitivity prediction values ​​of the multiple candidate monomer core pathways are standardized to obtain the prediction standard values ​​corresponding to the multiple candidate monomer core pathways, and the pathway contribution weight values ​​of the multiple candidate monomer core pathways are standardized to obtain the weight standard values ​​corresponding to the multiple candidate monomer core pathways; the prediction standard value and the weight standard value of each candidate monomer core pathway are fused using a preset fusion formula to obtain the pathway importance of each candidate monomer core pathway; based on the size relationship between multiple pathway importances and the preset pathway importance, multiple target monomer core pathways are selected from all candidate monomer core pathways, and the multiple target monomer core pathways are integrated into a cross-dimensional core pathway set.

[0120] In one embodiment, the maximum impact value of each channel in the single channel perturbation analysis is and the absolute weight values ​​in the model weight analysis Perform standardization processing, where the formula for standardization processing is as follows:

[0121]

[0122]

[0123] in, is the predicted standard value, is the weight standard value, is the drug sensitivity prediction value, is the contribution weight value of the path, where and The value range of is [0,1]. It is the monomer core pathway with the largest impact value among the candidate monomer core pathways. is the monomer core pathway with the largest absolute weight value; the predicted standard value and weight standard value of each candidate monomer core pathway are fused to obtain the pathway importance of each candidate monomer core pathway, wherein the fusion formula is as follows:

[0124]

[0125] in, For access The importance of the pathway, is the weight coefficient, which can be 0.5. After completing the weight integration calculation for all pathways, the integrated importance scores are used to sort and obtain the top N pathways as the input for subsequent network construction. The default number is 50.

[0126] In this embodiment, in the process of constructing a cross-dimensional core pathway set, multiple candidate monomer core pathways that have intersections in the first monomer core pathway set and the second monomer core pathway set are first screened, and then standardized respectively to eliminate the dimensional differences between different values. Finally, the target monomer core pathway is determined by fusion of the obtained pathway importance, and a cross-dimensional core pathway set is constructed with a target monomer core pathway that has both high weight and high sensitivity. Therefore, the recognition accuracy of genome sequencing data variation identification is further improved.

[0127] In one embodiment, the association relationship includes association similarity; based on the association relationship between multiple pairs of cross-dimensional core pathways, a cross-dimensional pathway network co-regulated by the target drug in multiple tissue samples is constructed, including:

[0128] According to the size relationship between the association similarity and the preset association similarity threshold, multiple pairs of cross-dimensional core pathways with association relationships are determined; by constructing cross-dimensional regulatory edges of multiple pairs of cross-dimensional core pathways and setting edge weights for the cross-dimensional regulatory edges, a cross-dimensional pathway network that is jointly regulated by target drugs in multiple tissue samples is constructed.

[0129] As an example, cross-dimensional core pathways with correlation similarity greater than a preset correlation similarity threshold are selected as multiple cross-dimensional core pathway pairs, wherein the calculation formula of the correlation similarity is as follows:

[0130]

[0131] in, is the association similarity, which is used to characterize the proportion of genes shared between each pair of pathways. G1 is the gene set included in pathway p1, and G2 is the gene set included in pathway p2 if and only if When ≥0.05, a cross-dimensional core pathway pair is selected; by constructing cross-dimensional regulatory edges of multiple pairs of cross-dimensional core pathway pairs and setting edge weights for the cross-dimensional regulatory edges, a cross-dimensional pathway network in which the target drug is jointly regulated in multiple tissue samples is constructed, wherein the cross-dimensional regulatory edge represents a directed edge of the regulatory relationship between the cross-dimensional core pathway pairs, the direction of the arrow indicates the regulatory direction, and the edge weight is used to quantify the numerical value of the cross-dimensional regulatory edge strength, reflecting the significance and consistency of the regulatory relationship between the pathway pairs. For example, in an implementable manner, in the pathway and access roads A cross-dimensional control edge is established between them, where is the edge weight.

[0132] In this embodiment, by integrating data from multiple tissue samples, the constructed pathway network reflects the core regulatory mechanism of drug action, avoiding the randomness of single tissue samples. At the same time, the edge weights clarify the strength of the regulatory relationship between pathway pairs, facilitating the screening of key regulatory nodes. Furthermore, by constructing multiple pairs of cross-dimensional core pathway pairs' cross-dimensional regulatory edges and setting edge weights for the cross-dimensional regulatory edges, a cross-dimensional pathway network in which the target drug is jointly regulated in multiple tissue samples is constructed. This can be expanded into a cross-dimensional pathway network to analyze changes in the pathway network under different drug concentrations or treatment times. Therefore, this lays the foundation for further improving the interpretability of drug sensitivity prediction.

[0133] In one embodiment, the regulatory association feature includes betweenness centrality; in a cross-dimensional pathway network, based on the regulatory association feature of the target drug on each cross-dimensional core pathway, multiple pairs of interpretable initial pathway groups are generated, including:

[0134] In a cross-dimensional pathway network, according to the betweenness centrality of the target drug to each cross-dimensional core pathway, all cross-dimensional core pathways are divided into multiple pairs of first cross-dimensional core pathways and multiple pairs of second cross-dimensional core pathways, wherein the betweenness centrality of multiple pairs of first cross-dimensional core pathways is greater than the preset betweenness centrality threshold, and the betweenness centrality of multiple pairs of second cross-dimensional core pathways is less than the preset betweenness centrality threshold; according to the edge weights of multiple pairs of first cross-dimensional core pathways, the first target cross-dimensional core pathway is selected from all first cross-dimensional core pathways; according to the edge weights of multiple pairs of second cross-dimensional core pathways, the second target cross-dimensional core pathway is selected from all second cross-dimensional core pathways; the third target cross-dimensional core pathway is selected from all first cross-dimensional core pathways and all second cross-dimensional core pathways; the first target cross-dimensional core pathway, the second target cross-dimensional core pathway and the third target cross-dimensional core pathway are used to generate multiple pairs of initial pathway groups.

[0135] It should be noted that betweenness centrality is used to measure the importance of cross-dimensional core pathways. In the cross-dimensional pathway network, cross-dimensional core pathways with betweenness centrality greater than the preset betweenness centrality threshold play a key connection and regulatory role in the network; cross-dimensional core pathways with betweenness centrality less than the preset betweenness centrality threshold are less critical in the network than the first cross-dimensional core pathway, but still play a certain role in the drug action process; the initial pathway group is to screen out pathway pairs with significant regulation, known functions and close interactions, which can be used as the primary analysis unit for drug mechanism research. In the process of generating multiple pairs of initial pathway groups with interpretability, In , different types of target cross-dimensional core pathways can be generated based on the combination strategy to ensure the richness of cross-dimensional core pathways. Among them, the first target cross-dimensional core pathway refers to the cross-dimensional core pathway with the strongest regulatory association or the most important role selected from all first cross-dimensional core pathways according to the size of the edge weight. The second target cross-dimensional core pathway refers to the key cross-dimensional core pathway selected according to the size of the edge weight from all second cross-dimensional core pathways. The third target cross-dimensional core pathway refers to the cross-dimensional core pathway selected from all first cross-dimensional core pathways and second cross-dimensional core pathways by comprehensive consideration, which can be selected by random selection.

[0136] As an example, in an interdimensional pathway network, multiple pairs of interdimensional core pathways whose betweenness centrality is greater than a preset betweenness centrality threshold are divided into first interdimensional core pathways, and multiple pairs of interdimensional core pathways whose betweenness centrality is less than or equal to the preset betweenness centrality threshold are divided into second interdimensional core pathways; multiple first interdimensional core pathways whose edge weights are greater than a first preset edge weight threshold are selected from all first interdimensional core pathways as first target interdimensional core pathways, wherein the first preset edge weight threshold can be specifically 0.5; multiple second interdimensional core pathways whose edge weights are equal to a second preset edge weight threshold are selected from all second interdimensional core pathways as second target interdimensional core pathways, wherein the second preset edge weight threshold can be specifically 0; a preset number of third target interdimensional core pathways are randomly selected from all first interdimensional core pathways and all second interdimensional core pathways; the first target interdimensional core pathways, the second target interdimensional core pathways and the third target interdimensional core pathways are combined into multiple pairs of initial pathway groups.

[0137] In one implementation, betweenness centrality can be calculated using the following formula:

[0138]

[0139] in, is the total number of shortest paths from the inter-dimensional core path s to the inter-dimensional core path t, is the number of shortest paths from s to t through the interdimensional core path p, is the betweenness centrality of the cross-dimensional core path p.

[0140] In another practicable manner, for multiple pairs of initial path groups The design is as follows:

[0141] 1) The first target cross-dimensional core pathway can refer to a strongly connected pathway combination, that is, a closely related pathway pair. During the design process, for each pair of first cross-dimensional core pathways, , get the edge weight of the first cross-dimensional core path , sort all the first cross-dimensional core paths in descending order according to the size of the edge weight, and select the top The high-weight combination S of , where the expression of S is as follows:

[0142]

[0143] Among them, S represents the first target cross-dimensional core pathway, represents each pair of first interdimensional core pathways, represents the edge weight of each pair of first cross-dimensional core pathways;

[0144] 2) The second-target cross-dimensional core pathway can refer to a weakly connected pathway combination, that is, a functionally independent pathway pair. During the design process, the second-target cross-dimensional core pathway with an edge weight of 0 is selected as the functionally independent combination W, where W is expressed as follows:

[0145]

[0146] Among them, W represents the first target cross-dimensional core pathway, represents each pair of first interdimensional core pathways, Represents the edge weight of each pair of first cross-dimensional core pathways, so that it can be guaranteed that there are both strongly correlated pathways and independent pathways in multiple pairs of initial pathway groups;

[0147] 3) To ensure that multiple pairs of initial pathway groups meet the requirements of the preset number of pathway groups, a third target interdimensional core pathway is randomly selected from all first interdimensional core pathways and all target interdimensional core pathways. In addition, for example, in one feasible method, the number of preset channel groups of multiple pairs of initial channel groups is 30, the number of first target cross-dimensional core channels is 12, the number of second target cross-dimensional core channels is 12, and the number of third target cross-dimensional core channels is 6.

[0148] In this embodiment, by dividing the first cross-dimensional core pathway and the second cross-dimensional core pathway based on betweenness centrality and selecting the target cross-dimensional core pathway based on edge weight to generate an initial pathway group, it is possible to screen out pathway combinations with significant regulation, known functions, and close interactions from two important dimensions: the criticality of the pathway in the network and the strength of the association between pathways. At the same time, by selecting a third target cross-dimensional core pathway from all first cross-dimensional core pathways and all second cross-dimensional core pathways, it is possible to further simulate the randomness of pathway interactions in real biological systems. Then, by integrating the first target cross-dimensional core pathway, the second target cross-dimensional core pathway, and the third target cross-dimensional core pathway together to obtain an initial pathway group, which can serve as the primary analysis unit for drug mechanism research and clearly demonstrate the interaction relationship between the target drug and the cross-dimensional core pathway. Generating multiple pairs of initial pathway groups in the above manner can more accurately focus on the cross-dimensional core pathways that play a key role in the sensitivity of the target drug, making the sensitivity prediction of the target drug more targeted, thereby effectively and simultaneously improving the prediction accuracy and interpretability of the drug prediction.

[0149] In one embodiment, after predicting the sensitivity of the target drug based on the disturbance change value, the method further includes:

[0150] After predicting the sensitivity of the target drug according to the disturbance change value, the method further includes:

[0151] Obtain pathway network characteristic values ​​of a cross-dimensional pathway network; generate cascade regulation breadth information and perturbation effect intensity information corresponding to the target drug based on the pathway network characteristic values ​​and perturbation change values; and display the cascade regulation breadth information and perturbation effect intensity information on a preset display interface.

[0152] It should be noted that the network characteristic value is a quantitative indicator used to characterize the topological structure of the cross-dimensional pathway network, which can describe the global characteristics of the network, specifically the characteristic path length or clustering coefficient, etc.; the cascade regulation breadth information is used to describe the breadth of propagation of the disturbance signal of the target drug in the cross-dimensional pathway network, specifically the number of pathways x affected by the cascade and the propagation level y; the disturbance effect intensity information is used to quantify the regulatory efficacy of the target drug on a specific cross-dimensional core pathway in the cross-dimensional pathway network, specifically the absolute amplitude of the disturbance change By displaying the cascade control breadth information and the disturbance effect intensity information, the analysis results can be presented intuitively, supporting user interactive exploration and decision-making. It is used to reflect the intensity of the disturbance impact. For example, in one practicable manner, the disturbance change value can be analyzed. The corresponding relationship between the number of pathways x affected by the cascade can further evaluate the correlation between the scope of the cascade effect and the change in drug efficacy, providing a basis for identifying key regulatory pathways.

[0153] In one practicable manner, the calculation formula of the global connectivity efficiency E is as follows:

[0154]

[0155] Where E is the global connectivity efficiency, n is the number of pathways in the cross-dimensional pathway network, It is the shortest path distance between the cross-dimensional core pathway s and the cross-dimensional core pathway t; it can be understood that the local network efficiency provides a quantitative indicator of the overall signal conduction capability of the network. A high efficiency value indicates that the disturbance signal in the network can be transmitted more efficiently. This indicator can be recorded as a benchmark value of the pathway network characteristics.

[0156] It should be noted that the cascade effect simulation algorithm based on the above embodiment can further predict the adverse reactions of drugs. For example, the cascade disturbance is calculated under the conditions of the default cascade propagation parameters (trigger threshold) and 0.7 (propagation probability). For each disturbance combination, the number of pathways affected by the cascade, the absolute amplitude of the disturbance change, and the absolute amplitude of the disturbance change are recorded. and global connectivity efficiency E; further, with the number of pathways affected by the cascade x as the X-axis, the absolute amplitude of the disturbance change For the Y axis, draw a bar chart, and the bar color is based on the disturbance change value The sign of the IC50 is used to determine the effect, with an increase in IC50 represented by red and a decrease in IC50 represented by green. This allows for the identification of high-risk combinations when combined perturbations lead to large-scale cascade effects and when perturbation changes significantly. At the same time, the recording of the global connectivity efficiency E helps identify key cross-dimensional core pathways.

[0157] As an example, the pathway network characteristic values ​​of the cross-dimensional pathway network are obtained; based on the pathway network characteristic values, the number of pathways affected by the cascade under the action of the target drug is calculated, and the absolute amplitude of the disturbance change is calculated by inputting the disturbance change value into the preset disturbance change calculation formula; with the number of pathways affected by the cascade as the x-axis and the absolute amplitude of the disturbance change as the y-axis, a bar chart is generated, and the bar chart is displayed on a preset display interface.

[0158] In this embodiment, the pathway network characteristic values ​​of the cross-dimensional pathway network are first obtained, and then the perturbation behavior of the target drug is quantified to generate cascade regulation breadth information and perturbation effect intensity information, respectively. The cascade regulation breadth information and perturbation effect intensity information are visualized, which can clearly demonstrate the complex biological network and drug action mechanism to non-professionals, thereby further improving the interpretability of drug sensitivity prediction.

[0159] In one practicable manner, referring to Figure 3 , Figure 3 This is a schematic diagram of the overall process for drug sensitivity prediction. First, the baseline prediction value is obtained. When the pathway database is not available, single pathway perturbation analysis and model weight analysis are performed, and the pathway importance is obtained by weighted fusion. Finally, multiple pairs of cross-dimensional core pathways are obtained. When the pathway database is available, multiple pairs of cross-dimensional core pathways are directly selected to generate pathway combinations. The cascade effect is simulated through combined perturbation experiments, and the perturbation change value is finally calculated to predict the sensitivity of the target drug.

[0160] Since, in the process of target drug sensitivity assessment, cross-dimensional pathway network modeling and probabilistic cascade perturbation analysis can characterize the regulatory dependencies between pathways and simulate the cascade effects of pathway networks under drug action, the purpose of converting the traditional black-box prediction method into a traceable and verifiable prediction process is achieved, rather than only outputting quantitative prediction results. Therefore, the technical defect of being difficult to quantify the specific contribution of pathways to the prediction results of deep neural network models due to the complex synergistic and antagonistic relationships between pathways is overcome, which makes it easy to lead to clinical and scientific research risks. Therefore, the interpretability of drug sensitivity prediction is improved.

[0161] It should be understood that, although the steps in the flowcharts of the above embodiments are shown in sequence as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be performed in other orders. Moreover, at least a portion of the steps in the flowcharts of the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily to be performed in sequence, but can be performed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0162] Based on the same inventive concept, embodiments of the present application also provide a drug sensitivity prediction device for implementing the aforementioned drug sensitivity prediction method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of the embodiments of the one or more test drug sensitivity prediction devices provided below can be found in the above-mentioned limitations of the drug sensitivity prediction method and will not be repeated here.

[0163] In an exemplary embodiment, Figure 4 As shown, a drug sensitivity prediction device is provided, comprising: a determination module 301, a construction module 302, a generation module 303, a perturbation module 304 and a prediction module 305, wherein:

[0164] Determination module 301 is used to determine a cross-dimensional core pathway set that is co-regulated by a target drug in multiple tissue samples, wherein the dimension spanned by the cross-dimensional core pathway set includes at least one of a sample dimension and a drug dimension, and the cross-dimensional core pathway set includes multiple pairs of cross-dimensional core pathways;

[0165] A construction module 302 is used to construct a cross-dimensional pathway network that is co-regulated by the target drug in multiple tissue samples based on the association relationships between multiple pairs of cross-dimensional core pathways;

[0166] A generation module 303 is used to generate multiple pairs of interpretable initial pathway groups in the cross-dimensional pathway network based on the regulatory association characteristics of the target drug on each cross-dimensional core pathway;

[0167] The perturbation module 304 is configured to set a first pathway group to be perturbed based on the multiple pairs of initial pathway groups, and perform a probabilistic cascade iterative perturbation on the pathway response network under the action of the target drug based on the pathway characteristic data of the first pathway group to be perturbed, until a perturbation change value of the target drug under the pathway response network is obtained;

[0168] The prediction module 305 is used to predict the sensitivity of the target drug according to the disturbance change value.

[0169] In one embodiment, the determining module 301 is further configured to:

[0170] Based on the drug characteristic information of the target drug and the sample characteristic information of multiple tissue samples, the drug sensitivity baseline value of the target drug in the multiple tissue samples is predicted; a second pathway group to be disturbed that is jointly regulated in the multiple tissue samples is set for the target drug, wherein the second pathway group to be disturbed includes multiple monomer core pathways; based on the drug sensitivity baseline value, the multiple monomer core pathways are separately perturbed and analyzed using a preset drug sensitivity prediction model to obtain multiple drug sensitivity prediction values; among the multiple monomer core pathways, a cross-dimensional core pathway set is constructed based on the multiple drug sensitivity prediction values.

[0171] In one embodiment, the determining module 301 is further configured to:

[0172] Determine the pathway contribution weight mean corresponding to the preset drug sensitivity prediction model and the pathway contribution weight values ​​corresponding to each of the multiple monomer core pathways; based on the size relationship between the pathway contribution weight mean and the multiple pathway contribution weight values, select multiple first monomer core pathways from all monomer core pathways, and construct a first monomer core pathway set based on the multiple first monomer core pathways; based on the multiple drug sensitivity prediction values, select multiple second monomer core pathways from all monomer core pathways, and construct a second monomer core pathway set based on the multiple second monomer core pathways; construct a cross-dimensional core pathway set based on the first monomer core pathway set and the second monomer core pathway set.

[0173] In one embodiment, the determining module 301 is further configured to:

[0174] Determine multiple candidate monomer core pathways that have intersections in the first monomer core pathway set and the second monomer core pathway set; standardize the drug sensitivity prediction values ​​and pathway contribution weight values ​​of the multiple candidate monomer core pathways to obtain the corresponding prediction standard values ​​and weight standard values ​​of the multiple candidate monomer core pathways; obtain the pathway importance of each candidate monomer core pathway by fusing the prediction standard value and weight standard value of each candidate monomer core pathway; select multiple target monomer core pathways from all candidate monomer core pathways based on multiple pathway importances, and integrate the multiple target monomer core pathways into a cross-dimensional core pathway set.

[0175] In one embodiment, the association relationship includes association similarity; the construction module 302 is further configured to:

[0176] According to the size relationship between the association similarity and the preset association similarity threshold, multiple pairs of cross-dimensional core pathways with association relationships are determined; by constructing cross-dimensional regulatory edges of multiple pairs of cross-dimensional core pathways and setting edge weights for the cross-dimensional regulatory edges, a cross-dimensional pathway network that is jointly regulated by target drugs in multiple tissue samples is constructed.

[0177] In one embodiment, the regulatory association feature includes betweenness centrality; the generating module 303 is further configured to:

[0178] In a cross-dimensional pathway network, according to the betweenness centrality of the target drug to each cross-dimensional core pathway, all cross-dimensional core pathways are divided into multiple pairs of first cross-dimensional core pathways and multiple pairs of second cross-dimensional core pathways, wherein the betweenness centrality of multiple pairs of first cross-dimensional core pathways is greater than the preset betweenness centrality threshold, and the betweenness centrality of multiple pairs of second cross-dimensional core pathways is less than the preset betweenness centrality threshold; according to the edge weights of multiple pairs of first cross-dimensional core pathways, the first target cross-dimensional core pathway is selected from all first cross-dimensional core pathways; according to the edge weights of multiple pairs of second cross-dimensional core pathways, the second target cross-dimensional core pathway is selected from all second cross-dimensional core pathways; the third target cross-dimensional core pathway is selected from all first cross-dimensional core pathways and all second cross-dimensional core pathways; the first target cross-dimensional core pathway, the second target cross-dimensional core pathway and the third target cross-dimensional core pathway are used to generate multiple pairs of initial pathway groups.

[0179] In one embodiment, the drug sensitivity prediction device is further used to:

[0180] Obtain pathway network characteristic values ​​of a cross-dimensional pathway network; generate cascade regulation breadth information and perturbation effect intensity information corresponding to the target drug based on the pathway network characteristic values ​​and perturbation change values; and display the cascade regulation breadth information and perturbation effect intensity information on a preset display interface.

[0181] Each module in the drug sensitivity prediction device described above can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0182] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 5As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface, the display unit and the input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, it realizes a method for predicting drug sensitivity. Those skilled in the art can understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0183] In one embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0184] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0185] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0186] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), data processing logic devices based on quantum computing, and the like.

[0187] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0188] The above embodiments merely illustrate several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for predicting drug sensitivity, characterized in that: The method comprises: Determining a cross-dimensional core pathway set that is co-regulated by a target drug in multiple tissue samples, wherein the dimension spanned by the cross-dimensional core pathway set includes at least one of a sample dimension and a drug dimension, and the cross-dimensional core pathway set includes multiple pairs of cross-dimensional core pathways; Constructing a cross-dimensional pathway network co-regulated by the target drug in the multiple tissue samples based on the association relationships between the multiple pairs of cross-dimensional core pathways; In the cross-dimensional pathway network, multiple pairs of initial pathway groups with interpretability are generated based on the regulatory association characteristics of the target drug on each cross-dimensional core pathway; A first pathway group to be disturbed is set based on the multiple pairs of initial pathway groups, and based on the pathway characteristic data of the first pathway group to be disturbed, a probabilistic cascade iterative perturbation is performed on the pathway response network under the action of the target drug until a perturbation change value of the target drug under the pathway response network is obtained; The sensitivity of the target drug is predicted according to the disturbance change value.

2. The method according to claim 1, characterized in that The cross-dimensional core pathway set co-regulated by the target drug in multiple tissue samples includes: predicting a drug sensitivity baseline value of the target drug in the multiple tissue samples based on the drug characteristic information of the target drug and the sample characteristic information of the multiple tissue samples; Setting a second group of pathways to be disturbed that are commonly regulated in multiple tissue samples for the target drug, wherein the second group of pathways to be disturbed includes multiple monomer core pathways; According to the drug sensitivity benchmark value, performing perturbation analysis on the multiple monomer core pathways respectively using a preset drug sensitivity prediction model to obtain multiple drug sensitivity prediction values; Among the multiple monomer core pathways, the cross-dimensional core pathway set is constructed according to the multiple drug sensitivity prediction values.

3. The method according to claim 2, characterized in that The cross-dimensional core pathway set is constructed according to the multiple drug sensitivity prediction values ​​in the multiple monomer core pathways, including: Determining a mean pathway contribution weight corresponding to the preset drug sensitivity prediction model and a pathway contribution weight value corresponding to each of the multiple monomer core pathways; Selecting a plurality of first monomer core pathways from all monomer core pathways according to a relationship between the pathway contribution weight mean and the plurality of pathway contribution weights, and constructing a first monomer core pathway set based on the plurality of first monomer core pathways; selecting a plurality of second monomer core pathways from all the monomer core pathways according to the plurality of drug sensitivity prediction values, and constructing a second monomer core pathway set based on the plurality of second monomer core pathways; The cross-dimensional core pathway set is constructed based on the first monomer core pathway set and the second monomer core pathway set.

4. The method according to claim 3, characterized in that The constructing of the cross-dimensional core pathway set according to the first monomer core pathway set and the second monomer core pathway set includes: Determine a plurality of candidate monomer core pathways having intersections between the first monomer core pathway set and the second monomer core pathway set; Normalizing the drug sensitivity prediction values ​​and pathway contribution weight values ​​of the multiple candidate monomer core pathways to obtain prediction standard values ​​and weight standard values ​​corresponding to each of the multiple candidate monomer core pathways; By fusing the predicted standard value and weight standard value of each candidate monomer core pathway, the pathway importance of each candidate monomer core pathway is obtained; According to the importance of multiple pathways, multiple target monomer core pathways are selected from all candidate monomer core pathways, and the multiple target monomer core pathways are integrated into the cross-dimensional core pathway set.

5. The method according to claim 1, wherein The association relationship includes an association similarity; and constructing a cross-dimensional pathway network co-regulated by the target drug in the multiple tissue samples based on the association relationship between the multiple pairs of cross-dimensional core pathways includes: Determining, based on a magnitude relationship between the correlation similarity and a preset correlation similarity threshold, a plurality of pairs of cross-dimensional core pathways having a correlation relationship among the plurality of pairs of cross-dimensional core pathways; By constructing cross-dimensional regulatory edges of the multiple pairs of cross-dimensional core pathways and setting edge weights for the cross-dimensional regulatory edges, a cross-dimensional pathway network that is co-regulated by the target drug in the multiple tissue samples is constructed.

6. The method according to claim 5, characterized in that The regulatory association feature includes betweenness centrality; in the cross-dimensional pathway network, based on the regulatory association feature of the target drug on each cross-dimensional core pathway, multiple pairs of initial pathway groups with interpretability are generated, including: In the cross-dimensional pathway network, all cross-dimensional core pathways are divided into multiple pairs of first cross-dimensional core pathways and multiple pairs of second cross-dimensional core pathways according to the betweenness centrality of the target drug to each cross-dimensional core pathway, wherein the betweenness centralities of the multiple pairs of first cross-dimensional core pathways are all greater than a preset betweenness centrality threshold, and the betweenness centralities of the multiple pairs of second cross-dimensional core pathways are all less than the preset betweenness centrality threshold; Selecting a first target inter-dimensional core path from all first inter-dimensional core paths according to the edge weights of the multiple pairs of first inter-dimensional core paths; Selecting a second target inter-dimensional core pathway from all second inter-dimensional core pathways according to the edge weights of the multiple pairs of second inter-dimensional core pathways; Selecting a third target interdimensional core pathway from all of the first interdimensional core pathways and all of the second interdimensional core pathways; The first target cross-dimensional core pathway, the second target cross-dimensional core pathway and the third target cross-dimensional core pathway are used to generate the multiple pairs of initial pathway groups.

7. The method according to claim 1, characterized in that After predicting the sensitivity of the target drug according to the disturbance change value, the method further includes: Obtaining a pathway network characteristic value of the cross-dimensional pathway network; Generating cascade regulation breadth information and perturbation effect intensity information corresponding to the target drug according to the pathway network characteristic value and the perturbation change value; The cascade control breadth information and the disturbance effect intensity information are displayed on a preset display interface.

8. A drug sensitivity prediction device, characterized in that: The device comprises: a determination module, configured to determine a set of cross-dimensional core pathways co-regulated by a target drug in multiple tissue samples, wherein the dimensions spanned by the cross-dimensional core pathway set include at least one of a sample dimension and a drug dimension, and the cross-dimensional core pathway set includes multiple pairs of cross-dimensional core pathways; A construction module, configured to construct a cross-dimensional pathway network co-regulated by the target drug in the multiple tissue samples based on the association relationships between the multiple pairs of cross-dimensional core pathways; A generation module is used to generate a plurality of pairs of initial pathway groups with interpretability in the cross-dimensional pathway network according to the regulatory association characteristics of the target drug on each cross-dimensional core pathway; a perturbation module, configured to set a first pathway group to be perturbed based on the multiple pairs of initial pathway groups, and perform a probabilistic cascade iterative perturbation on the pathway response network under the action of the target drug based on the pathway characteristic data of the first pathway group to be perturbed, until a perturbation change value of the target drug under the pathway response network is obtained; The prediction module is used to predict the sensitivity of the target drug according to the disturbance change value.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

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

Images