Disturbance response scanning analysis-based lung squamous cell carcinoma drug discovery method and system

Abstract

The invention discloses a lung squamous cell carcinoma drug discovery method and system based on disturbance response scanning analysis, and belongs to the field of biological medicines.The method comprises the steps that lung squamous cell carcinoma protein expression profile data is downloaded and preprocessed, a robust network module is constructed and multi-scale analysis and evaluation are conducted, and drug genomic data are obtained based on drug genomic data. Predicting the drug sensitivity of the patient by using machine learning, and generating a DRN through difference analysis of the drug sensitivity; predicting drug-target protein binding affinity in combination with deep learning, quantifying node sensitivity through a PRS technology, calculating a DPI disturbance score and ranking drug priorities; high-ranking drugs are screened through comprehensive sensitivity analysis and literature investigation; a drug for a lung squamous carcinoma cell line is screened out through a pre-experiment and is relocated. According to the framework developed by the invention, by integrating proteomics, pharmacogenomics and dynamic network analysis, systematic analysis is provided for relocation drug identification of lung squamous cell carcinoma, and good news is brought to treatment of lung squamous cell carcinoma.

Description

Technical Field

[0001] The present invention belongs to the field of biomedicine, and in particular relates to a method and system for discovering drugs for squamous cell lung cancer based on perturbation response scanning analysis. Background Art

[0002] Squamous cell lung carcinoma (SCC) is a major subtype of non-small cell lung cancer (NSCLC), accounting for approximately 20%-30% of all cases. Its incidence is high and is often closely associated with a history of smoking. Despite recent progress in targeted therapy and immunotherapy for NSCLC, the treatment of SCC remains challenging, including the lack of effective targeted drugs and the emergence of acquired drug resistance. Therefore, the development of new therapeutic strategies is urgently needed. In this context, drug repositioning (DR), particularly network biology-based approaches, has opened up a promising avenue for cancer treatment by analyzing how drugs affect cellular networks at a systemic scale. However, most network-based DR methods only provide a static view of the problem and are often limited by the incompleteness and inherent biases of network data. With the advancement of omics technologies, drug-related experimental data continues to accumulate, and the resulting large-scale pharmacogenomic data provide rich data support for effectively predicting detailed association maps between disease and drug effects. Furthermore, while traditional genomic analyses have provided important insights into the complex mechanisms of SCC, they remain limited in characterizing the dynamics of tumor progression and guiding personalized treatment. In recent years, the emergence of large-scale proteomic data has provided a new dimension for elucidating specific regulatory mechanisms and is expected to become an important support for precision medicine.

[0003] Therefore, based on the proteomic map and drug genome data of lung squamous cell carcinoma, how to combine network dynamics models and evaluate the overall perturbation effect of drugs on the system has become a key scientific issue in DR research. Summary of the Invention

[0004] In view of the above existing problems, the present invention provides a drug repositioning framework based on perturbation response scanning analysis for drug discovery of squamous cell lung cancer.

[0005] To solve the above technical problems, a drug discovery method for lung squamous cell carcinoma based on perturbation response scanning analysis was proposed, including:

[0006] Protein expression profile data of lung squamous cell carcinoma were downloaded and preprocessed. Differential proteins were obtained through differential expression analysis and mapped to the protein-protein interaction (PPI) network to construct a differential network. The differential network was divided into modules to obtain robust network modules. The importance of robust network modules was evaluated through multi-scale analysis, and the target modules were screened and the generalization ability was verified. Based on drug genome data, machine learning was used to predict patient drug sensitivity, and a drug response network (DRN) was generated through differential analysis of drug sensitivity. Deep learning was combined to predict drug-target protein binding affinity, and node sensitivity was quantified through perturbation response scanning (PRS) technology. The drug-protein interaction (DPI) perturbation score was calculated and the drug priority was ranked. Comprehensive sensitivity analysis and literature research were used to screen high-ranking drugs. Preliminary experiments were conducted to screen out drugs for repositioning against lung squamous cell carcinoma cell lines.

[0007] As a preferred embodiment of the perturbation response scanning analysis-based drug discovery method for squamous cell lung cancer described in the present invention, the robust network module includes collecting tumor samples, downloading protein expression profile data for squamous cell lung cancer, and performing preprocessing to delete proteins missing in more than 50% of the tumor samples from the protein expression profile data, and replacing the remaining missing values ​​in the tumor samples with the mean of the expression profile data;

[0008] The preprocessed expression profile data were analyzed for differential expression between tumor samples and adjacent normal samples using the first test value obtained by the Wilcoxon rank sum test. The fold change (FC) of proteins between the cancer group and adjacent normal samples was calculated. Proteins with a first test value < 0.01 and |log2(FC)| > 1 were defined as differentially expressed proteins.

[0009] The differentially expressed proteins were mapped to the PPI network to preliminarily extract the differential network of lung squamous cell carcinoma. The Pearson correlation coefficient between the nodes of each PPI was calculated, and the Pearson correlation coefficient threshold was set. When the absolute value of the Pearson correlation coefficient was greater than the Pearson correlation coefficient threshold, a secondary screening was performed to obtain the final differential network.

[0010] The ne-PCA algorithm was used to divide the difference network into modules and obtain the robust network modules. That is, the hypergeometric test was used to perform the association test on all module pairs derived from the ne-PCA algorithm to obtain the test values. The common parts of the modules with test values ​​< 0.05 were finally identified as stable modules.

[0011] Network modules with at least 10 nodes were selected, and GO and KEGG pathway enrichment analysis was performed on the protein set of each module to evaluate the main biological functions of the module.

[0012] As a preferred embodiment of the lung squamous cell carcinoma drug discovery method based on perturbation response scanning analysis of the present invention, the multi-scale analysis includes kinetic analysis, survival analysis, enrichment analysis and independent cohort test analysis;

[0013] Perform dynamic analysis on the robust network module and perform PRS analysis to obtain the average response degree of the module;

[0014] Combining proteomic expression and clinical data, survival analysis was performed using the survminer and survival packages in R language to comprehensively evaluate the prognostic value of each module;

[0015] The GSEA enrichment analysis method was used to evaluate the activity status of cancer-related hallmarks in each module and obtain the status value. Entries with a status value < 0.05 were considered significantly enriched and marked as target modules.

[0016] For the target module, independent cohort testing analysis was performed in the proteomic and transcriptomic cohorts, and ridge regression was used to calculate the correlation between each protein and sample, extract risk features and calculate risk scores to distinguish tumors from adjacent tumors.

[0017] As a preferred embodiment of the lung squamous cell carcinoma drug discovery method based on perturbation response scanning analysis of the present invention, generating a DRN includes, based on large-scale cancer drug genome data, using a machine learning algorithm to predict the clinical drug response of tumor patients for the drug's target module, and generating a patient-drug matrix;

[0018] Patients were divided into two groups, high expression and low expression, based on protein expression information. The second test value was obtained by comparing the difference in drug sensitivity scores between the two groups using the Wilcoxon rank sum test.

[0019] If there is a difference in the sensitivity score between the groups of any drug, that is, the second test value is <0.05, it is considered that there is an interaction between the current drug and the query protein, and the current interaction is integrated into the PPI network as a new node and edge to generate a DRN.

[0020] As a preferred embodiment of the lung squamous cell carcinoma drug discovery method based on perturbation response scanning analysis described in the present invention, the drug priority sorting includes: drug sorting based on PRS, using deep learning tools to predict the binding affinity between drugs and proteins and use it as the weight of the edge in DRN, selecting message passing neural network and convolutional neural network as molecular encoders, and embedding the learned features into a multilayer perceptron decoder to generate the final affinity prediction

[0021]

[0022] in, represents the specific affinity of the drug for protein x, and T is the transposed matrix.

[0023] As a preferred embodiment of the lung squamous cell carcinoma drug discovery method based on perturbation response scanning analysis described in the present invention, the drug priority ranking further includes introducing PRS into the PPI network of DRN to measure the sensitivity of the nodes, converting the PPI network into a Laplacian matrix and performing eigenvalue decomposition to derive the covariance matrix, and normalizing each row of the covariance matrix by the diagonal elements to obtain an asymmetric PRS matrix S PRS , where the rows and columns correspond to the perturbation and response nodes, respectively, and the column average represents the sensitivity of each node;

[0024] The perturbation effect of drug binding to the interacting protein is calculated by performing matrix multiplication of the force vector and the PRS matrix:

[0025]

[0026] Among them, S PRS is the identity matrix that applies a consistent force to every node;

[0027] The perturbation score ps of a drug is defined as the average perturbation of all nodes in the network:

[0028] ps=μ(ΔR DPI )

[0029] Performing data transformation and simplification on the average disturbance yields:

[0030]

[0031] Among them, r pk =[p 1k p 2k …p nk ] T It's S PRS The k-th displacement vector, DPI is sorted in descending order according to ps, p 1k p2k …p nk represents the disturbance at each node;

[0032] The final perturbation score of drugs acting on multiple targets is defined as the cumulative score of all DPI normalized scores for drug ranking.

[0033] As a preferred embodiment of the lung squamous cell carcinoma drug discovery method based on perturbation response scanning analysis of the present invention, the repositioning includes obtaining the top 100 drugs according to drug ranking, and screening candidate sensitive drugs in combination with drug sensitivity analysis based on AUC;

[0034] The candidate sensitive drugs screened in the first screening are narrowed down based on database search and literature research;

[0035] From the narrowed-down drug range, drugs that showed anti-tumor activity against lung squamous cell carcinoma cell lines were screened through preliminary experiments;

[0036] The cell lines were treated with different concentration gradients of the drug exhibiting anti-tumor activity, and cell viability was determined by cell proliferation assay;

[0037] At different drug concentrations, the effects of drugs on cell migration and invasion were determined by wound healing assay and Transwell assay.

[0038] Another object of the present invention is to provide a lung squamous cell carcinoma drug discovery system based on perturbation response scanning analysis.

[0039] To solve the above technical problems, the present invention provides the following technical solutions: a lung squamous cell carcinoma drug discovery system based on perturbation response scanning analysis, comprising a data preprocessing module, a difference network construction module, a multi-scale analysis module, a drug response prediction module, and a drug repositioning screening module;

[0040] The data preprocessing module is responsible for downloading and preprocessing the protein expression profile data of lung squamous cell carcinoma;

[0041] The differential network construction module uses the preprocessed expression profile data to perform differential expression analysis between tumor samples and adjacent adjacent samples through the Wilcoxon rank sum test, identifies differentially expressed proteins and maps them to the PPI network, constructs a differential network, and uses the ne-PCA algorithm to divide the differential network into modules to generate a robust network module;

[0042] The multi-scale analysis module performs multi-scale analysis on the robust network module, evaluates the prognostic value of the module, and identifies the target module;

[0043] The drug response prediction module uses a machine learning algorithm to analyze the drug target module data, generate a patient-drug matrix, evaluate the drug sensitivity of the patient, compare the sensitivity scores of different groups through the Wilcoxon test, and generate a DRN;

[0044] The drug repositioning screening module screens candidate drugs based on drug ranking and combines AUC-based drug sensitivity analysis, conducts systematic search and literature research to narrow the drug range, and evaluates the anti-tumor activity of drugs against lung squamous cell carcinoma cell lines through preliminary experiments to confirm the effectiveness of the drugs.

[0045] The present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, the steps of the method for drug discovery for lung squamous cell carcinoma based on perturbation response scanning analysis are implemented.

[0046] The present invention provides 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 for drug discovery for lung squamous cell carcinoma based on perturbation response scanning analysis are implemented.

[0047] The invention's beneficial effects: By integrating proteomics, pharmacogenomics, and dynamic network analysis, the framework developed by the invention provides a systematic analysis for identifying repositionable drugs for squamous cell lung cancer, bringing new benefits to its treatment. The repositionable drugs disclosed in the invention can be used in the treatment of patients with squamous cell lung cancer. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive work, among which:

[0049] Figure 1 This is an overall flow chart of a method for drug discovery for squamous cell lung cancer based on perturbation response scanning analysis provided by one embodiment of the present invention.

[0050] Figure 2 This invention provides a method for drug discovery for squamous cell lung cancer based on perturbation response scanning analysis, and verifies the therapeutic effect of repositioned drugs in an embodiment of the invention. DETAILED DESCRIPTION

[0051] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0052] Example 1, with reference to Figure 1-Figure 2 , which is the first embodiment of the present invention, provides a lung squamous cell carcinoma drug discovery method based on perturbation response scanning analysis, comprising:

[0053] S1: Download protein expression profile data of lung squamous cell carcinoma and perform preprocessing. Differential proteins are obtained through differential expression analysis and mapped to the PPI network to construct a differential network. The differential network is divided into modules to obtain a robust network module.

[0054] S101: Download the protein expression profile of lung squamous cell carcinoma from the Clinical Proteomic Tumor Analysis Consortium (CPTAC) database.

[0055] The cohort collected data from patients with squamous cell lung cancer in seven different countries and characterized 108 tumor samples from primary cancer patients and 99 paired adjacent adjacent tumor samples.

[0056] For expression profile data, proteins missing in more than 50% of the samples were deleted, and the remaining missing values ​​were replaced by the mean.

[0057] S102: Using the Wilcoxon rank sum test to obtain the first test value, differential expression analysis was performed on the tumor samples and adjacent normal samples; at the same time, the differential expression fold (FC) of the protein between the cancer group and the adjacent normal group was calculated.

[0058] Among them, proteins with the first test value < 0.01 and |log2(FC)| > 1 were defined as differentially expressed proteins.

[0059] S103: Download a high-quality physical protein-protein interaction (PPI) network of humans. The network contains 387,626 edges between 17,660 proteins involved in regulatory, metabolic, kinase substrate, signaling, or binding relationships.

[0060] The differentially expressed proteins were mapped to the network, and the differential network of lung squamous cell carcinoma was preliminarily extracted.

[0061] Furthermore, the Pearson correlation coefficient (r) between nodes X and Y of each PPI was calculated based on the protein expression profile and was defined as:

[0062]

[0063] Where n is the number of samples, X i and Y i is the expression level in the i-th sample, and is the mean value. The range of r is between -1 and 1, and the larger its absolute value, the stronger the correlation.

[0064] The present invention screens PPIs with greater biological significance under the condition of |r|>0.3, thereby generating the final difference network.

[0065] S104: In the network, the difference network is divided into modules using the ne-PCA algorithm to obtain a robust network module;

[0066] Specifically, this method integrates two complementary algorithms to identify network modules, namely the top-down Girvan-Newman (GN) algorithm starting from the network level and the bottom-up Label Propagation Algorithm (LPA) starting from the protein level.

[0067] The GN algorithm completes module partitioning by iteratively deleting the edge with the highest betweenness in the network until the modularity is maximized; while the LPA gradually updates the label value of the node according to the adjacent nodes, so that similar nodes are divided into the same cluster.

[0068] Furthermore, the hypergeometric test was used to perform an association test on all module pairs derived from GN and LPA to obtain the test value, and the common parts of the module pairs with the test value < 0.05 were finally identified as stable modules.

[0069] S105: Select network modules with at least 10 nodes and assign unique terms to each module through Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis to identify functional modules and evaluate the main biological functions of the modules.

[0070] S2: Evaluate the importance of robust network modules through multi-scale analysis, screen target modules and verify generalization ability.

[0071] It should be noted that the multi-scale analysis includes kinetic analysis, survival analysis, enrichment analysis and independent cohort test analysis.

[0072] Specifically,

[0073] S201: Perturbation response scanning (PRS), rooted in linear response theory, can evaluate the network dynamics of node responses to external forces with the help of elastic network models (ENM).

[0074] A PRS analysis was performed for each module to obtain the average response degree of the module.

[0075] S202: Combining proteomic expression and clinical data, survival analysis was performed using the survminer and survival packages in R language.

[0076] S2021: Use the surv_cutpoint() function to calculate the optimal threshold of protein expression. The optimal threshold can maximize the survival differences between patients, rather than simply relying on the median or mean.

[0077] S2022: Patients were divided into high expression group and low expression group according to the optimal threshold, and the survival curves between the two groups were compared by Log-Rank test to obtain the difference value;

[0078] The difference value <0.05 was used as the significance threshold to screen out proteins that were significantly associated with the patient's overall survival;

[0079] Proteins significantly associated with patients' overall survival were divided into two categories, namely unfavorable proteins and favorable proteins: among them, unfavorable proteins were defined as proteins whose high expression levels were significantly associated with patients' poor survival outcomes, while high expression of favorable proteins indicated a better prognosis for patients.

[0080] S2023: By integrating the dysregulation status of proteins and their correlation with survival, we define the types of proteins that have a positive, negative, or no significant impact on patient survival, enabling a comprehensive assessment of the prognostic value of each module.

[0081] S203: For functional modules, the gene set enrichment analysis (GSEA) method was used to evaluate the active status of cancer-related Hallmarks in each module and obtain the status value. Among them, entries with a status value <0.05 were considered to be significantly enriched and marked as target modules.

[0082] S204: Based on the above analysis, the results show that the extracellular matrix module is of great significance for the progression of lung squamous cell carcinoma.

[0083] Independent cohort testing analysis was performed for the target module: S2041: For the selection of risk features in the module, ridge regression was used to calculate the correlation between each protein and tumor and adjacent paracancerous samples;

[0084] S2042: The protein expression data were divided into a training cohort and a test cohort in a 1:1 ratio. The model was built on the training cohort using the glmfit() function of the R language glmnet package (α = 0). In the results, proteins corresponding to non-zero coefficients were considered risk signatures.

[0085] S2043: At the same time, the Firth correction method of the logistf package was used to calculate the odds ratio (OR) of the risk signature in the training cohort.

[0086] S2044: Generate a risk score (Riskscore) in the test cohort and the additional independent cohort, namely:

[0087]

[0088] Where n is the total number of proteins belonging to the risk signature, z i is the normalized Z-score expression value of protein i, β i is the logarithm of the OR for protein i.

[0089] S2045: The difference in risk scores between tumors and adjacent normal tissues was compared by Wilcoxon test, and the ability of the score to distinguish between tumors and adjacent normal tissues was evaluated by receiver operating characteristic (ROC) curve analysis, which was implemented by the R package pROC.

[0090] S3: Based on drug genomic data, machine learning is used to predict patient drug sensitivity and generate DRN through differential analysis of drug sensitivity.

[0091] Furthermore, based on large-scale cancer drug genomic data, machine learning algorithms are used to predict the clinical drug response of cancer patients based on the drug's target module. Specifically, protein expression data in tumor tissue is used as input to oncoPredict, and its built-in training dataset is used to train the model to obtain a patient-drug matrix representing sensitivity scores.

[0092] The patients were divided into high expression group and low expression group by the median expression value of each protein in the module, and the Wilcoxon rank sum test was used to analyze the difference in drug sensitivity scores between the two groups to obtain the second test value;

[0093] If there is a difference in the sensitivity score between the groups of any drug (second test value < 0.05), it is considered that there is an interaction between the current drug and the query protein, and the current interaction is integrated into the PPI network as a new node and edge to generate a DRN.

[0094] S4: Combine deep learning to predict drug-target protein binding affinity, quantify node sensitivity through PRS technology, calculate drug-protein interaction (DPI) perturbation scores and prioritize drugs.

[0095] S401: Use the deep learning tool DeepPurpose to predict the binding affinity between drugs and proteins and regard it as the weight of the edge in DRN.

[0096] Specifically, drugs and proteins are converted into simplified molecular linear input canonical symbols and amino acid sequences through PubChem and Uniprot databases, respectively, as inputs of DeepPurpose.

[0097] Message passing neural networks and convolutional neural networks are selected as molecular encoders, and the learned features are embedded into a multilayer perceptron decoder to generate the final affinity prediction, where the affinity of the drug to its interacting protein is represented as a force vector, namely:

[0098]

[0099] in, represents the specific affinity of the drug for protein x.

[0100] S402: Apply the PRS policy to the PPI network of the DRN.

[0101] The structure of the PPI network is represented by an ENM, where the nodes are proteins and the edges are interactions between proteins modeled as springs. This allows the application of forces (e.g. ), thereby generating a perturbation effect that propagates throughout the network and allows quantitative measurement of the coordinated responses of all other nodes.

[0102] Specifically, the PPI network is converted into a Laplacian matrix and eigenvalue decomposition is performed to derive the covariance matrix.

[0103] Each row of the covariance matrix is ​​normalized by the diagonal elements, resulting in an asymmetric PRS matrix S PRS , where rows and columns correspond to perturbation and response nodes, respectively, and the column mean represents the sensitivity of each node.

[0104] The perturbation effect on the binding of a drug to its interacting protein is calculated by performing matrix multiplication of the force vector and the PRS matrix:

[0105]

[0106] Among them, S PRS is the identity matrix that applies a uniform force to each node, computed by the Python package ProDy.

[0107] The perturbation score ps of a drug is defined as the average perturbation of all nodes in the network, thus capturing the comprehensive impact of DPI on the entire DRN:

[0108] ps=μ(ΔR DPI )

[0109] Performing data transformation and simplification on the average disturbance yields:

[0110]

[0111] Among them, r pk =[p 1k p 2k …p nk ] T It's S PRS The k-th displacement vector, DPI is sorted in descending order according to ps, p 1k p 2k …p nk represents the disturbance at each node;

[0112] DPIs are sorted in descending order according to their ps, and the final perturbation score of a drug acting on multiple targets is defined as the cumulative score of all its DPI normalized scores.

[0113] S5: Comprehensive sensitivity analysis and literature research were used to screen high-ranking drugs; drugs for lung squamous cell carcinoma cell lines were screened through preliminary experiments for repositioning.

[0114] S501: For the top 100 drugs mentioned above, further screen candidate sensitive drugs by combined drug sensitivity analysis.

[0115] The CTRP drug activity data and corresponding gene expression levels of lung squamous cell carcinoma cell lines were downloaded from the CellMinerCDB database, and drug sensitivity analysis based on AUC was performed.

[0116] S5011: Drugs with missing activity scores in more than 50% of the cell lines and cell lines with missing activity scores in more than 80% of the drugs were removed, and the remaining drug activity data were imputed using the R package impute.

[0117] S5012: Cell lines were stratified according to the median activity score, and ROC analysis was performed to calculate the AUC based on the target gene expression.

[0118] S5013: Simultaneously, a univariate linear regression model based on target gene expression levels was constructed to predict drug activity scores, and the overall significance of the model was assessed using analysis of variance. A gene was considered significantly associated with the sensitivity of a cell line to a drug when the overall significance was <0.05 and the AUC was ≥0.75; a gene was considered significantly associated with drug resistance when the overall significance was <0.05 and the AUC was ≤0.25.

[0119] S502: For sensitive drugs, collect their indications, mechanisms of action and their correlation with lung squamous cell carcinoma through database retrieval and literature research.

[0120] Further cell model experiments will be conducted to verify drugs that are not currently used in the treatment of squamous cell lung cancer, clinical trials, or biological experiments.

[0121] Preliminary experimental results showed that piperlongumamide and fingolimod exhibited significant antitumor activity against lung squamous cell carcinoma cell lines H226 and H1703. Piperlongumamide is a natural product that has been characterized as an inducer of reactive oxygen species. This compound not only has multiple biological activities such as anti-inflammatory, antibacterial, anti-cardiovascular disease and anti-diabetes, but also can inhibit cell proliferation and induce cell apoptosis in cell lines such as pancreatic cancer, lung cancer, kidney cancer and breast cancer. Fingolimod is an immunomodulatory drug that, as a modulator of sphingosine-1-phosphate (S1P) receptors, has been approved by the FDA for patients with relapsing-remitting multiple sclerosis and is being studied for the treatment of pulmonary complications of pneumonia caused by the new coronavirus, asthma or stroke, etc.

[0122] S503: Further, different concentration gradients of drugs were used to treat lung squamous cell carcinoma cell lines, and the half-maximal inhibitory concentration (IC50) was determined. The following experiments were also performed:

[0123] S5031: Cell Proliferation Assay: Cell proliferation was measured using the Cell Counting Kit-8 (CCK-8) assay. H1703 and H226 stable cells were seeded in 96-well plates at a density of 2 × 10 cells per well and cultured under normal conditions for 24, 48, and 72 hours. Cell viability was determined according to the manufacturer's instructions. Experiments were performed in triplicate.

[0124] S5032: Wound healing assay: H1703 and H226 cells were seeded into 6-well cell culture plates and cultured until the cells reached 70-80% monolayer confluence. A fresh 10 μl pipette tip was used to gently and slowly scratch the center of the well, aiming to create a gap with a width equal to the outer diameter of the pipette tip. Another vertical scratch was also created in each well to form a cross shape. Then, the cells were gently washed twice with 1× PBS (phosphate buffer saline) to remove detached cells. Fresh culture medium was added to the wells and cultured for another 24 hours. Cell images were observed and photographed using a microscope at the same magnification and settings. The width of the scratches was quantitatively evaluated by Photoshop.

[0125] S5033: Transwell Assay: Cell migration and invasion assays were performed using 8.0 μm pore size Transwell chambers. For migration assays, 800 μl of RPMI-1640 medium supplemented with 10% fetal bovine serum was added to the bottom layer of each Transwell chamber.

[0126] The stable cells were trypsinized, and then 5×10 4 cells were suspended in a culture medium containing 1% FBS, seeded into the upper chamber, and incubated in a humidified incubator at 37° C. for 24 h.

[0127] Cells that migrated to the lower surface of the chamber were fixed with 100% methanol for 20 min, air-dried for 10 min, stained with 0.1% crystal violet overnight, and washed twice with 1× PBS.

[0128] Cells were photographed and counted. For invasion assays, chambers were first coated with Matrigel matrix (BD Science, Sparks, MD, USA) diluted in serum-free medium and then incubated at 37°C for 2 h. The remaining steps were similar to those for migration assays. Each experiment was repeated three times.

[0129] The experimental results showed that both piperlongum amide and fingolimod significantly reduced the viability of lung squamous cell carcinoma cells in a dose-dependent manner ( Figure 2 a and b). Among them, after 48h treatment, the IC50 of piperlongumamide on H226 and H1703 cells were 5.959μM and 3.241μM, respectively, while the IC50 of fingolimod on H226 and H1703 cells were 8.001μM and 10.4μM, respectively. Subsequently, in order to further verify the effects of the two compounds on the migration and invasion ability of lung squamous cell carcinoma cells, the present invention conducted wound healing experiments and Transwell experiments. The results showed that both piperlongumamide and fingolimod treatment significantly inhibited the migration and invasion ability of H226 and H1703 cells in a dose-dependent manner ( Figure 2cf). These results suggest that both compounds have antitumor potential, with piperlongumamide potentially exhibiting a stronger therapeutic effect.

[0130] Embodiment 2 The second embodiment of the present invention is different from the previous embodiment in that:

[0131] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0132] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0133] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0134] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0135] Example 3 is the third embodiment of the present invention, which provides a lung squamous cell carcinoma drug discovery system based on perturbation response scanning analysis, including a data preprocessing module, a difference network construction module, a multi-scale analysis module, a drug response prediction module, and a drug repositioning screening module;

[0136] The data preprocessing module is responsible for downloading and preprocessing the protein expression profile data of lung squamous cell carcinoma;

[0137] The differential network construction module uses the preprocessed expression profile data to perform differential expression analysis between tumor samples and adjacent adjacent samples through the Wilcoxon rank sum test, identifies differentially expressed proteins and maps them to the PPI network, constructs a differential network, and uses the ne-PCA algorithm to divide the differential network into modules to generate a robust network module;

[0138] The multi-scale analysis module performs multi-scale analysis on the robust network module, evaluates the prognostic value of the module, and identifies the target module;

[0139] The drug response prediction module uses a machine learning algorithm to analyze the drug target module data, generate a patient-drug matrix, evaluate the drug sensitivity of the patient, compare the sensitivity scores of different groups through the Wilcoxon test, and generate a DRN;

[0140] The drug repositioning screening module screens candidate drugs based on drug ranking and combines AUC-based drug sensitivity analysis, conducts systematic search and literature research to narrow the drug range, and evaluates the anti-tumor activity of drugs against lung squamous cell carcinoma cell lines through preliminary experiments to confirm the effectiveness of the drugs.

[0141] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A drug discovery method for squamous cell lung cancer based on perturbation response scanning analysis, characterized by: Download protein expression profile data of lung squamous cell carcinoma and perform preprocessing. Differential expression analysis is performed to obtain differential proteins, which are then mapped to the PPI network to construct a differential network. Divide the difference network into modules to obtain robust network modules; Evaluate the importance of robust network modules through multi-scale analysis, screen target modules and verify generalization capabilities; Based on drug genome data, machine learning is used to predict patient drug sensitivity and generate DRN through differential analysis of drug sensitivity; Combine deep learning to predict drug-target protein binding affinity, quantify node sensitivity through PRS technology, calculate DPI perturbation scores and prioritize drugs; Comprehensive sensitivity analysis and literature research were used to screen high-ranking drugs; Preliminary experiments were conducted to screen and reposition drugs targeting lung squamous cell carcinoma cell lines.

2. The method for drug discovery for lung squamous cell carcinoma based on perturbation response scanning analysis according to claim 1, wherein: The robust network module includes collecting tumor samples, downloading protein expression profile data of lung squamous cell carcinoma, and performing preprocessing to delete proteins missing in more than 50% of the tumor samples from the protein expression profile data, and replacing the remaining missing values ​​in the tumor samples with the mean of the expression profile data; The pre-processed expression profile data were analyzed for differential expression between tumor samples and adjacent normal samples using the first test value obtained by the Wilcoxon rank sum test. At the same time, the differential expression fold (FC) of proteins between the cancer group and adjacent normal samples was calculated. Proteins with a first test value < 0.01 and |log2(FC)| > 1 were defined as differentially expressed proteins. The differentially expressed proteins were mapped to the PPI network to preliminarily extract the differential network of lung squamous cell carcinoma. The Pearson correlation coefficient between the nodes of each PPI was calculated, and the Pearson correlation coefficient threshold was set. When the absolute value of the Pearson correlation coefficient was greater than the Pearson correlation coefficient threshold, a secondary screening was performed to obtain the final differential network. The ne-PCA algorithm was used to divide the difference network into modules and obtain robust network modules. That is, the hypergeometric test was used to perform association tests on all module pairs derived from the ne-PCA algorithm to obtain test values. The common parts of the modules with test values ​​< 0.05 were finally identified as stable modules. Network modules with at least 10 nodes were selected, and GO and KEGG pathway enrichment analysis was performed on the protein set of each module to evaluate the main biological functions of the module.

3. The method for drug discovery for lung squamous cell carcinoma based on perturbation response scanning analysis according to claim 2, wherein: The multi-scale analysis includes kinetic analysis, survival analysis, enrichment analysis and independent cohort test analysis; Perform dynamic analysis on the robust network module and perform PRS analysis to obtain the average response degree of the module; Combining proteomic expression and clinical data, survival analysis was performed using the survminer and survival packages in R language to comprehensively evaluate the prognostic value of each module; The GSEA enrichment analysis method was used to evaluate the activity status of cancer-related Hallmarks in each module and obtain the status value. Entries with a status value < 0.05 were considered significantly enriched and marked as target modules. For the target module, independent cohort testing analysis was performed in the proteomic and transcriptomic cohorts, and ridge regression was used to calculate the correlation between each protein and sample, extract risk features and calculate risk scores to distinguish tumors from adjacent tumors.

4. The method for drug discovery for lung squamous cell carcinoma based on perturbation response scanning analysis according to claim 3, wherein: Generating the DRN includes, based on large-scale cancer drug genome data, using a machine learning algorithm to predict the clinical drug response of tumor patients for the drug's target module, and generating a patient-drug matrix; Patients were divided into two groups, high expression and low expression, based on protein expression information, and the difference in drug sensitivity scores between the two groups was compared again by using the Wilcoxon rank sum test to obtain the second test value; If there is a difference in the sensitivity score between the groups of any drug, that is, the second test value is <0.05, it is considered that there is an interaction between the current drug and the query protein, and the current interaction is integrated into the PPI network as a new node and edge to generate a DRN.

5. The method for drug discovery for lung squamous cell carcinoma based on perturbation response scanning analysis according to claim 4, wherein: The drug prioritization process includes: drug ranking based on PRS, using deep learning tools to predict the binding affinity between drugs and proteins and using it as the weight of the edges in DRN, selecting message passing neural networks and convolutional neural networks as molecular encoders, and embedding the learned features into a multi-layer perceptron decoder to generate the final affinity prediction. in, represents the specific affinity of the drug for protein x, and T is the transposed matrix.

6. The method for drug discovery for lung squamous cell carcinoma based on perturbation response scanning analysis according to claim 5, characterized in that: The drug priority ranking further includes introducing PRS into the PPI network of the DRN to measure the sensitivity of the nodes, converting the PPI network into a Laplacian matrix and performing eigenvalue decomposition to derive a covariance matrix, wherein each row of the covariance matrix is ​​normalized by the diagonal elements to obtain an asymmetric PRS matrix S PRS , where the rows and columns correspond to the perturbation and response nodes, respectively, and the column average represents the sensitivity of each node; The perturbation effect of drug binding to the interacting protein is calculated by performing matrix multiplication of the force vector and the PRS matrix: Among them, S PRS is the identity matrix that applies a consistent force to every node; The perturbation score ps of a drug is defined as the average perturbation of all nodes in the network: ps=μ(ΔR DPI ) Performing data transformation and simplification on the average disturbance yields: in, It's S PRS The k-th displacement vector, DPI is sorted in descending order according to ps, p 1k p 2k …p nk represents the disturbance at each node; The final perturbation score of drugs acting on multiple targets is defined as the cumulative score of all DPI normalized scores for drug ranking.

7. The method for drug discovery for lung squamous cell carcinoma based on perturbation response scanning analysis according to claim 6, wherein: The repositioning includes obtaining the top 100 drugs according to the drug ranking, and screening candidate sensitive drugs in combination with drug sensitivity analysis based on AUC; The candidate sensitive drugs screened in the first screening are narrowed down based on database search and literature research; From the narrowed-down drug range, drugs that showed anti-tumor activity against lung squamous cell carcinoma cell lines were screened through preliminary experiments; The cell lines were treated with different concentration gradients of the drug exhibiting anti-tumor activity, and cell viability was determined by cell proliferation assay; At different drug concentrations, the effects of drugs on cell migration and invasion were determined by wound healing assay and Transwell assay.

8. A system for drug discovery for squamous cell lung cancer based on perturbation response scanning analysis, employing the method for drug discovery for squamous cell lung cancer based on perturbation response scanning analysis according to any one of claims 1 to 7, characterized in that: It includes data preprocessing module, difference network construction module, multi-scale analysis module, drug response prediction module, and drug repositioning screening module; The data preprocessing module is responsible for downloading and preprocessing the protein expression profile data of lung squamous cell carcinoma; The differential network construction module uses the preprocessed expression profile data to perform differential expression analysis between tumor samples and adjacent adjacent samples through the Wilcoxon rank sum test, identifies differentially expressed proteins and maps them to the PPI network, constructs a differential network, and uses the ne-PCA algorithm to divide the differential network into modules to generate a robust network module; The multi-scale analysis module performs multi-scale analysis on the robust network module, evaluates the prognostic value of the module, and identifies the target module; The drug response prediction module uses a machine learning algorithm to analyze the drug target module data, generate a patient-drug matrix, evaluate the drug sensitivity of the patient, compare the sensitivity scores of different groups through the Wilcoxon test, and generate a DRN; The drug repositioning screening module screens candidate drugs based on drug ranking and combines AUC-based drug sensitivity analysis, conducts systematic search and literature research to narrow the drug range, and evaluates the anti-tumor activity of drugs against lung squamous cell carcinoma cell lines through preliminary experiments to confirm the effectiveness of the drugs.

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 lung squamous cell carcinoma drug discovery method based on perturbation response scanning analysis 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 lung squamous cell carcinoma drug discovery method based on perturbation response scanning analysis according to any one of claims 1 to 7 are implemented.

Images