Multi-target anticancer drug screening optimization method based on transcriptomics and machine learning

Abstract

The present invention provides a multi-target anticancer drug screening and optimization method based on transcriptomics and machine learning, which relates to the field of biomedicine technology. The present invention first obtains three sets of transcriptome data and uses DESeq2 to perform basic transcriptomic analysis. Then, key biological processes and pathways of prostate cancer are identified through DAVID enrichment analysis. The confirmed data are then submitted to the CMap database, and SuperPred is used to predict the targets of bisoprolol. The dual targets ADRB3 and hERG are confirmed through dynamic simulation. Finally, cell experiments are used to verify the potential effect of bisoprolol in inhibiting cancer cell proliferation. The method provided by the present invention has broad application prospects and can be used to discover therapeutic drugs for other types of cancer.

Description

Technical Field

[0001] The present invention relates to the field of biomedicine technology, and in particular to a multi-target anticancer drug screening and optimization method based on transcriptomics and machine learning. Background Art

[0002] Cancer is a serious disease that poses a significant threat to human health and life. Existing anticancer drugs often act only on a single target or pathway, making it difficult to achieve comprehensive therapeutic effects against complex cancers. Therefore, the development of multi-target anticancer drugs has become a hot topic and a challenge in cancer research.

[0003] With the development of high-throughput sequencing technologies, more and more gene expression data have been accumulated and made publicly available. This data provides strong support for drug discovery based on gene expression profiling. However, how to mine cancer-related genes and pathways from this data and how to use machine learning methods for effective drug screening and optimization remain challenging. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a multi-target anticancer drug screening and optimization method based on transcriptomics and machine learning.

[0005] A multi-target anticancer drug screening and optimization method based on transcriptomics and machine learning, comprising the following steps:

[0006] Step 1: Collect transcriptome sequencing data of a specific type of tumor, and obtain 3-5 groups of tumor tissue samples and normal tissue samples respectively;

[0007] Step 2: Use bioinformatics software to perform differential analysis on transcriptome sequencing data, screen differential genes, and obtain up-regulated and down-regulated genes with a difference degree higher than the set threshold;

[0008] Step 3: Obtain the intersection of up-regulated and down-regulated genes from the screened differentially expressed genes. Use gene analysis and prediction software to intersect the differentially expressed genes and draw a Venn diagram to identify key gene clusters associated with the current cancer disease.

[0009] Step 4: Perform GO and KEGG analysis on key gene clusters to obtain functional and pathway information related to current cancer; specifically: the user selects the latest version of the biological database, uses DAVID 2021, submits the gene names of each key gene cluster to the analysis window of the database, submits the data and selects the categories of cellular component CC, biological process BP and molecular function MF, as well as the Kyoto Encyclopedia of Genes and Genomes pathway KEGG for analysis, and sets the threshold for selecting key functions and pathways to P value <0.05 and counts ≥2.

[0010] Step 5: Input the differentially expressed genes into the CMap database for screening;

[0011] Specifically: upload the differentially expressed genes to the database analysis window, with up-regulated genes and down-regulated genes uploaded separately, and no less than 15 of each; determine the query parameters for "Gene Expression L1000" and "Individual Query"; select the same cell line as the cancer being studied; and analyze to obtain a list of gene expression changes caused by a small molecule interfering with the cells. Each row in the gene expression change list represents a small molecule, indicating the changes in its genes after the small molecule interferes with the selected cell line, and contains a score from -100 to 100. A score of 100 indicates that the effect of the small molecule is exactly the same as the up- and down-regulation of the differentially expressed genes provided, and a score of -100 indicates the exact opposite. Small molecules with scores below the threshold are deemed to have interference effects on cells that are opposite to the up- and down-regulation of genes caused by cancer, and the higher the possibility of their anti-cancer effects.

[0012] Step 6: Combine the analysis results of step 4 and the small molecule scores in step 5 to select candidate drugs;

[0013] Step 7: Conduct biological experiments including but not limited to inhibiting the proliferation and migration of cancer cells to verify the candidate drug to determine its effectiveness.

[0014] Step 8: Use target prediction tools to perform machine learning-based target prediction on effective drugs to obtain possible drug targets;

[0015] Step 9: Use computational biology methods to verify possible targets, use molecular docking software to perform molecular docking, and observe the stable connection relationship between small molecules and targets.

[0016] The beneficial effects of adopting the above technical solution are:

[0017] This invention provides a multi-target anticancer drug screening optimization method based on transcriptomics and machine learning. This method can effectively identify anticancer drugs and targets with potential therapeutic effects through comprehensive analysis of transcriptomic data, drug screening, and biological experimental verification. Specifically, it has the following beneficial effects:

[0018] 1. High efficiency: This method can quickly and accurately screen differentially expressed genes associated with specific cancer diseases in large-scale genomic data, thereby enabling efficient drug screening and development.

[0019] 2. Multi-target screening: This method takes into account the expression changes of multiple genes and can consider multiple targets simultaneously in drug screening, thereby improving drug efficacy and reducing side effects.

[0020] 3. Reliability: This method verifies the screened drugs through multiple steps, including biological experiments and molecular simulations, which can effectively reduce the probability of false positives and improve the reliability of drug development.

[0021] 4. Data richness: This method uses multiple sets of paired tumor tissue samples and normal tissue samples for transcriptome sequencing, and combines them with external databases for drug screening and target prediction, which can make full use of rich data resources to improve the success rate of drug development.

[0022] 5. Reproducibility: This method uses common bioinformatics software and online tools in each step, making it easy for other operators to replicate and reproduce. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 Flowchart of the multi-target cancer drug screening optimization method in an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The following embodiments of the present invention are described in further detail with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention but are not intended to limit the scope of the present invention.

[0025] A multi-target anticancer drug screening optimization method based on transcriptomics and machine learning, such as Figure 1 As shown, the following steps are included:

[0026] Step 1: Collect transcriptome sequencing data of a specific type of tumor and obtain 3-5 groups of tumor tissue samples and normal tissue samples respectively; these samples are used for subsequent differential analysis to determine the upregulated and downregulated differentially expressed genes in each group of data.

[0027] In this example, we chose to download the gene expression dataset from the GEO database, and it is recommended to use the map based on GPL570 (Affymetrix Human Genome U133 Plus 2.0 array);

[0028] Step 2: Use bioinformatics software such as DESeq2 or edgeR to perform differential analysis on the transcriptome sequencing data, screen for differentially expressed genes, and obtain up-regulated and down-regulated genes with a difference greater than a set threshold. In this example, DESeq2 was used for differential analysis, with a threshold set at logFC > 1 and P value < 0.05.

[0029] Step 3: Obtain the intersection of up-regulated and down-regulated genes from the screened differentially expressed genes. Use the gene analysis and prediction software FunRich to intersect the differentially expressed genes and draw a Venn diagram to identify key gene clusters related to the current cancer disease.

[0030] Step 4: Perform GO and KEGG analysis on key gene clusters to obtain information on cancer-related functions and pathways. Specifically, the user selects the latest version of a biological database and, using DAVID 2021, submits the gene names of each key gene cluster to the database's analysis window. After submitting the data, select the categories of Cellular Component (CC), Biological Process (BP), and Molecular Function (MF), as well as the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway for analysis to further elucidate the potential functions represented by the key gene clusters. The threshold for selecting key functions and pathways is set at a P value < 0.05 and a count of ≥ 2.

[0031] Step 5: In this example, differentially expressed genes were entered into the CMap database, a gene expression profile database developed by the Broad Institute based on the CLUE website, for screening to reveal functional relationships between small molecule compounds, genes, and disease states. Specifically, the differentially expressed genes were uploaded to the database analysis window, with both up-regulated and down-regulated genes uploaded separately, with no fewer than 15 genes in each category. The query parameters for "Gene Expression L1000" and "Individual Query" were determined. The same cell line used in the cancer study was selected, and a list of gene expression changes in the cells affected by the small molecule interference was generated. Each row in the gene expression change list represented a small molecule, indicating the gene changes observed after the small molecule interfered with the selected cell line. The list contained a score ranging from -100 to 100, with a score of 100 indicating that the small molecule had the exact same effect as the up- or down-regulated profile of the provided differentially expressed gene, while a score of -100 indicated the exact opposite. Small molecules with scores below the threshold were considered to have an interference effect on the cells that was opposite to the gene up- or down-regulated profile caused by the cancer, and were therefore more likely to have anti-cancer effects.

[0032] The threshold for selecting low-scoring small molecules set in this example is score<-95, thereby obtaining a candidate drug list consisting of many small molecules.

[0033] Step 6: Combine the analysis results of step 4 and the small molecule scores in step 5 to select candidate drugs;

[0034] In this example, the scores in the candidate drug list were combined with the KEGG analysis results to obtain candidate drugs. The pathways regulated by the candidate drugs highly overlapped with the KEGG analysis results.

[0035] Step 7: Conduct biological experiments including but not limited to inhibiting the proliferation and migration of cancer cells to verify the candidate drug to determine its effectiveness.

[0036] According to the needs of the experimenter, verification methods such as MCC experiment and CCK-8 experiment can be selected to evaluate the effect of drugs on cancer cell proliferation.

[0037] Step 8: Use target prediction tools to perform machine learning-based target prediction on effective drugs to obtain possible drug targets;

[0038] In this example, target prediction tools such as SwissTargetPrediction and SuperPred are used. For example, drug candidates are submitted to SuperPred's online service to predict potential binding targets. In this example, the target prediction model is selected with a threshold accuracy of 90%, ensuring that targets with a ligand binding probability greater than 80% are considered candidate targets.

[0039] Step 9: Use computational biology methods to verify possible targets, use molecular docking software to perform molecular docking, and observe the stable connection relationship between small molecules and targets.

[0040] In this example, molecular docking software AutoDock was used for molecular docking, and Amber was used for dynamic simulation of the complex to observe whether there was a stable connection between the small molecule and the target;

[0041] The embodiment of the present invention takes prostate cancer as the research object. First, three sets of transcriptome data were obtained, and basic transcriptomic analysis was performed using DESeq2. Then, key biological processes and pathways of prostate cancer were identified through DAVID enrichment analysis. Next, we submitted the confirmed data to the CMap database and used SuperPred to predict the targets of bisoprolol, and confirmed the dual targets ADRB3 and hERG through dynamic simulation. Finally, cell experiments were used to verify the potential effect of bisoprolol in inhibiting cancer cell proliferation. The results of this study show that bisoprolol is a promising prostate cancer drug with multi-target effects and potential anti-cancer effects. In addition, this research method is expected to discover potential cancer therapeutic drugs by utilizing the concept of drug side effects that lead to anti-cancer effects. Therefore, the method provided by the present invention has broad application prospects and can be used to discover therapeutic drugs for other types of cancer.

[0042] Using the method provided by the present invention, we successfully discovered bisoprolol, a multi-target drug for prostate cancer, and preliminarily verified its anticancer efficacy. This method, which leverages changes in gene expression profiles and machine learning, boasts high throughput and accuracy. Furthermore, comprehensive data analysis using multiple methods, including enrichment analysis and target prediction, enhanced the accuracy and reliability of the discovery of potential anticancer drugs. Therefore, the method provided by the present invention has broad application prospects and market potential.

[0043] The above description is merely a preferred embodiment of the present disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also encompass other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the above-mentioned inventive concept. For example, a technical solution formed by mutually replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in the embodiments of the present disclosure.

Claims

1. A multi-target anticancer drug screening and optimization method based on transcriptomics and machine learning, characterized in that: The following steps are involved: Step 1: Collect transcriptome sequencing data of a specific type of tumor and obtain tumor tissue samples and normal tissue samples respectively; Step 2: Use bioinformatics software to perform differential analysis on transcriptome sequencing data, screen differential genes, and obtain up-regulated and down-regulated genes with a difference degree higher than the set threshold; Step 3: Obtain the intersection of up-regulated and down-regulated genes from the screened differentially expressed genes. Use gene analysis and prediction software to intersect the differentially expressed genes and draw a Venn diagram to identify key gene clusters associated with the current cancer disease. Step 4: Perform GO and KEGG analysis on key gene clusters to obtain functional and pathway information related to current cancer; The user selects the latest version of the biological database, submits the names of each gene of the key gene cluster to the analysis window of the database, submits the data and selects the categories of cellular component CC, biological process BP and molecular function MF as well as Kyoto Encyclopedia of Genes and Genomes pathway KEGG for analysis; Step 5: Input the differentially expressed genes into the CMap database for screening to obtain small molecule scores; Step 6: Combine the function and pathway information obtained in step 4 and the small molecule score in step 5 to select candidate drugs; Step 7: Conduct biological experiments to verify the effectiveness of candidate drugs; Step 8: Use target prediction tools to perform machine learning-based target prediction on effective drugs to obtain possible drug targets; Step 9: Use computational biology methods to verify possible targets, use molecular docking software to perform molecular docking, and observe the stable connection relationship between small molecules and targets.

2. The multi-target anticancer drug screening optimization method based on transcriptome and machine learning according to claim 1, characterized in that: The biological database is DAVID 2021.

3. The multi-target anticancer drug screening optimization method based on transcriptome and machine learning according to claim 1, characterized in that: The analysis described in step 4 specifically set the threshold for selecting key functions and pathways to be P value < 0.05 and counts ≥ 2.

4. The multi-target anticancer drug screening optimization method based on transcriptome and machine learning according to claim 1, characterized in that: The step 5 specifically includes: uploading the differentially expressed genes to the database analysis window, wherein the up-regulated genes and down-regulated genes are uploaded separately, and the number of each is no less than 15, determining the query parameters of "Gene Expression L1000" and "Individual Query", and selecting the same cell line as the cancer studied, and analyzing to obtain a list of gene expression changes in cells interfered with by a small molecule.

5. The multi-target anticancer drug screening optimization method based on transcriptome and machine learning according to claim 4, characterized in that: Each row in the gene expression change list represents a small molecule, indicating the changes in the genes of the selected cell line after the small molecule interferes with it, and contains a score from -100 to 100. A score of 100 means that the effect of the small molecule is exactly the same as the up- or down-regulation of the provided differentially expressed genes, and a score of -100 means it is completely opposite. Small molecules with scores below the threshold are deemed to have interference effects on cells that are opposite to the up- or down-regulation of genes caused by cancer, and are more likely to have anti-cancer effects.

6. The multi-target anticancer drug screening optimization method based on transcriptome and machine learning according to claim 1, characterized in that: The biological experiments described in step 7 include experiments on inhibiting the proliferation and migration of cancer cells.

Images