Search

CN-121999878-A - Gingko seed toxicity mechanism analysis method based on network toxicology and machine learning and experimental verification

CN121999878ACN 121999878 ACN121999878 ACN 121999878ACN-121999878-A

Abstract

The invention provides a gingko seed toxicity mechanism analysis method based on network toxicology, machine learning and experimental verification, which comprises screening and intersection analysis of gingko seed toxicity component related targets and gastrointestinal toxicity related targets, screening and weighting gene co-expression network analysis (WGCNA) of gastrointestinal toxicity related Differential Expression Genes (DEGs), intersection analysis and function enrichment of key targets, screening and verification of key targets driven by machine learning, functional analysis of core genes, molecular docking and molecular dynamics simulation and experimental verification. Breaks through the limitation of focusing single component or target in the traditional toxicology research, constructs a component-target-disease network by a network toxicology method, captures the complex interaction between toxic components and a biological system, and comprehensively reveals the multi-component and multi-target action characteristics of the toxicity of the ginkgo seeds.

Inventors

  • SONG YAOHUI
  • LI DAN

Assignees

  • 武汉市第八医院(武汉市肛肠医院)

Dates

Publication Date
20260508
Application Date
20251230

Claims (10)

  1. 1. A gingko seed toxicity mechanism analysis method based on network toxicology, machine learning and experimental verification is characterized by comprising the following steps: s1, screening and intersection analysis of a ginkgo seed toxic component related target and a gastrointestinal tract toxicity related target; a. Aiming at the main toxic components of ginkgo seeds, searching potential targets through a plurality of bioinformatics platforms, and obtaining a medicine-target set after standardized treatment and de-duplication; b. taking gastrointestinal toxicity related diseases as keywords, searching disease related targets in a plurality of databases, and obtaining a disease-target set after standardized treatment and merging; c. performing intersection analysis on the medicine-target set and the disease-target set to obtain a common target, constructing a component-target-disease network and a protein interaction (PPI) network, and screening a core gene; S2, screening a gastrointestinal toxicity related Differential Expression Gene (DEGs) and analyzing a weighted gene co-expression network (WGCNA); a. obtaining a gastrointestinal toxicity related public gene expression data set from a GEO database, and screening differential expression genes after standardized treatment and batch effect correction (DEGs); b. Carrying out WGCNA analysis on the corrected gene expression matrix, screening a gene co-expression module which is obviously related to gastrointestinal toxicity, and identifying a core gene; s3, intersection analysis and function enrichment of key targets; a. performing intersection analysis on the common target obtained in the step (S1) and the DEGs and WGCNA core genes screened in the step (S2) to obtain an overlapped gene set; b. Performing GO and KEGG function enrichment analysis on the overlapped gene sets to determine related biological processes and signal paths; s4, screening and verifying key targets driven by machine learning; a. Combining three machine learning algorithms of LASSO regression, random Forest (RF) and support vector machine recursive feature elimination (SVM-RFE), refining overlapping gene sets, and screening core genes; b. the identification capacity of the core gene is verified through violin diagram and ROC curve analysis; c. determining the contribution degree and interaction of the core genes to toxicity prediction by SHAP analysis; s5, functional analysis of the core gene; a. Carrying out Gene Set Enrichment Analysis (GSEA) on the key core genes, and exploring the related paths of the key core genes; b. Performing immune cell infiltration analysis by adopting a CIBERSORT algorithm, and determining the association of a core gene and an immune microenvironment by combining immune function enrichment analysis and immune cell correlation analysis; s6, molecular docking and molecular dynamics simulation; a. Acquiring the main toxic components of ginkgo seeds and the three-dimensional structure of a core target, performing molecular docking analysis, and screening a compound with high binding affinity; b. performing molecular dynamics simulation on the complex with high binding affinity, and verifying the binding stability; S7, experimental verification; a. preparing ginkgo seed drug-containing serum, and identifying toxic components in the drug-containing serum by adopting UHPLC-QExactiveHFX-MS technology; b. Performing in vitro experiments by using gastrointestinal epithelial cell lines, including cell proliferation detection, cytotoxicity detection, oxidative stress detection and apoptosis detection; c. verifying the expression of the core gene by RT-qPCR, westernBlot and immunofluorescence staining; d. Direct interactions of toxic components with core targets were verified using cell thermal displacement analysis (CETSA).
  2. 2. The method for analyzing the toxicity mechanism of ginkgo seeds based on network toxicology, machine learning and experimental verification of claim 1, wherein the ginkgo seeds in the step (S1) mainly comprise Ginkgolic Acid (GAs), 4 '-O-Methyl Pyridoxine (MPN) and 4' -O-methyl pyridoxine-5-glucoside (MPNG), and the bioinformatics platform comprises TCMSP, chEMBL, swissTargetPrediction, pharmMapper and SEA.
  3. 3. The method for analyzing the toxicity mechanism of ginkgo seeds based on network toxicology, machine learning and experimental verification of claim 1, wherein the database in the step (S1) comprises OMIM, geneCards, pharmGKB and TTD, and the gastrointestinal toxicity related diseases comprise Inflammatory Bowel Disease (IBD) and gastritis.
  4. 4. The method for analyzing the toxicity mechanism of ginkgo seeds based on network toxicology, machine learning and experimental verification of claim 1, wherein in the step (S2), the GEO data sets comprise IBD related data sets GSE3365, GSE36807 and gastritis related data sets GSE233973 and GSE5081, the screening threshold of DEGs is a module with a |log2 fold change (log 2 FC) | >0.415 and a P value <0.05, soft threshold power is set to be 5 in WGCNA analysis, a high threshold merging module of 0.25 is adopted, and a module with a Pearson correlation coefficient |r| >0.5 and a P <0.05 is screened.
  5. 5. The method for analyzing the toxicity mechanism of the ginkgo seeds based on network toxicology, machine learning and experimental verification according to claim 1, wherein in the step (S4), the LASSO regression analysis adopts 10-fold cross verification to determine the optimal lambda value, the random forest analysis is set to 500 decision trees, the feature importance is estimated through average reduction of the base index, the SVM-RFE analysis is combined with recursive feature elimination and 10-fold cross verification, the area under the curve (AUC) and the 95% confidence interval are calculated in the ROC curve analysis, and the optimal critical value is determined based on the about-log index.
  6. 6. The method for analyzing the toxicity mechanism of ginkgo seeds based on network toxicology, machine learning and experimental verification according to claim 1, wherein the gastrointestinal epithelial cell line in the step (S7) comprises a human gastric epithelial cell line GES-1 and a colorectal adenocarcinoma cell line Caco-2, the cell proliferation detection adopts a CCK-8 method and an EdU staining method, the cytotoxicity detection adopts an LDH release and total LDH activity detection, the oxidative stress detection adopts a GSH detection and a DCFH-DA fluorescent probe to detect ROS levels, and the apoptosis detection adopts a Hoechst/PI double staining method and an annexin V-FITC/PI double staining flow cytometry.
  7. 7. The method for analyzing the toxicity mechanism of ginkgo seeds based on network toxicology, machine learning and experimental verification of claim 1, wherein the preparation process of the serum containing drugs in the step (S7) is that SD rats are selected and randomly divided into GBS treatment groups and control groups, the GBS treatment groups are administrated by oral gavage according to the dosage of 2000 mg/kg/day for 3 continuous days, the volume of distilled water administrated by the control groups is obtained after the last gavage for 1 hour, serum is collected by centrifugation, and the serum is heat-inactivated and filtered for later use.
  8. 8. The analysis method of the ginkgo seed toxicity mechanism based on network toxicology, machine learning and experimental verification of claim 1, wherein AutoDockVina software is adopted for molecular docking in the step (S6), a conformation with the binding energy less than or equal to-6 kcal/mol is reserved, GROMACS2022 software is adopted for molecular dynamics simulation, simulation parameters are AMBER14SB/GAFF/TIP3P force field, 150mM NaCl, 100psNVT balance +100psNPT balance, 298K, 1bar, LINCS constraint and 2fs step length, and simulation time is 100ns.
  9. 9. The method for analyzing the toxicity mechanism of ginkgo seeds based on network toxicology, machine learning and experimental verification of claim 1, wherein the functional enrichment analysis in the step (S3) adopts an R software package clusterProfiler, the screening threshold is P value <0.05 and the error discovery rate (FDR) is <0.05.
  10. 10. The method for analyzing the toxicity mechanism of ginkgo seeds based on network toxicology, machine learning and experimental verification according to claim 1, wherein in the step (S5), the GSEA analysis adopts a c2.cp.kegg.Hs.symbols.gmt gene set, pvalueCutoff =1 and P.adjustCutoff=0.05, and a pathway with a standardized enrichment score (NES) >1 or < -1 and P <0.05 is defined as a significant enrichment pathway.

Description

Gingko seed toxicity mechanism analysis method based on network toxicology and machine learning and experimental verification Technical Field The invention relates to the technical field of ginkgo seed toxicity mechanism analysis methods, in particular to a ginkgo seed toxicity mechanism analysis method based on network toxicology, machine learning and experimental verification. Background Semen Ginkgo (Ginkgo biloba L.) as one of the oldest gymnosperms is known as "activated stone", and seeds thereof (Chinese medicinal name is "semen Ginkgo") have been used as medicinal and edible resources in east Asia for hundreds of years, and modern pharmacological researches show that semen Ginkgo contains bioactive components such as flavonoids, terpene lactones, polysaccharides and the like, has various pharmacological effects such as antibacterial, neuroprotection, cardiovascular protection and the like, and increasingly clinical and epidemiological evidences show that excessive eating of semen Ginkgo seeds may cause gastrointestinal toxicity, mainly manifested as symptoms such as nausea, vomiting, abdominal pain and the like, and may cause damage to multiple organs in severe cases. The toxicity of ginkgo seeds mainly originates from three components of Ginkgolic Acid (GAs), 4 '-O-Methyl Pyridoxine (MPN) and 4' -O-methyl pyridoxine-5-glucoside (MPNG), wherein ginkgolic acid can generate genetic toxicity and cytotoxicity through inducing DNA damage, cell cycle arrest and mitochondrial dysfunction, MPN is used as a vitamin B6 antagonist, can inhibit glutamate decarboxylase activity and reduce gamma-aminobutyric acid (GABA) level, causes convulsion and gastrointestinal discomfort, MPNG can be hydrolyzed by beta-glucosidase in the gastrointestinal tract to release MPN, enhances overall toxicity, and can cause dynamic interconversion with the MPN through heat treatment. Although researches prove that the eating of ginkgo seeds is related to the gastrointestinal toxicity, the potential toxicology mechanism is still not completely defined, the traditional toxicology researches focus on single components or isolated targets, complex interactions between toxic components of the ginkgo seeds and biological systems are difficult to capture, and the toxic action characteristics of multiple components and multiple targets cannot be comprehensively revealed. Disclosure of Invention The invention provides a gingko seed toxicity mechanism analysis method based on network toxicology, machine learning and experimental verification, which aims to solve the technical problems that traditional toxicology research focuses on single components or isolated targets, complex interaction between gingko seed toxic components and biological systems is difficult to capture, and toxic action characteristics of multiple components and multiple targets cannot be comprehensively disclosed. The invention solves the technical problems by the following technical proposal: The invention provides a gingko seed toxicity mechanism analysis method based on network toxicology, machine learning and experimental verification, which comprises the following steps: s1, screening and intersection analysis of a ginkgo seed toxic component related target and a gastrointestinal tract toxicity related target; a. Aiming at the main toxic components of ginkgo seeds, searching potential targets through a plurality of bioinformatics platforms, and obtaining a medicine-target set after standardized treatment and de-duplication; b. taking gastrointestinal toxicity related diseases as keywords, searching disease related targets in a plurality of databases, and obtaining a disease-target set after standardized treatment and merging; c. performing intersection analysis on the medicine-target set and the disease-target set to obtain a common target, constructing a component-target-disease network and a protein interaction (PPI) network, and screening a core gene; S2, screening a gastrointestinal toxicity related Differential Expression Gene (DEGs) and analyzing a weighted gene co-expression network (WGCNA); a. obtaining a gastrointestinal toxicity related public gene expression data set from a GEO database, and screening differential expression genes after standardized treatment and batch effect correction (DEGs); b. Carrying out WGCNA analysis on the corrected gene expression matrix, screening a gene co-expression module which is obviously related to gastrointestinal toxicity, and identifying a core gene; s3, intersection analysis and function enrichment of key targets; a. performing intersection analysis on the common target obtained in the step (S1) and the DEGs and WGCNA core genes screened in the step (S2) to obtain an overlapped gene set; b. Performing GO and KEGG function enrichment analysis on the overlapped gene sets to determine related biological processes and signal paths; s4, screening and verifying key targets driven by machine learning; a. Combining three machin