| 徐子皓,刘逸藩,徐俊,等.整合机器学习与网络毒理学预测栀子在原发性胆汁性胆管炎中的潜在毒性作用机制及关键靶点[J].安徽医药,待发表. |
| 整合机器学习与网络毒理学预测栀子在原发性胆汁性胆管炎中的潜在毒性作用机制及关键靶点 |
|
| 投稿时间：2025-10-04 录用日期：2025-10-29 |
| DOI： |
| 中文关键词: 栀子 原发性胆汁性胆管炎 网络毒理学 机器学习 SHAP 分子对接 |
| 英文关键词: |
| 基金项目:国家自然科学(项目编号:82274367) |
|
| 摘要点击次数: 81 |
| 全文下载次数: 0 |
| 中文摘要: |
| 目的：本研究采用网络毒理学，生物信息学和机器学习相结合的方法，系统预测了栀子对原发性胆汁性胆管炎（PBC）潜在毒性作用的分子机制，并筛选出潜在的关键靶点。方法：通过GeneCards（https://www.genecards.org/）和OMIM（https://omim.org/）数据库检索PBC相关靶点，并从GEO（https://www.ncbi.nlm.nih.gov/geo/）数据库获取PBC转录组数据，通过TCMSP数据库（https://www.tcmsp-e.com/tcmsp.php）筛选栀子的主要化学成分，并使用PubChem数据库（https://pubchem.ncbi.nlm.nih.gov/）获取其SMILES结构式，结合ADMETlab V3.0数据库（https://admetlab3.scbdd.com/）分析活性成分的毒理信息。结果：最终预测出15个活性成分和13个毒性成分，构建了药物毒性成分与PBC靶点的交集网络，筛选出54个交集靶点。基于GSE119600数据集，筛选出9个在正常与PBC样本间的表达存在显著差异的交集靶点。通过多种机器学习算法筛选出潜在核心靶点FRAP1、ABCB1、NLRP3和PIK3CG，并运用SHAP解释了模型的有效性。分子对接分析结果表明，栀子的主要成分与核心靶点之间具有良好的结合亲和力。此外，GO和KEGG分析显示靶点显著参与多个生物过程和信号通路，如IL-17信号通路和氧化应激反应。结论：本研究系统预测了栀子在PBC中的潜在毒理作用机制，提出了一个有待实验验证的“多成分-多靶点”毒性假说，为后续机制研究与临床安全用药提供计算依据与理论参考。 |
| 英文摘要: |
| Objective: This study systematically predicted the molecular mechanisms of the potential toxic effects of Gardenia on primary biliary cholangitis (PBC) by combining network toxicology, bioinformatics, and machine learning, and screened potential key targets. Methods: PBC-related targets were retrieved from the GeneCards (https://www.genecards.org/) and OMIM (https://omim.org/) databases, and PBC transcriptome data were obtained from the GEO (https://www.ncbi.nlm.nih.gov/geo/) database. The main chemical compositions of Gardenia were screened using the TCMSP database (https://www.tcmsp-e.com/tcmsp.php), and its SMILES structure was obtained from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/). The toxicological information of active components was analyzed in conjunction with the ADMETlab V3.0 database (https://admetlab3.scbdd.com/). Results: A total of 15 active components and 13 toxic components were predicted, constructing an intersection network of drug toxic components and PBC targets, identifying 54 intersection targets. Based on the GSE119600 dataset, 9 intersection targets with significant expression differences between normal and PBC samples were screened. Potential core targets FRAP1, ABCB1, NLRP3, and PIK3CG were identified through various machine learning algorithms, and the model's effectiveness was explained using SHAP. Molecular docking analysis results indicated that the main components of Gardenia have good binding affinity with core targets. Furthermore, GO and KEGG analyses showed that the targets significantly participate in multiple biological processes and signaling pathways, such as the IL-17 signaling pathway and oxidative stress response. Conclusion: This study systematically predicted the potential toxicological mechanisms of Gardenia in PBC, proposing a "multi-component-multi-target" toxicity hypothesis that awaits experimental validation, providing computational basis and theoretical reference for subsequent mechanism research and clinical safety medication. |
|
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
