Scholarlink

<p>Drug-induced reproductive toxicity is a critical concern in drug safety evaluation, whereas conventional assessment methods are often constrained by high costs and long experimental cycles. In this study, a machine learning-based predictive model for reproductive toxicity was developed and integrated with data from the FDA Adverse Event Reporting System (FAERS), network toxicology analysis, molecular docking, and molecular dynamics simulation to systematically evaluate the post-marketing reproductive toxicity risk of drugs and explore their potential mechanisms. Among the evaluated machine learning algorithms, LightGBM demonstrated the best overall performance, achieving an F1-score of 0.854, a ROC-AUC of 0.933, a PR-AUC of 0.931, and an MCC of 0.705 on the independent test set, with robust generalization confirmed by ten-fold cross-validation. Among drugs approved between 2015 and 2024, 72 were predicted to have a high risk of reproductive toxicity. FAERS-based signal comparison showed that 55 of these drugs (76.39%) were associated with reproductive toxicity-related adverse event reports, indicating consistency between model predictions and FAERS-reported reproductive toxicity-related adverse events. Network toxicology analysis identified 12 key targets, including ESR1, IGF1, and AKT1, that may be involved in reproductive toxicity. Molecular docking showed that drugs with high predicted reproductive toxicity risk could bind effectively to multiple toxicity-related targets, while molecular dynamics simulations confirmed stable interactions between selected drugs and ESR1, mainly through hydrogen-bonding and hydrophobic interactions. Favorable binding free energies further supported their potential multi-target effects. Overall, this integrated strategy combining predictive modeling with FAERS-based signal comparison provides a useful framework for drug safety evaluation and mechanistic investigation of reproductive toxicity.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Prediction and mechanistic insights into drug-induced reproductive toxicity through integrated machine learning, FAERS-based signal comparison, and network toxicology analyses

  • 期刊论文

  • 发表日期：

    2026年7月11日

查看全文
  • Jiawang Yan,
  • Yaofeng Zhou,
  • Yihuan Zhao,
  • Fushan Tang

摘要

Drug-induced reproductive toxicity is a critical concern in drug safety evaluation, whereas conventional assessment methods are often constrained by high costs and long experimental cycles. In this study, a machine learning-based predictive model for reproductive toxicity was developed and integrated with data from the FDA Adverse Event Reporting System (FAERS), network toxicology analysis, molecular docking, and molecular dynamics simulation to systematically evaluate the post-marketing reproductive toxicity risk of drugs and explore their potential mechanisms. Among the evaluated machine learning algorithms, LightGBM demonstrated the best overall performance, achieving an F1-score of 0.854, a ROC-AUC of 0.933, a PR-AUC of 0.931, and an MCC of 0.705 on the independent test set, with robust generalization confirmed by ten-fold cross-validation. Among drugs approved between 2015 and 2024, 72 were predicted to have a high risk of reproductive toxicity. FAERS-based signal comparison showed that 55 of these drugs (76.39%) were associated with reproductive toxicity-related adverse event reports, indicating consistency between model predictions and FAERS-reported reproductive toxicity-related adverse events. Network toxicology analysis identified 12 key targets, including ESR1, IGF1, and AKT1, that may be involved in reproductive toxicity. Molecular docking showed that drugs with high predicted reproductive toxicity risk could bind effectively to multiple toxicity-related targets, while molecular dynamics simulations confirmed stable interactions between selected drugs and ESR1, mainly through hydrogen-bonding and hydrophobic interactions. Favorable binding free energies further supported their potential multi-target effects. Overall, this integrated strategy combining predictive modeling with FAERS-based signal comparison provides a useful framework for drug safety evaluation and mechanistic investigation of reproductive toxicity.