Scholarlink

<p>The prediction of absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties plays a critical role in early-stage drug discovery. While artificial intelligence (AI) has demonstrated transformative potential in revolutionizing this field, existing computational tools remain constrained by limitations in throughput, interpretability, and immobilized modeling frameworks. To address these challenges, we developed ADMETPred, an innovative platform that integrates machine learning and graph neural networks to deliver rapid, accurate, and comprehensive ADMET profiling. Trained on rigorously curated datasets comprising 120,616 compounds, ADMETPred employs 189 models combining LightGBM, XGBoost, Random Forest, and graph attention network to predict 27 drug pharmacokinetic, metabolism, and toxicity endpoints. Compared with current tools, ADMETPred demonstrates superior predictive accuracy by leveraging multi-algorithm synergy, high-throughput batch processing capabilities with parallelized architecture, and customizable workflows for improved prediction flexibility. Notably, the platform integrates an interpretable, attention-driven substructure highlighting module to bridge predictions with actionable structural optimization insights. Case studies spanning post-market drug surveillance, natural product toxicity screening, and lead compound preclinical safety assessment demonstrated alignment with experimental and clinical evidence. In summary, ADMETPred provides a practical resource to enhance early-stage drug development by combining lowered usage barriers with reliable ADMET profiling, freely accessible at <a href="http://admetpred.pumc.ai-tcm.cn/">http://admetpred.pumc.ai-tcm.cn/</a>.</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

ADMETPred: a high-throughput ADMET prediction platform integrating multi-model algorithms and interpretable substructure identification

  • 期刊论文

  • 发表日期：

    2026年3月26日

查看全文
  • Chuipu Cai,
  • Zhuang Chen,
  • Zhe Wang,
  • Lingyu Wu,
  • Zhihong Liu,
  • Jiansong Fang,
  • Ailin Liu

摘要

The prediction of absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties plays a critical role in early-stage drug discovery. While artificial intelligence (AI) has demonstrated transformative potential in revolutionizing this field, existing computational tools remain constrained by limitations in throughput, interpretability, and immobilized modeling frameworks. To address these challenges, we developed ADMETPred, an innovative platform that integrates machine learning and graph neural networks to deliver rapid, accurate, and comprehensive ADMET profiling. Trained on rigorously curated datasets comprising 120,616 compounds, ADMETPred employs 189 models combining LightGBM, XGBoost, Random Forest, and graph attention network to predict 27 drug pharmacokinetic, metabolism, and toxicity endpoints. Compared with current tools, ADMETPred demonstrates superior predictive accuracy by leveraging multi-algorithm synergy, high-throughput batch processing capabilities with parallelized architecture, and customizable workflows for improved prediction flexibility. Notably, the platform integrates an interpretable, attention-driven substructure highlighting module to bridge predictions with actionable structural optimization insights. Case studies spanning post-market drug surveillance, natural product toxicity screening, and lead compound preclinical safety assessment demonstrated alignment with experimental and clinical evidence. In summary, ADMETPred provides a practical resource to enhance early-stage drug development by combining lowered usage barriers with reliable ADMET profiling, freely accessible at http://admetpred.pumc.ai-tcm.cn/.