ADMETPred
A High-Throughput ADMET Prediction Platform Integrating Multi-Model Algorithms and Interpretable Substructure Identification.
A High-Throughput ADMET Prediction Platform Integrating Multi-Model Algorithms and Interpretable Substructure Identification.
Cai Chuipu, Chen Zhuang, Wang Zhe, Wu Lingyu, Liu Zhihong, Fang Jiansong*, Liu Ailin*. (2026).
ADMETPred: A High-Throughput ADMET Prediction Platform Integrating Multi-Model Algorithms and Interpretable Substructure Identification.
SCIENCE CHINA Life Sciences, 69, 1–17.
doi: 10.1007/s11427-025-3166-8
ADMETPred combines 189 models (LightGBM, XGBoost, Random Forest, and GAT) trained on 120,616 rigorously curated compounds to predict 27 ADMET endpoints, while integrating an interpretable, attention-driven substructure highlighting module. The platform demonstrates superior predictive accuracy and efficiency by leveraging multi-algorithm synergy, high-throughput batch processing capabilities with parallelized architecture, and customizable workflows for improved flexibility.
Integrating multi-model approaches trained on 120,616 rigorously curated data entries enables accurate prediction of 27 ADMET endpoints.
Attention-driven substructure highlighting module to bridge predictions with actionable structural optimization insights.
Unlimited batch processing with field-leading rapid computational speed via parallel architecture.
User-customizable workflow allow dynamic selection of endpoints, algorithms, and feature representations.