ADMET property prediction through combinations of molecular fingerprints

James H. Notwell, Michael W. Wood

2023-09-29Prediction TDC ADMET Benchmarking Group

Abstract

While investigating methods to predict small molecule potencies, we found random forests or support vector machines paired with extended-connectivity fingerprints (ECFP) consistently outperformed recently developed methods. A detailed investigation into regression algorithms and molecular fingerprints revealed gradient-boosted decision trees, particularly CatBoost, in conjunction with a combination of ECFP, Avalon, and ErG fingerprints, as well as 200 molecular properties, to be most effective. Incorporating a graph neural network fingerprint further enhanced performance. We successfully validated our model across 22 Therapeutics Data Commons ADMET benchmarks. Our findings underscore the significance of richer molecular representations for accurate property prediction.

Results

Task	Dataset	Metric	Value	Model
Drug Discovery	tdcommons	TDC.Caco2_Wang	0.276	MapLight
Therapeutics Data Commons	tdcommons	TDC.Caco2_Wang	0.276	MapLight

Related Papers

Multi-Strategy Improved Snake Optimizer Accelerated CNN-LSTM-Attention-Adaboost for Trajectory Prediction2025-07-21 Generative Click-through Rate Prediction with Applications to Search Advertising2025-07-15 Conformation-Aware Structure Prediction of Antigen-Recognizing Immune Proteins2025-07-11 Foundation models for time series forecasting: Application in conformal prediction2025-07-09 Predicting Graph Structure via Adapted Flux Balance Analysis2025-07-08 Speech Quality Assessment Model Based on Mixture of Experts: System-Level Performance Enhancement and Utterance-Level Challenge Analysis2025-07-08 A Wireless Foundation Model for Multi-Task Prediction2025-07-08 High Order Collaboration-Oriented Federated Graph Neural Network for Accurate QoS Prediction2025-07-07