MEGA-CL: A Molecular Foundation Model for Generalizable ADMET Prediction through Graph External Attention and Contrastive Learning

2026-07-27

AI Summary

This work addresses the low accuracy in extrapolating in vitro to in vivo ADMET properties and poor generalization of existing models for small-molecule drugs by proposing MEGA-CL, a novel molecular foundation model. MEGA-CL innovatively integrates self-supervised contrastive learning with a multi-head external attention mechanism within an enhanced message-passing architecture, enabling simultaneous modeling of local substructures and global graph relationships while effectively mitigating the oversmoothing issue common in graph neural networks. Evaluated across 13 benchmark datasets and 21 ADMET tasks, MEGA-CL substantially outperforms state-of-the-art methods: it achieves prediction errors within three-fold for over 75% of tasks, predicts human liver microsomal (HLM) clearance within two-fold error for more than half of 18 novel compounds, and demonstrates prospective validation with all HLM clearance errors below 2.5-fold and 73.3% accuracy in CYP450 inhibition classification.

Abstract

Predicting the absorption, distribution, metabolism, excretion and toxicity (ADMET) properties of small molecules remains a major challenge in drug discovery. Here, we present MEGA-CL, a foundation graph neural network framework for universal molecular ADMET prediction. MEGA-CL integrates self-supervised contrastive learning with a multi-head external attention mechanism and an enhanced message-passing architecture, enabling simultaneous modeling of local chemical substructures and global inter-graph relationships while mitigating over-smoothing effects commonly observed in deep graph networks. Across 13 benchmark datasets and 21 downstream ADMET tasks, MEGA-CL consistently outperforms state-of-the-art baseline models. In particular, the framework demonstrates robust performance on challenging regression tasks, including clearance (CL) and steady-state volume of distribution (VDss), while maintaining strong generalization ability in independent external validation. Clinically relevant predictive accuracy was achieved, with more than 75% of predictions falling within a 3-fold error range. In an external evaluation on 18 novel compounds derived from recently approved FDA drugs, over 50% of human liver microsome clearance (HLMC) predictions were within a 2-fold error range. To further assess its practical applicability, MEGA-CL was prospectively evaluated on three preclinical drug candidates using in vitro hepatic microsomal metabolism assays and CYP450 inhibition assays guided by model predictions. The predicted HLMC values for all candidates were within 2.5-fold of the experimentally measured values, and 73.3% of CYP450 inhibition endpoints (11/15) were correctly classified. These results demonstrate the potential of MEGA-CL as a generalizable framework for accelerating in silico ADMET evaluation and early-stage drug candidate optimization.

Problem

• ADMET prediction
• drug discovery
• molecular property prediction
• toxicity
• pharmacokinetics

Innovation

• contrastive learning
• external attention
• graph neural network
• ADMET prediction
• foundation model