Quality molecular representations are key to foundation model development in bio-medical research. Previous efforts have typically focused on a single representation or molecular view, which may have strengths or weaknesses on a given task. We develop Multi-view Molecular Embedding with Late Fusion (MMELON), an approach that integrates graph, image and text views in a foundation model setting and may be readily extended to additional representations. Single-view foundation models are each pre-trained on a dataset of up to 200M molecules. The multi-view model performs robustly, matching the performance of the highest-ranked single-view. It is validated on over 120 tasks, including molecular solubility, ADME properties, and activity against G Protein-Coupled receptors (GPCRs). We identify 33 GPCRs that are related to Alzheimer's disease and employ the multi-view model to select strong binders from a compound screen. Predictions are validated through structure-based modeling and identification of key binding motifs.
@article{arxiv.2410.19704,
title = {Multi-view biomedical foundation models for molecule-target and property prediction},
author = {Parthasarathy Suryanarayanan and Yunguang Qiu and Shreyans Sethi and Diwakar Mahajan and Hongyang Li and Yuxin Yang and Elif Eyigoz and Aldo Guzman Saenz and Daniel E. Platt and Timothy H. Rumbell and Kenney Ng and Sanjoy Dey and Myson Burch and Bum Chul Kwon and Pablo Meyer and Feixiong Cheng and Jianying Hu and Joseph A. Morrone},
journal= {arXiv preprint arXiv:2410.19704},
year = {2026}
}
Comments
40 pages including supplement. 10 figures, 8 tables