Multi-view biomedical foundation models for molecule-target and property prediction

  • 2024-10-25 18:22:33
  • Parthasarathy Suryanarayanan, Yunguang Qiu, Shreyans Sethi, Diwakar Mahajan, Hongyang Li, Yuxin Yang, Elif Eyigoz, Aldo Guzman Saenz, Daniel E. Platt, Timothy H. Rumbell, Kenney Ng, Sanjoy Dey, Myson Burch, Bum Chul Kwon, Pablo Meyer, Feixiong Cheng, Jianying Hu, Joseph A. Morrone
  • 0

Abstract

Foundation models applied to bio-molecular space hold promise to acceleratedrug discovery. Molecular representation is key to building such models.Previous works have typically focused on a single representation or view of themolecules. Here, we develop a multi-view foundation model approach, thatintegrates molecular views of graph, image and text. Single-view foundationmodels are each pre-trained on a dataset of up to 200M molecules and thenaggregated into combined representations. Our multi-view model is validated ona diverse set of 18 tasks, encompassing ligand-protein binding, molecularsolubility, metabolism and toxicity. We show that the multi-view models performrobustly and are able to balance the strengths and weaknesses of specificviews. We then apply this model to screen compounds against a large (>100targets) set of G Protein-Coupled receptors (GPCRs). From this library oftargets, we identify 33 that are related to Alzheimer's disease. On thissubset, we employ our model to identify strong binders, which are validatedthrough structure-based modeling and identification of key binding motifs.

 

Quick Read (beta)

loading the full paper ...