Compound Protein Interaction Prediction

Abstract

Quals talk:

Protein-ligand interactions are central to many biological processes. Drug-target, molecule-enzyme interactions are both facets of this problem. Predicting these interactions enables drug discovery, drug repurposing, drug side effect analysis, and even synthetic biology to generate molecules or proteins with specific interaction properties. Computational methods like statistical, similarity based and deep learning based methods have been applied to the protein-ligand interaction prediction problem for some time now. In this research, we apply Contrastive Multiview Coding, successfully used previously in image recognition, to improve the quality of interaction prediction. We propose a novel general purpose data stratification approach to show improved model performance for multiple datasets over the baseline model without contrastive multiview coding.

Please join the meeting in Cummings #610 or via Zoom: https://tufts.zoom.us/j/96722224228

Dial-in not available for this talk.