Predicting Host-Pathogen Interactions Using Deep Learning on Protein Structural Data

Pathogens, or agents which spread disease within humans, enter host (e.g., human) cells in order to carry out their processes and disrupt normal host cell functions.

For this project, Draizen and Lanchantin hypothesize that pathogenic proteins mimic normal human proteins in order to competitively bind the host protein's interaction partner. They will focus on predicting the interaction sites of host-host proteins so that scientists can better understand how these proteins interact for cell development. In particular, the researchers plan to leverage the recent success of graph neural networks, which they hypothesize will be able to learn protein interaction surfaces and generalize to new pathogens of interest.

Completed in: