We focus on machine learning in precision medicine.
Protein-protein interactions are the center of inter- and intra-cell communication and signaling. Many diseases such as cancer, neurodegenerative disorders involve malfunctioning proteins which result in erroneous signaling. We develop computational methods to predict how proteins interact and study how genomic variations and mutations rewires signaling and their relation to diseases, particularly in cancer.
Protein-protein interface prediction
Proteins recognize each other through their binding surfaces. Predicting which sides form a protein-protein interaction is a difficult and important problems. DeepInterface projects uses deep CNNs to predict which protein-binding sites are biologically relevant.
Development of a new drug is extremely costly and lengthy process. An alternative strategy is to repositioning known drugs to new diseases by predicting protein interface-drug interactions.
Microbe-human protein interactions
Human microbiome includes around 100 trillion bacterial cells (more than our own cells). We develop methods to unravel how microbes (bacteria, virus) modulate protein interactions and cause diseases.