Health Research and Policy


DATE: April 4, 2013
TIME: 1:15 - 3:00 pm
LOCATION: Medical School Office Building, Rm x303
TITLE: Kernels for Vertex Classification in Graphs: a Case Study in Predicting Protein Function and Molecular Mechanisms of Disease
SPEAKER: Predrag Radivojac
Associate Professor of Computer Science and Informatics, Indiana University - Bloomington

Graph kernels for supervised learning and inference on graphs have been around for more than a decade. However, the problem of designing robust kernel functions that can effectively compare graph neighborhoods in a variety of practical scenarios (e.g. the presence of incomplete and/or noisy data, auxiliary information) remains much less explored. Here, I will present our methods for vertex classification in large, sparse, and labeled graphs and their application to predicting protein function as well as molecular mechanisms of disease. Before presenting methods and results of our research, I will introduce necessary concepts related to proteins, protein function prediction, and its importance for modern biology and medicine. If time permits, I will briefly discuss how we use bioinformatics to guide wet lab experiments towards understanding molecular mechanisms of disease.

Suggested readings:
Original graphlet kernel paper:

Automated inference of molecular mechanisms of disease (from sequence data):

Our recent review on prediction of functional residues from structure:

Stanford Medicine Resources:

Footer Links: