Dr. Junyuan Lin, Loyola Marymount University

In this project, we present the award-winning algorithm that ranked No.1 in the Disease Module Identification DREAM international bioinformatics challenge. The goal of the challenge was to systematically assess module identification methods on the latest molecular networks, and to discover novel network modules/pathways underlying complex diseases. We defined a new diffusion-based metric for the Protein-Protein Interaction networks to measure protein similarity based on network properties. To compute the similarity between proteins in large interaction networks, we applied a modified Algebraic Multi-grid (AMG) solver and random projection method to reduce computational complexity. Finally, we applied spectral clustering to partition the protein network into disease modules and predict their functionalities.

Zoom