Integrating biological insights with topological characteristics for improved complex prediction from protein interaction networks

Abstract

This thesis is about devising and developing computational techniques for protein complex prediction from protein interaction (PPI) networks. We integrate a variety of biological insights - core-attachment structure, functional interactions and temporal information - with topology-based clustering of PPI networks for complex prediction. We substantially improve the available 'state-of-the-art' techniques and also identify a significant number of complexes that are sparse in the PPI network.