Protein complex prediction via bottleneck-based graph partitioning

Detecting protein complexes is one of essential and fundamental tasks in understanding various biological functions or processes. Therefore, precise identification of protein complexes is indispensible. For more precise detection of protein complexes, we propose a novel data structure which employs bottleneck proteins as partitioning points for detecting the protein complexes. The partitioning process allows overlapping between resulting protein complexes. We applied our algorithm to several PPI (Protein-Protein Interaction) networks of Saccharomyces cerevisiae and Homo sapiens, and validated our results using public databases of protein complexes. Our algorithm resulted in overlapping protein complexes with significantly improved F1 score, which comes from higher precision.