关键词: 蛋白质相互作用（PPI）, 鸽子优化算法, 蛋白质复合物, 聚类

Abstract: Detecting protein complexes is crucial to understand the principles of cellular organization and predict diseases. Yet up to now, the performance of existing protein complex detection algorithms is not very ideal for the deficiencies of low accuracy, sensitive to noisy data and so on. This paper proposes a novel algorithm named pigeon-inspired optimization clustering (PIOC) algorithm. It puts forward the concepts of cluster's closely adjacent nodes and attachment's attachment degree on core, and identifies protein complexes by simulating the process of pigeons finding home based on these two concepts. The PIOC algorithm combines the characters of first global search then local search of pigeon-inspired optimization (PIO) algorithm and the core-attachment structure of protein complexes. Particularly, it first develops the cores of protein complexes by the global search of PIO's map and compass operator, and then forms the protein complexes by gathering in the attachment proteins to the cores based on the local search of PIO landmark operation. The experimental results on yeast protein DIP dataset demonstrate that PIOC is more   effective in detecting protein complexes than the state-of-the-art complex detection algorithms.

Key words:  protein-protein interaction (PPI), pigeon-inspired optimization algorithm, protein complex, clustering