A new distributed modified extremal optimization for optimizing protein structure alignment

Abstract

Identifying similar structures in proteins has emerged as one of the most attractive research topics in the post-genome era. Protein structure alignment, which is similar to sequence alignment, identifies the structural homology between two protein structures according to their three-dimensional conformation. One of the simplest yet most robust techniques for optimizing protein structure alignment is the contact map overlap maximization problem (the CMO problem). In this paper, we focus on heuristics for the CMO problem. In our previous work, we proposed a bio-inspired heuristic using distributed modified extremal optimization (DMEO) for the CMO problem. DMEO is a hybrid of population-based modified extremal optimization (PMEO) and the island model. DMEO enhances population diversity; however, individual evolution is extremely monotonous because evolutions of it is based on the greedy moving approach. To address this issue, we propose a novel bio-inspired heuristic, i.e., DMEO with different evolutionary strategy (DMEODES). DMEODES is also based on the island model; however, some of the islands, called hot-spot islands, have a different evolutionary strategy. To evaluate DMEODES, we used actual protein structures. Experimental results showed that DMEODES outperforms DMEO.