A Novel Binary Particle Swarm Optimization for Multiple Sequence Alignment

Multiple sequence alignment (MSA) is a basic operation in bioinformatics, which is known to be an NP-complete problem. With the advent of the post genetic era, the number of biological sequences is increasing exponentially, it is urgent to develop high-efficiency heuristic parallel algorithms to solve MSA. In this paper, a novel binary particle swarm optimization (NBPSO) is proposed for MSA. Firstly, a novel binary code is designed to encode particle to adapt PSO to MSA, velocity and position updating formulas are refined accordingly. Regarding of the ‘illegal’ phenomenon of MSA after positions updating, based on probability theory, an auxiliary particle re-initialization strategy is proposed. Multi-swarm PSO is used to make the searching more parallel, diversity index dt is proposed based on Hamming Distance, mutation takes place when dt falls below threshold. The proposed NBPSO-MSA is operated on two benchmarks, i.e., OX-Bench and BAliBASE. Experiments show that, NBPSO-MSA outperforms other 12 state-of-the art aligners in accuracy, the trajectories of particles show that, diversity index and mutation help multi-swarm NBPSO jump out of local optima and lead to rapid convergence.