Multi-objective two-level swarm intelligence approach for multiple RNA sequence-structure alignment

• Published in: Swarm and evolutionary computation 2017-06, Vol.34, p.130-144
• Main Authors: Lalwani, Soniya, Kumar, Rajesh, Deep, Kusum
• Format: Article
• Language: English
• Subjects: Conflicting objectives; Minimum free energy; Multi-objective optimization; Multiple sequence alignment; Non-dominated solutions; Pareto optimal solution; Particle swarm optimization; RNA secondary structure
• ISSN: 2210-6502
• DOI: 10.1016/j.swevo.2017.02.002

This paper proposes a novel two-level particle swarm optimization algorithm for multi-objective optimization (MO-TLPSO) employed to a challenging problem of bioinformatics i.e. RNA sequence-structure alignment. Level one of the proposed approach optimizes the dimension of each swarm which is sequence length for the addressed problem, whereas level two optimizes the particle positions and then evaluates both the conflicting objectives. The conflicting objectives of the addressed problem are obtaining optimal multiple sequence alignment as well as optimal secondary structure. Optimal secondary structure is obtained by TL-PSOfold, the structure is further used for computing the contribution of base pairing of individual sequence and the co-variation between aligned positions of sequences so as to make the structure closer to the natural one. The results are tested against the popular softwares for pairwise and multiple alignment at BRAlibase benchmark datasets. Proposed work is so far the first multi-objective optimization based approach for structural alignment of multiple RNA sequences without converting the problem into single objective. Also, it is the first swarm intelligence based approach that addresses sequence-structure alignment issue of RNA sequences. Simulation results are compared with the state-of-the-art and competitive approaches. MO-TLPSO is found well competent in producing pairwise as well as multiple sequence-structure alignment of RNA. The claim is supported by performing statistical significance testing using one way ANOVA followed by Bonferroni post-hoc analysis for both kind of alignments.