An efficient discrete PSO coupled with a fast local search heuristic for the DNA fragment assembly problem

Abstract This paper focuses on Particle Swarm Optimization (PSO) applied to the DNA fragment assembly problem. Existing PSO algorithms for this permutation-based combinatorial problem use the Smaller Position Value (SPV) rule to transform continuous vectors into permutations of integers. However, this approach has limitations and is not suitable for this NP-hard problem. Here we propose a new discrete PSO that works directly in the search space of permutations and effectively addresses the fragment assembly problem. In our proposal, the fact that relative ordering of DNA fragments is most indicative of assembly accuracy is exploited in the particle update mechanism. This is implemented through a new operator called Probabilistic Edge Recombination (PER). This operator builds a new position through the probabilistic recombination of edges (adjacency relations) between fragments from the current position, the personal best, and the group best. Additionally, we design variants of the proposed PSO algorithm by applying heuristic information and/or local search. With this aim, we develop a new fast variant of the best state-of-the-art local search algorithm for the assembly problem. Extensive experiments have been conducted to demonstrate the efficiency and effectiveness of the algorithms used. In comparison with the state-of-the-art assembly techniques, our algorithms achieve a better performance.