On the design of shared memory approaches to parallelize a multiobjective bee-inspired proposal for phylogenetic reconstruction

Current efforts in solving computationally demanding optimization problems in bioinformatics rely on the combination of bioinspired computing and parallelism. The multiobjective nature shown by a wide variety of these problems represents an additional challenge, as the optimization of multiple objective functions involves growing computational requirements. In this work, a multiobjective metaheuristic inspired by honey bees is applied to tackle the phylogenetic inference problem. For this purpose, two parallel implementations for shared memory architectures are proposed: a synchronous generational model and a novel non-generational design inspired by the asynchronous behaviour of bees in nature. Experiments on six real biological data sets and comparisons with other parallel biological methods point out the relevance of applying nature-inspired parallelization strategies, addressing the performance pitfalls shown by traditional parallelization schemes.