Comparing Different Operators and Models to Improve a Multiobjective Artificial Bee Colony Algorithm for Inferring Phylogenies

Maximum parsimony and maximum likelihood approaches to phylogenetic reconstruction were proposed with the aim of describing the evolutionary history of species by using different optimality principles. These discrepant points of view can lead to situations where discordant topologies are inferred from a same dataset. In recent years, research efforts in Phylogenetics try to apply multiobjective optimization techniques to generate phylogenetic topologies which suppose a consensus among different criteria. In order to generate high quality topologies, it is necessary to perform an exhaustive study about topological search strategies as well as to decide the most fitting molecular evolutionary model in agreement with statistical measurements. In this paper we report a study on different operators and models to improve a Multiobjective Artificial Bee Colony algorithm for inferring phylogenies according to the parsimony and likelihood criteria. Experimental results have been evaluated using the hypervolume metrics and compared with other multiobjective proposals and state-of-the-art phylogenetic software.