Performance evaluation of dominance-based and indicator-based multiobjective approaches for phylogenetic inference

One of the main research lines in bioinformatics focuses on the optimization of biological processes involving several objective functions. Due to the variety of multiobjective strategies which are available, comparative studies are needed to decide which algorithmic designs lead to improved results. This work tackles the inference of phylogenetic relationships by means of multiobjective metaheuristics. More specifically, we perform the comparative assessment of two lines of multiobjective schemes: dominance-based and indicator-based approaches. On the dominance-based side, we consider two algorithms: Fast Non-Dominated Sorting Genetic Algorithm II and Strength Pareto Evolutionary Algorithm 2. Indicator-based designs are represented by the Indicator-Based Evolutionary Algorithm and a new Indicator-Based Multiobjective Bat Algorithm. The experimental evaluation of these methods is conducted over six real biological datasets, making comparisons with multiple state-of-the-art phylogenetic tools. Our experimentation verifies the significant performance achieved when combining indicator-based approaches and swarm intelligence. Particularly, different multiobjective metrics (hypervolume, set coverage, and spacing) and biological testing procedures highlight the promising results reported by this kind of algorithmic designs.