A Survey of Evolutionary Crossover Operators as Applied to Phylogenetic Inferencing

Inferring phylogenetic trees from sequence or morphological characters is computationally intensive, and requires the use of heuristics. The search space and fitness landscape of treespace for even modest datasets is large and multi-modal. Traditional computational approaches typically involve hill-climbing using intra-tree branch-swapping techniques such as subtree pruning and re-grafting (SPR) or tree bisection and reconnection (TBR) to propose incremental modifications to topology. Several genetic algorithms for phylogenetic inferencing have been proposed and implemented, all of which involve crossover operators to recombine putative building blocks in selected parent trees. In this paper, we survey four different crossover operators present in the literature. Specifically, we evaluate their expressiveness, destructiveness, ergodicity, and bias. An overall assessment of these distinct crossover approaches is summarized.