A genetic algorithm for character state live phylogeny

Abstract

Character state live phylogeny generalizes character state phylogeny in the sense that they relate taxonomic units based on their similarities over a set of characters, but allowing live ancestors. An approach for character state live phylogeny reconstruction is called parsimony, where one tries to minimize the total number of character state changes along the edges of the tree. The problem of finding a tree that minimizes this number is known as large live parsimony problem. When the tree topology is also given as input, the problem is known as small live parsimony problem. We propose a genetic algorithm to solve the large live problem, which uses extended versions of the algorithms of Fitch and Sankoff to solve the small live problem, both devised in this work. Besides, we performed two experiments. In the first one, a multiple alignment of H1N1 and H3N2 viruses from different countries, taken as input, allowed to obtain interesting live phylogenies, representing alternative evolutionary hypothesis. The second experiment took as input a multiple alignment of the HIV virus env gene, from one patient, read in different dates through 12 years. The generated live phylogenies were similar to the ones generated by PAUP, where dates close to each other were grouped into clusters, but suggesting new evolutionary stories.