Multilocus Phylogeny Estimation Using Probabilistic Topic Modeling

Abstract

AbstractInferring the evolutionary history of species or populations employing multilocus analysis is gaining ground in phylogenetic analysis. We developed an alignment-free method to infer the multilocus species tree, which is implemented in the Python package TopicContml. The method operates in two primary stages. First, it uses probabilistic topic modeling (specifically, Latent Dirichlet Allocation or LDA) to extract topic frequencies from k-mers, which are in turn derived from multilocus DNA sequences. Second, these extracted frequencies serve as an input for the program Contmlin the PHYLIP package, which is used to generate a species tree. We evaluated the performance of our method with biological and simulated datasets: a dataset with 14 DNA sequence loci from 78-92 individuals from two Australian bird species distributed in 9 populations, and a second dataset of 67317 autosomal loci and 4157 X-chromosome loci of 6 species in theAnopheles gambiaecomplex, and several simulated data sets. Our empirical results and simulated data suggest that our method is efficient and statistically accurate. We also assessed the uncertainty of the estimated relationships among clades using a bootstrap procedure for aligned data and fork-mers.