A Maximum-Likelihood Method to Construct Phylogenetic Trees Using Low-Homoplasy Markers: New Insights into Neoaves Phylogeny

Abstract

We present a novel probabilistic method for species tree reconstruction from low-homoplasy markers, termed MLRE (Maximum Likelihood for Retrotransposed Elements). MLRE uses an infinite-sites model for introducing new markers, which then evolve following a Markov process with three states: polymorphic, fixed, and lost. The model allows MLRE to optimize the likelihood of a tree with edge lengths in coalescent units. The method further incorporates variable parameters for insertion rates across lineages and missing data rates across taxa. For simulated data, MLRE consistently performs better than existing tree inference methods. Reanalyzing a retrotransposon dataset for Neoaves, we find better agreement with published trees from other data types, and we propose one strongly supported novel clade. These results underscore the value of retrotransposon markers in phylogenetics and highlight MLRE as a robust framework for species tree inference under challenging evolutionary conditions.