The Impact of Multiple Sequence Alignment Error on Phylogenetic Estimation under Variable-Across-Phylogeny Substitution Models

In many different species, it has been observed that nucleotide compositions are not identical on the genic and even genomic scale. This observation contradicts a commonly held assumption in most maximum likelihood based phylogenetic estimation methods - that the process governing DNA evolution is identical across lineages. We show that when DNA evolution is nonhomogeneous, topological estimation and continuous parameter estimation are impacted both by alignment quality and model misspecification due to the homogeneity-across-lineages assumption.