The impact and interplay of long and short branches on phylogenetic information content

In molecular systematics, evolutionary trees are reconstructed from sequences at the tips under simple models of site substitution. A central question is how much sequence data is required to reconstruct a tree accurately? The answer depends on the lengths of the branches (edges) of the tree, with very short and very long edges requiring long sequences for accurate tree inference, particularly when these branch lengths are arranged in certain ways. For four-taxon trees, the sequence length question has been investigated for the case of a rapid speciation event in the distant past. Here, we generalize results from this earlier study, and show that the same sequence length requirement holds even when the speciation event is recent, provided that at least one of the four taxa is distantly related to the others. However, this equivalence disappears if a molecular clock applies, since the length of the long outgroup edge becomes largely irrelevant in the estimation of the tree topology for a recent divergence. We also discuss briefly some extensions of these results to models in which substitution rates vary across sites and to settings where more than four taxa are involved.