A tale of too many trees: a conundrum for phylogenetic regression

AO_SCPLOWBSTRACTC_SCPLOWJust exactly which tree(s) should we assume when testing evolutionary hypotheses? This question has plagued comparative biologists for decades. Given a perfectly estimated tree (if this is even possible in practice), we seldom know with certainty whether such a tree is truly best (or even adequate) to represent the evolutionary history of our studied traits. Regardless of our certainty, choosing a tree is required for all phylogenetic comparative methods. Yet, phylogenetic conflict and error are ubiquitous in modern comparative biology, and we are still learning about their dangers when testing evolutionary hypotheses. Here we investigated the consequences of gene tree-species tree mismatch for phylogenetic regression in the presence of incomplete lineage sorting. Our simulation experiments reveal excessively high false positive rates for mismatched phylogenetic regression with both small and large trees, simple and complex traits, and known and estimated phylogenies. In some cases, we find evidence of a directionality of error: incorrectly assuming a species tree for traits that evolved according to a gene tree sometimes fares worse than the opposite. To explore difficult yet realistic regression scenarios, we also used estimated rather than known trees to conduct case studies, as well as an expansive gene expression dataset to investigate an arguably best-case scenario in which one may have a better chance to match tree with trait. Though never meant to be a panacea for all that may ail phylogenetic comparative methods, we found promise in the application of a robust estimator as a potential, albeit imperfect, solution to some issues raised by tree mismatch, perhaps offering a path forward. Collectively, our results emphasize the importance of careful study design for comparative methods, highlighting the need to fully appreciate the role of adequate phylogenetic modeling for testing evolutionary hypotheses.