Performance of a phylogenetic independent contrast method and an improved pairwise comparison under different scenarios of trait evolution after speciation and duplication

Abstract

O_LIDespite the importance of gene function to evolutionary biology, the applicability of comparative methods to gene function is poorly known. A specific case which has crystalized methodological questions is the "ortholog conjecture", the hypothesis that function evolves faster after duplication (i.e., in paralogs), and conversely is conserved between orthologs. Since the mode of functional evolution after duplication is not well known, we investigate under which reasonable evolutionary scenarios phylogenetic independent contrasts or pairwise comparisons can recover a putative signal of different functional evolution between orthologs and paralogs. C_LIO_LIWe investigate three different simulation models, which represent reasonable but simplified hypotheses about the evolution of a gene function trait. These are time dependent trait acceleration, correlated changes in rates of both sequence and trait evolution, and asymmetric trait jump. For each model we tested phylogenetic independent contrasts and an improved pairwise comparison method which accounts for interactions between events and node age. C_LIO_LIBoth approaches lose power to detect the trend of functional evolution when the functional trait accelerates for a long time following duplication for trees with many duplications, with better power of phylogenetic contrasts under intermediate scenarios. Concomitant increase in evolutionary rates of sequence and of trait after duplication can lead to both an incorrect rejection of the null under null simulations of trait evolution, and a false rejection of the ortholog conjecture under ortholog conjecture simulations, by phylogenetic independent contrasts. Improved pairwise comparisons are robust to this bias. Both approaches perform equally well under rapid shifts in traits. C_LIO_LIConsidering our ignorance of gene function evolution, and the potential for bias under simple models, we recommend methodological pluralism in studying gene family evolution. Functional phylogenomics is complex and results supported by only one method should be treated with caution. C_LI