Comparing evolutionary rates between trees, clades and traits

Abstract

The tempo of evolutionary change through time is among the most heavily studied dimensions of macroevolution using phylogenies. Here, we present a simple, likelihood‐based method for comparing the rate of phenotypic evolution for continuous characters between trees. Our method is derived from a previous approach published by Brian O'Meara and colleagues in 2006. We examine the statistical performance of the method and find that it suffers from the typical downward bias expected for maximum likelihood estimates of the variance, but only for very small trees. We find that evolutionary rates are estimated with minimal bias for trees of even relatively modest size. We also find that type I error rates based on a likelihood‐ratio test are minimally elevated above the nominal level, even for small phylogenies. The type I error rate can be reduced to a level at or below its nominal value by substituting a test‐statistic distribution obtained via simulation under the null hypothesis of no difference in evolutionary rate among trees. We discuss the consequences of failing to account for uncertainty in the estimation of species means or in the phylogeny, and describe strategies for taking this uncertainty into consideration during estimation. We also identify how our approach is related to previous methods for comparing the rate of evolution among different clades of a single tree or between different phenotypic traits. Finally, we describe how the method can be applied to different evolutionary models and to discrete characters—options that are already implemented in software. Evolutionary biologists continue to be intrigued by changes in the tempo of phenotypic evolution across the tree of life. The method described herein should be useful for contexts in which changes in the evolutionary rate or process between two or more clades of distant or unknown relationship are of interest.

Elevational and microclimatic drivers of thermal tolerance in Andean Pristimantis frogs

“…We also used the function "ratebytree" from the R package "phytools" (Revell, 2012) to evaluate whether there are differences in the rate of CTmax and CTmin evolution. This function allows the comparison of phenotypic evolution of continuous traits between trees under different models of evolution: "random walk" (Brownian motion, BM) and the adaptive models Ornstein-Uhlenbeck (OU) and early burst (EB) (Revell et al, 2018). We first fitted the three different models of evolution to each physiological trait (CTmax and CTmin) using the "fitContinuous" function from the R package geiger (Harmon, Weir, Brock, Glor, & Challenger, 2008).…”

Plant evolutionary history mainly explains the variance in biomass responses to climate warming at a global scale

“…These advantages with thinner roots could increase the plant's efficiency to absorb water and nutrients (Chen et al ., ), and shape the different responses of biomass to warming between gymnosperms and angiosperms. However, in general, the divergence age could not predict the magnitude of the phylogenetic components well (Tables S3, S4), indicating that the evolutionary rate might vary among different clades (Revell et al ., ).…”

“…These methods relied on the underlying evolutionary models such as Brownian motion and Ornstein-Uhlenbeck models. However, as the evolutionary processes might be more complicated than the model assumption (Castiglione et al, 2018;Revell et al, 2018), these methods could underestimate the contribution of phylogenetic effects (Cadotte et al, 2017). For that reason, we applied a purely data-driven variance partitioning approach, the phylogenetic eigenvector regression (PVR; Diniz-Filho et al, 1998;Desdevises et al, 2003), to partition the total variance in biomass responses.…”

BBMV: an R package for the estimation of macroevolutionary landscapes

“…To deal with this problem, the BBMV package calculates the likelihood of the FPK model over multiple clades by multiplying the likelihoods in all clades. This procedure is similar to one that was recently published for the BM model (Revell et al 2018). Functions exist to fit 1) a model in which all clades share the same macroevolutionary landscape and the same rate of evolution, 2) a model in which all clades share the same macroevolutionary landscape but have different rates of evolution, or 3) a model in which both the macroevolutionary landscape and the rate of evolution can vary between clades.…”