Linking phenotypic and genotypic variation: a relaxed phylogenetic approach using the probabilistic programming language Stan

Abstract

AbstractPhyloG2P methods link genotype and phenotype by integrating evidence from across a phylogeny. I introduce a Bayesian approach to jointly modelling a continuous trait and a multiple sequence alignment, given a background tree and substitution rate matrix. The aim is to ask whether faster sequence evolution is linked to faster phenotypic evolution. Per-branch substitution rate multipliers (for the alignment) are linked to per-branch variance rates of a Brownian diffusion process (for the trait) via the flexible logistic function. The Brownian diffusion process can evolve on the same tree used to describe the alignment, or on a second tree, for example a tree with branch lengths in units of time. Simulation studies suggest the model can be well estimated using relatively short alignments and reasonably sized trees. An application of the model in both its one-tree and two-tree variants is provided as an example. Notably, the method is implemented concisely using the general-purpose probabilistic programming language Stan.