A novel method for jointly modeling the evolution of discrete and continuous traits

Abstract

The correlated evolution of multiple characters is a crucial aspect of evolutionary change. If change in a particular character influences the evolution of a separate trait, then modeling these features independently can mislead our understanding of the evolutionary process. Progress toward jointly modeling several characters has involved modeling multivariate evolution of the same class of character, but there are far fewer options when jointly modeling traits when one character is discrete and the other is continuous. Here, we develop such a framework to explicitly estimate the joint likelihood for discrete and continuous characters. Specifically, our model combines the probability of observing the continuous character under a generalized OU process with the probability of the discrete character under a hidden Markov model, linked by a shared underlying regime. We use simulation studies to demonstrate that this approach, hOUwie, can accurately evaluate parameter values across a broad set of models. We then apply our model to test whether fleshy and dry fruits of Ericaceae lineages are correlated with their climatic niche evolution as represented by the aridity index. Consistent with expectations, we find that the climatic niche of lineages with fleshy fruits is more conserved while lineages with dry fruits have higher rates of climatic niche evolution and a more humid climatic optimum.

Keywords: Ornstein-Uhlenbeck; continuous character; discrete character; hidden rates; phylogenetic comparative methods; rate variation.