Phylogenetic Comparative Methods: A User's Guide for Paleontologists

Abstract

Recent advances in statistical approaches called Phylogenetic Comparative Methods (PCMs) have provided paleontologists with a powerful set of analytical tools for investigating evolutionary tempo and mode in fossil lineages. However, attempts to integrate PCMs with fossil data often present workers with practical challenges or unfamiliar literature. In this paper, we present guides to the theory behind, and application of, PCMs with fossil taxa. Based on an empirical dataset of Paleozoic crinoids, we present example analyses to illustrate common applications of PCMs to fossil data, including investigating patterns of correlated trait evolution, and macroevolutionary models of morphological change. We emphasize the importance of accounting for sources of uncertainty, and discuss how to evaluate model fit and adequacy. Finally, we discuss several promising methods for modelling heterogenous evolutionary dynamics with fossil phylogenies. Integrating phylogeny-based approaches with the fossil record provides a rigorous, quantitative perspective to understanding key patterns in the history of life. Bapst 2014b). The dataset we use here is for fossil crinoids (Eucladida, Echinodermata), a morphologically diverse clade of marine invertebrates with a well-sampled fossil record. Our use of this dataset is primarily intended to demonstrate the different tree-based analytical tools that can be applied, rather than to glean specific inferences about crinoid macroevolution and we caution readers our results should be viewed in this light. Key packages in R that contain implementations of PCMs that are commonly applied to fossil data, are ape (standard format and processing for phylogenies in R, [Paradis et al. 2004]), nlme (fitting Gaussian models e.g. least squares regression, [Pinheiro et al. 2019]), geiger (versatile package that performs and plots many PCMs, [Harmon et al. 2008;Pennell et al. 2014]), phytools (additional plotting and simulation functions, [Revell 2012]) and OUwie (heterogeneous macroevolutionary model fitting e.g. Brownian Motion or Early Burst, [Beaulieu and O'Meara 2016]). There are many others, so it is valuable to spend time exploring the different tools that might be best suited to answering a particular question. Once a user has gained familiarity with the above key packages, a very extensive list of all the packages that can be used for phylogenetic approaches in R can be found at this website: https://cran.rproject.org/web/views/Phylogenetics.html.

Permian–Triassic phylogenetic and morphologic evolution of rhynchonellide brachiopods

“…The low posterior support of nodes (i.e., highly variable topology) is not uncommon in empirical datasets (O'Reilly and Donoghue 2018; Barido-Sottani et al 2020), and the tree topology and divergence times affect the results of phylogenetic comparative analysis (Bapst 2014; Bapst and Hopkins 2017). However, the downstream comparative analyses are not impossible if multiple trees, rather than a single point estimate of phylogeny, are analyzed (Wright et al 2015; Bapst et al 2016; Soul and Wright 2021).…”

“…In this study, many nodes recovered by the MCC tree are not well supported (see "Results"). To avoid utilizing the MCC tree as the only input in the following analysis, multiple posterior trees were also analyzed to account for the uncertainty of the tree topology and divergence times (Wright et al 2015(Wright et al , 2021Bapst et al 2016;Soul and Wright 2021).…”

Palatal morphology predicts the paleobiology of early salamanders

“…The remaining fossil taxa in our dataset with age ranges determined from published literature (Dataset in Dryad) and the Paleobiology Database ( https://www.paleobiodb.org ) were incorporated into this MCC tree based on a recent cladogram ( Jia et al, 2021a : Figure 6) we constructed for living and fossil cryptobranchoids ( Figure 1D ). Adding fossil taxa increases the accuracy for multivariate phylogenetic comparative analyses and ancestral state reconstruction (see below; Soul and Wright, 2021 ), but will inevitably bring zero-length branches when the age of internal nodes was considered equal to that of its immediate oldest descendant, resulting in difficulties in multivariate phylogenetic comparative analyses. To circumvent this problem, any zero-length branch in the cladogram was treated to have a same branch length with its first none zero-length ancestral branch by using the ‘equal’ dating method in the ‘DatePhylo’ function from the R package ‘strap’ (version 1.4; Bell et al, 2015 ).…”

Palatal morphology predicts the paleobiology of early salamanders

“…6) we constructed for living and fossil cryptobranchoids (Figure 1D). Adding fossil taxa increases the accuracy for multivariate phylogenetic comparative analyses and ancestral state reconstruction (see below; Soul and Wright, 2021), but will inevitably bring zero-length branches when the age of internal nodes 25 was considered equal to that of its immediate oldest descendant, resulting in difficulties in multivariate phylogenetic comparative analyses. To circumvent this problem, any zero-length branch in the cladogram was treated to have a same branch length with its first none zero-length ancestral branch by using the "equal" dating method in the "DatePhylo" function from the R package 'strap' (version 1.4; Bell and Lloyd, 2015).…”