Premium
Phylogenetic eigenvector maps: a framework to model and predict species traits
Author(s) -
Guénard Guillaume,
Legendre Pierre,
PeresNeto Pedro
Publication year - 2013
Publication title -
methods in ecology and evolution
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 3.425
H-Index - 105
ISSN - 2041-210X
DOI - 10.1111/2041-210x.12111
Subject(s) - phylogenetic tree , trait , phylogenetics , evolutionary biology , biology , phylogenetic comparative methods , tree (set theory) , phylogenetic network , set (abstract data type) , computer science , mathematics , genetics , combinatorics , gene , programming language
Summary Phylogenetic signals are the legacy related to evolutionary processes shaping trait variation among species. Biologists can use these signals to tackle questions related to the evolutionary processes underlying trait evolution, estimate the ancestral state of a trait and predict unknown trait values from those of related species (i.e. ‘phylogenetic modelling’). Approaches to model phylogenetic signals rely on quantitative descriptors of the structures representing the consequences of evolution on trait differences among species. Here, we propose a novel framework to model phylogenetic signals: P hylogenetic E igenvectors M aps ( PEM ). PEM are a set of eigenfunctions obtained from the structure of a phylogenetic graph, which can be a standard phylogenetic tree or a phylogenetic tree with added reticulations. These eigenfunctions depict a set of potential patterns of phenotype variation among species from the structure of the phylogenetic graph. A subset of eigenfunctions from a PEM is selected for the purpose of predicting the phenotypic values of traits for species that are represented in a tree, but for which trait data are otherwise lacking. This paper introduces a comprehensive view and the computational details of the PEM framework (with calculation examples), a simulation study to demonstrate the ability of PEM to predict trait values and four real data examples of the use of the framework. Simulation results show that PEM are robust in representing phylogenetic signal and in estimating trait values. The method also performed well when applied to the real‐world data: prediction coefficients were high (0·76–0·88), and no notable model biases were found. Phylogenetic modelling using PEM is shown to be a useful methodological asset to disciplines such as ecology, ecophysiology, ecotoxicology, pharmaceutical botany, among others, which can benefit from estimating trait values that are laborious and often expensive to obtain.