Accuracy of phylogenetic reconstructions from continuous characters analyzed under parsimony and its parametric correlates

Abstract

Quantitative traits are a source of evolutionary information often difficult to handle in cladistics. Tools exist to analyze this kind of data without subjective discretization, avoiding biases in the delimitation of categorical states. Nonetheless, the ability of continuous characters to accurately infer relationships is incompletely understood, particularly under parsimony analysis. This study evaluates the accuracy of phylogenetic reconstructions from simulated matrices of continuous characters evolving under alternative evolutionary processes and analyzed by parsimony. We generated 100 trees to simulate 9,000 matrices containing 26 terminals and 100 continuous characters evolving under: Brownian-Motion (BM), Ornstein-Uhlenbeck (OU) and Early-Burst (EB) processes assuming variable parametrizations. Our comparisons of cladograms revealed that matrices analyzed by parsimony carry phylogenetic signals to infer relationships, but the accuracy is higher for matrices simulated under BM, regardless of the parameterization schemes. Implementation of equal or implied weighting with multiple penalization strengths against homoplasies did not affect cladogram inferences. Accuracy of continuous characters in resolving relationships is skewed toward apical nodes of the trees. Our simulations provide controlled tests of the usefulness of quantitative traits in phylogenetics, specifically under neutral evolution, and demonstrate their effectiveness in estimating shallower nodes among recently diverged species, regardless of parameters and weighting schemes.