%0 Journal Article
%T Undersampling Taxa Will Underestimate Molecular Divergence Dates: An Example from the South American Lizard Clade Liolaemini
%A James A. Schulte II
%J International Journal of Evolutionary Biology
%D 2013
%I Hindawi Publishing Corporation
%R 10.1155/2013/628467
%X Methods for estimating divergence times from molecular data have improved dramatically over the past decade, yet there are few studies examining alternative taxon sampling effects on node age estimates. Here, I investigate the effect of undersampling species diversity on node ages of the South American lizard clade Liolaemini using several alternative subsampling strategies for both time calibrations and taxa numbers. Penalized likelihood (PL) and Bayesian molecular dating analyses were conducted on a densely sampled (202 taxa) mtDNA-based phylogenetic hypothesis of Iguanidae, including 92 Liolaemini species. Using all calibrations and penalized likelihood, clades with very low taxon sampling had node age estimates younger than clades with more complete taxon sampling. The effect of Bayesian and PL methods differed when either one or two calibrations only were used with dense taxon sampling. Bayesian node ages were always older when fewer calibrations were used, whereas PL node ages were always younger. This work reinforces two important points: (1) whenever possible, authors should strongly consider adding as many taxa as possible, including numerous outgroups, prior to node age estimation to avoid considerable node age underestimation and (2) using more, critically assessed, and accurate fossil calibrations should yield improved divergence time estimates. 1. Introduction Alternative taxon sampling strategies are known to affect many facets of phylogenetic reconstruction [1¨C3]. Linder et al. [4] conducted one of the most thorough analyses assessing the impact of taxon sampling on divergence date estimation using resampling analyses. This study found that mean estimated ages of focal nodes were significantly younger with sparser taxon sampling than when taxa from the complete dataset were included. In addition, the more distant the focal node was from the calibration point, the more sensitive estimation effects were on nodal ages if the taxa were undersampled, especially for nonparametric rate-smoothing (NPRS) methods; but penalized likelihood (PL) and Bayesian methods also were prone to these effects especially if undersampling was large. Subsequent studies following up on these important results are lacking in the literature. This is especially important because numerous multigene studies estimate divergence times but often significantly undersample large clades using a single representative species in place of higher taxonomic groups. I argue that unless these large clades are densely sampled and calibration points are carefully chosen and spread
%U http://www.hindawi.com/journals/ijeb/2013/628467/