Effects of phylogenetic reconstruction method on the robustness of species delimitation using single-locus data.

Aelys M. Humphreys,Cuong Q. Tang,Emmanuel Paradis,Timothy G. Barraclough,Diego Fontaneto

doi:10.1111/2041-210x.12246

Abstract

Coalescent-based species delimitation methods combine population genetic and phylogenetic theory to provide an objective means for delineating evolutionarily significant units of diversity. The generalised mixed Yule coalescent (GMYC) and the Poisson tree process (PTP) are methods that use ultrametric (GMYC or PTP) or non-ultrametric (PTP) gene trees as input, intended for use mostly with single-locus data such as DNA barcodes. Here, we assess how robust the GMYC and PTP are to different phylogenetic reconstruction and branch smoothing methods. We reconstruct over 400 ultrametric trees using up to 30 different combinations of phylogenetic and smoothing methods and perform over 2000 separate species delimitation analyses across 16 empirical data sets. We then assess how variable diversity estimates are, in terms of richness and identity, with respect to species delimitation, phylogenetic and smoothing methods. The PTP method generally generates diversity estimates that are more robust to different phylogenetic methods. The GMYC is more sensitive, but provides consistent estimates for BEAST trees. The lower consistency of GMYC estimates is likely a result of differences among gene trees introduced by the smoothing step. Unresolved nodes (real anomalies or methodological artefacts) affect both GMYC and PTP estimates, but have a greater effect on GMYC estimates. Branch smoothing is a difficult step and perhaps an underappreciated source of bias that may be widespread among studies of diversity and diversification. Nevertheless, careful choice of phylogenetic method does produce equivalent PTP and GMYC diversity estimates. We recommend simultaneous use of the PTP model with any model-based gene tree (e.g. RAxML) and GMYC approaches with BEAST trees for obtaining species hypotheses.

Highlights

Species are a fundamental unit for many fields of biology, yet their identification and delimitation are rarely straightforward (Hebert et al 2003)
We assess how robust the generalised mixed Yule coalescent (GMYC) and Poisson tree process (PTP) are to different phylogenetic reconstruction and branch smoothing methods
The reduced number of analyses for the MM-GMYC is due to the method not accommodating trees with unresolved nodes without manual input, which would not have been achievable within the scope of the present study

Summary

Introduction

Species are a fundamental unit for many fields of biology, yet their identification and delimitation are rarely straightforward (Hebert et al 2003). A special branch of phylogenetic species delimitation (see Sites & Marshall 2003) is coalescent-based species delimitation methods (Pons et al 2006; Fontaneto et al 2007; Zhang et al 2013), which combine coalescent theory with diversification models to infer the transition point between population and species-level processes on a gene tree These approaches provide objective, clade-specific threshold(s) with which to delimit evolutionarily significant units (ESUs) of diversity (akin to species, as defined by the Evolutionary Species Concept – Simpson 1951). These methods provide an alternative to operational taxonomic unit (OTU) picking methods, which rely on arbitrary, clade-specific sequence similarity thresholds (Barraclough et al 2009)

Methods

Results

Discussion

Conclusion