Phylogenetic Signal and Noise: Predicting the Power of a Data Set to Resolve Phylogeny

Abstract

A principal objective for phylogenetic experimental design is to predict the power of a data set to resolve nodes in a phylogenetic tree. However, proactively assessing the potential for phylogenetic noise compared with signal in a candidate data set has been a formidable challenge. Understanding the impact of collection of additional sequence data to resolve recalcitrant internodes at diverse historical times will facilitate increasingly accurate and cost-effective phylogenetic research. Here, we derive theory based on the fundamental unit of the phylogenetic tree, the quartet, that applies estimates of the state space and the rates of evolution of characters in a data set to predict phylogenetic signal and phylogenetic noise and therefore to predict the power to resolve internodes. We develop and implement a Monte Carlo approach to estimating power to resolve as well as deriving a nearly equivalent faster deterministic calculation. These approaches are applied to describe the distribution of potential signal, polytomy, or noise for two example data sets, one recent (cytochrome c oxidase I and 28S ribosomal rRNA sequences from Diplazontinae parasitoid wasps) and one deep (eight nuclear genes and a phylogenomic sequence for diverse microbial eukaryotes including Stramenopiles, Alveolata, and Rhizaria). The predicted power of resolution for the loci analyzed is consistent with the historic use of the genes in phylogenetics.

The impact of incorporating molecular evolutionary model into predictions of phylogenetic signal and noise

“…The four-taxon tree has three possible tip-labeled subtrees, all of which can be supported by an AABB pattern, but only one of the three subtrees matches the actual four-taxon tree topology (c.f. Figure 1 in Townsend et al, 2012). Any other pattern yields no net contribution per site to phylogeny resolution under maximum parsimony and very minimal impact per site under ML analysis and Bayesian methods; thus any non-AABB pattern is conservatively assumed to make no contribution away from a polytomy by our approach.…”

“…As in Townsend et al (2012), we model phylogenetic signal and noise as accurate and inaccurate parsimony-informative sites for resolution of a four-taxon tree. A site is considered parsimonyinformative for resolving the four-taxon tree if it exhibits an AABB pattern of character states at the branch leaves, meaning that at the given site, two of the branch leaves feature an identical character state (suggesting these two branches belong to the same clade) while the other two branch leaves share a different character state.…”

“…Susko (2011) calculated via large sequence-length approximations the probabilities of correct phylogenetic reconstruction for ML estimation for a single poorly resolved internode. For a proposed discrete data set, Townsend et al (2012) applied empirical estimates of site substitution rates and an s-state Poisson model of molecular evolution (s ≥ 2) to predict the probabilities of a data set leading to correct, incorrect, and polytomous resolution of a quartet-taxon phylogeny. This analytical approach makes it possible to address the unresolved question of whether choice in model selection affects predictions of phylogenetic informativeness.…”

“…Not only does phylogenetic informativeness depend critically on estimates of the rate of evolution of the characters to be applied to a phylogenetic problem (Graybeal, 1994;Naylor and Brown, 1998;Yang, 1998;Rokas and Holland, 2000;Townsend, 2007;Townsend and Leuenberger, 2011;Townsend et al, 2012), but also the impact of characters on inference would likely depend upon the model of evolution to which they presumably adhere. To investigate this question, we generalize the Townsend et al (2012) phylogenetic signal and noise analysis based on the s-state Poisson model, to incorporate any time-reversible homogenous Markov model of nucleotide substitution. This generalization allows the signal and noise analysis to incorporate more realistic, higher parameterized assumptions of molecular evolution than the s-state Poisson model.…”

Sequence analysis of Maturase K (matK): A chloroplast-encoding gene in some selected pulses

“…According to Graham and Olmstead (2000), Wang et al (2009, rapidly evolving regions will be better used for shallow evolutionary histories while slowly evolving regions for deeper epochs. Their argument was premised on the fact that multiple hits confounded by extended time scale could be significant enough to conceal phylogenetic signals and elevate homoplasy, with saturation reaching levels that can negatively impact tree structure (Graybeal, 1994;Wenzel and Siddal, 1999;Klopfstein et al, 2010;Townsend et al, 2012). The accumulation of multiple hits in rapidly evolving regions is capable of obscuring potential synapomorphies as well as results in long branch alteration (Townsend, 2007;Magallon and Sanderson, 2002).On the contrary, however, the opposing school of thought opined that rapidly evolving regions promotes effectiveness and less constrained genomic regions in deep level phylogenetics (Yang, 1998;Hilu et al, 2008;Hilu and Liang, 1997;Worberg et al, 2007).According to Hilu et al (2014), phylogenetic signal from rapidly evolving and un-constrained matK provides by far the most structure and accuracy, whereas slowly evolving, constrained and un-constrained genes display decreasing degrees of informativeness and tree structure.…”