Evaluation of Ancestral Sequence Reconstruction Methods to Infer Nonstationary Patterns of Nucleotide Substitution

Abstract

Inference of gene sequences in ancestral species has been widely used to test hypotheses concerning the process of molecular sequence evolution. However, the approach may produce spurious results, mainly because using the single best reconstruction while ignoring the suboptimal ones creates systematic biases. Here we implement methods to correct for such biases and use computer simulation to evaluate their performance when the substitution process is nonstationary. The methods we evaluated include parsimony and likelihood using the single best reconstruction (SBR), averaging over reconstructions weighted by the posterior probabilities (AWP), and a new method called expected Markov counting (EMC) that produces maximum-likelihood estimates of substitution counts for any branch under a nonstationary Markov model. We simulated base composition evolution on a phylogeny for six species, with different selective pressures on G+C content among lineages, and compared the counts of nucleotide substitutions recorded during simulation with the inference by different methods. We found that large systematic biases resulted from (i) the use of parsimony or likelihood with SBR, (ii) the use of a stationary model when the substitution process is nonstationary, and (iii) the use of the Hasegawa-Kishino-Yano (HKY) model, which is too simple to adequately describe the substitution process. The nonstationary general time reversible (GTR) model, used with AWP or EMC, accurately recovered the substitution counts, even in cases of complex parameter fluctuations. We discuss model complexity and the compromise between bias and variance and suggest that the new methods may be useful for studying complex patterns of nucleotide substitution in large genomic data sets.

Faster‐X evolution: Theory and evidence from Drosophila

“…The numbers in brackets indicate the 95% confidence intervals, obtained by bootstrapping over genes. The branch length estimates were obtained by applying the AWP method of ancestral state reconstruction (Matsumoto, Akashi, & Yang, 2015) to the genome sequences of D. melanogaster, D. simulans and D. yakuba, as described by Jackson et al (2017). This method was chosen because it was found to be accurate in the face of departures from equilibrium.…”

Variation in the intensity of selection on codon bias over time causes contrasting patterns of base composition evolution in Drosophila

“…Use of maximum parsimony methods or maximum likelihood models that assume equilibrium base composition under such circumstances can lead to erroneous inferences although these two methods were used in many previous analyses of various Drosophila species (Akashi et al. 2007; Matsumoto et al. 2015).…”

Noncoding regions underpin avian bill shape diversification at macroevolutionary scales

“…In our model, however, are the effects of population size changes partially negated by a combined parametrizing of branches. Unless specifically accounted for, substitution rate models do not consider the effect of nonequilibria processes that could affect d N /d S estimates (Matsumoto et al 2015). For example, GCbiased gene conversion -described as the preferential conversion of 'A' or 'T' alleles to 'G' or 'C' during recombination induced repair -has been shown to significantly affect estimates of substitution rates, in particular at synonymous sites in birds (Galtier et al 2009;Weber et al 2014;Boĺivar et al 2016;Botero-Castro et al 2017;Corcoran et al 2017;Bolívar et al 2019).…”