On the interpretation of results returned by branch-site models
- Published
- Accepted
- Subject Areas
- Bioinformatics, Evolutionary Studies
- Keywords
- Codon models, Branch-site models, Positive selection
- Copyright
- © 2015 Guindon
- Licence
- This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, reproduction and adaptation in any medium and for any purpose provided that it is properly attributed. For attribution, the original author(s), title, publication source (PeerJ PrePrints) and either DOI or URL of the article must be cited.
- Cite this article
- 2015. On the interpretation of results returned by branch-site models. PeerJ PrePrints 3:e944v1 https://doi.org/10.7287/peerj.preprints.944v1
Abstract
In a recent study, Murrell et al. (2015) compared the performance of several branch-site models of codon evolution. Their interpretation of results published by Lu & Guindon (2014) suggests that the stochastic branch-site model implemented in the software fitmodel is anti-conservative altogether, i.e., positive selection is detected more often than expected when analyzing sequences evolving under a mixture of neutrality and negative selection. I argue here that this presentation of the performance of fitmodel is misleading and should not deter evolutionary biologists from using this approach in exploratory analyses of selection patterns at the molecular level.
Author Comment
In a recent article, Murrell and colleagues (Murrell et al., "Gene-Wide Identification of Episodic Selection", Mol. Biol. Evol., 2015) argue that the stochastic branch-site model implemented in fitmodel tends to find traces of positive selection from the analysis of coding sequences more often than expected. Here, I argue that their conclusions are exaggerated. Correct interpretation of the results derived from codon-based substitution model should not solely rely on likelihood ratio test for model comparison and careful interpretation of the parameter estimates is required. Examination of the results reported by Murrel et al. (2015) shows in fact that the branch-site model behaves fairly well, even in the least favorable cases where a large proportion of neutrally evolving sites and lineages can be interpreted as evidence for slight positive selection.