Estimating evolutionary rates using time-structured data: a general comparison of phylogenetic methods

Sebastián Duchêne,Jemma L. Geoghegan,Edward C. Holmes,Simon Y.W. Ho
DOI: https://doi.org/10.1093/bioinformatics/btw421
IF: 5.8
2016-07-13
Bioinformatics
Abstract:MOTIVATION: In rapidly evolving pathogens, including viruses and some bacteria, genetic change can accumulate over short time-frames. Accordingly, their sampling times can be used to calibrate molecular clocks, allowing estimation of evolutionary rates. Methods for estimating rates from time-structured data vary in how they treat phylogenetic uncertainty and rate variation among lineages. We compiled 81 virus data sets and estimated nucleotide substitution rates using root-to-tip regression, least-squares dating and Bayesian inference.RESULTS: Although estimates from these three methods were often congruent, this largely relied on the choice of clock model. In particular, relaxed-clock models tended to produce higher rate estimates than methods that assume constant rates. Discrepancies in rate estimates were also associated with high among-lineage rate variation, and phylogenetic and temporal clustering. These results provide insights into the factors that affect the reliability of rate estimates from time-structured sequence data, emphasizing the importance of clock-model testing.CONTACT: sduchene@unimelb.edu.au or garzonsebastian@hotmail.comSupplementary information: Supplementary data are available at Bioinformatics online.
biochemical research methods,biotechnology & applied microbiology,mathematical & computational biology
What problem does this paper attempt to address?