Estimating divergence dates from molecular sequences

The ability to date the time of divergence between lineages using moleculardata provides the opportunity to answer many important questions inevolutionary biology. However, molecular dating techniques have previouslybeen criticized for failing to adequately account for variation in the rateof molecular evolution. We present a maximum- likelihood approach toestimating divergence times that deals explicitly with the problem of ratevariation. This method has many advantages over previous approachesincluding the following: (1) a rate constancy test excludes data for whichrate heterogeneity is detected; (2) date estimates are generated withconfidence intervals that allow the explicit testing of hypothesesregarding divergence times; and (3) a range of sequences and fossil datesare used, removing the reliance on a single calculated calibration rate. Wepresent tests of the accuracy of our method, which show it to be robust tothe effects of some modes of rate variation. In addition, we test theeffect of substitution model and length of sequence on the accuracy of thedating technique. We believe that the method presented here offerssolutions to many of the problems facing molecular dating and provides aplatform for future improvements to such analyses.