Performance of genomic data sets on the estimation of the divergence time of New World and Old World anthropoids

The origin of New World anthropoids has received renewed attention since the advent of molecular dating methods that relax the assumption of a strict molecular clock. However, the studies conducted to date have estimated the time of the separation of New World and Old World anthropoids at values as different as 70 and 22 Ma. With the aim of investigating the source of the discrepancies in the inferred ages, we have compared the performance of mitochondrial and nuclear markers in two pairs of datasets. We show that in the larger genomic samples, the dates of the separation of New and Old World anthropoids estimated from nuclear and mitochondrial data are significantly different. The precision of the estimates demonstrated that both markers rendered significantly different estimates. However, parametric estimates from the large nuclear dataset were highly cross-correlated. Cross-correlation of absolute divergence times and evolutionary rates was as great as -96%. Consequently, the age estimates from the large nuclear data were not reproducible, because Markov chains were unable to reach the same parametric values independently, even with the adoption of additional information from calibration priors. Thus, because branch length decomposition was not achieved, a comparison of the genomic age estimates from nuclear and mitochondrial datasets was statistically impractical. We demonstrate the importance of examining the output of Markov chain Monte Carlo analyses for correlation between rate and time in studies that use phylogenomic datasets to examine the chronological scales of primate evolution.