Impact of the Partitioning Scheme on Divergence Times Inferred from Mammalian Genomic Data Sets

Data partitioning has long been regarded as an important parameter for phylogenetic inference. The division of heterogeneous multigene data sets into partitions with similar substitution patterns is known to increase the performance of probabilistic phylogenetic methods. However, the effect of the partitioning scheme on divergence time estimates has generally been ignored. To investigate the impact of data partitioning on the estimation of divergence times, we have constructed two genomic data sets. The first one with 15 nuclear genes comprising 50,928 bp were selected from the OrthoMam database; the second set was composed of complete mitochondrial genomes. We studied two partitioning schemes: concatenated supermatrices and partitioned gene analysis. We have also measured the impact of taxonomic sampling on the estimates. After drawing divergence time inferences using the uncorrelated relaxed clock in BEAST, we have compared the age estimates between the partitioning schemes. Our results show that, in general, both schemes resulted in similar chronological estimates, however the concatenated data sets were more efficient than the partitioned ones in attaining suitable effective sample sizes.