Soares André E R, Schrago Carlos G
Department of Genetics, Federal University of Rio de Janeiro, Rio de Janeiro 21941-590, Brazil.
J Theor Biol. 2015 Jan 7;364:31-9. doi: 10.1016/j.jtbi.2014.09.004. Epub 2014 Sep 11.
Although taxon sampling is commonly considered an important issue in phylogenetic inference, it is rarely considered in the Bayesian estimation of divergence times. In fact, the studies conducted to date have presented ambiguous results, and the relevance of taxon sampling for molecular dating remains unclear. In this study, we developed a series of simulations that, after six hundred Bayesian molecular dating analyses, allowed us to evaluate the impact of taxon sampling on chronological estimates under three scenarios of among-lineage rate heterogeneity. The first scenario allowed us to examine the influence of the number of terminals on the age estimates based on a strict molecular clock. The second scenario imposed an extreme example of lineage specific rate variation, and the third scenario permitted extensive rate variation distributed along the branches. We also analyzed empirical data on selected mitochondrial genomes of mammals. Our results showed that in the strict molecular-clock scenario (Case I), taxon sampling had a minor impact on the accuracy of the time estimates, although the precision of the estimates was greater with an increased number of terminals. The effect was similar in the scenario (Case III) based on rate variation distributed among the branches. Only under intensive rate variation among lineages (Case II) taxon sampling did result in biased estimates. The results of an empirical analysis corroborated the simulation findings. We demonstrate that taxonomic sampling affected divergence time inference but that its impact was significant if the rates deviated from those derived for the strict molecular clock. Increased taxon sampling improved the precision and accuracy of the divergence time estimates, but the impact on precision is more relevant. On average, biased estimates were obtained only if lineage rate variation was pronounced.
尽管分类群抽样通常被认为是系统发育推断中的一个重要问题，但在分歧时间的贝叶斯估计中却很少被考虑。事实上，迄今为止进行的研究结果并不明确，分类群抽样与分子定年的相关性仍不清楚。在本研究中，我们开展了一系列模拟，经过600次贝叶斯分子定年分析后，使我们能够在谱系间速率异质性的三种情况下评估分类群抽样对年代估计的影响。第一种情况使我们能够基于严格分子钟来研究终端数量对年龄估计的影响。第二种情况给出了谱系特异性速率变化的一个极端例子，第三种情况允许沿分支分布广泛的速率变化。我们还分析了所选哺乳动物线粒体基因组的实证数据。我们的结果表明，在严格分子钟情况下（情况I），分类群抽样对时间估计的准确性影响较小，尽管随着终端数量增加估计的精度更高。在基于分支间速率变化的情况（情况III）中效果类似。只有在谱系间存在强烈速率变化的情况下（情况II），分类群抽样才会导致有偏差的估计。实证分析结果证实了模拟结果。我们证明分类群抽样会影响分歧时间推断，但如果速率偏离严格分子钟得出的速率，其影响就会很显著。增加分类群抽样提高了分歧时间估计的精度和准确性，但对精度的影响更为重要。平均而言，只有当谱系速率变化显著时才会得到有偏差的估计。
