Divergence dating using mixed effects clock modelling: An application to HIV-1
Divergence dating using mixed effects clock modelling: An application to HIV-1
复制标题
DOI：
10.1093/ve/vez036
复制
发表时间：
2019-07-01
期刊：
影响因子：
5.3
通讯作者：
Lemey, Philippe
中科院分区：
文献类型：
--
作者：
Bletsa, Magda;Suchard, Marc A.;Lemey, Philippe
在各种替代率变异来源存在的情况下，需要估计进化历史中的分歧时间，这刺激了松弛分子时钟模型的丰富发展。病毒进化研究经常采用不相关的时钟模型作为一般的松弛分子时钟过程，但如果支系或谱系之间存在离散的速率差异，这可能会带来相当大的估计偏差。对于HIV-1 M组，亚型之间的比率差异已被证明导致与最近共同祖先估计的时间不一致。尽管这使现有的分子测年方法的充分性受到质疑，但到目前为止还没有提出解决这个问题的方法。在这里，我们研究了混合效应分子钟模型的使用，该模型结合了进化速率中的固定和随机效应，来估计发散时间。通过模拟，我们证明了该模型在贝叶斯框架下的性能优于现有的分子时钟模型，在速度变化的混合来源存在的情况下，该模型可以估计时间测量的系统发育，同时在更简单的场景中也保持了良好的性能。通过分析一个全面的HIV-1 M组完整基因组数据集，我们确认了亚型之间相当大的速率差异，这是不相关的松弛时钟模型所不能充分模拟的。混合效应时钟模型可以适应这种速率变化，并导致到1920年(1915-25年)HIV-1组M的最新共同祖先的时间，仅略早于相同数据集的不相关松弛时钟估计。使用完整的基因组数据似乎比分子时钟模型有更深远的影响，因为与基于短包络基因序列的类似估计相比，它将可信间隔减少了50%。
The need to estimate divergence times in evolutionary histories in the presence of various sources of substitution rate variation has stimulated a rich development of relaxed molecular clock models. Viral evolutionary studies frequently adopt an uncorrelated clock model as a generic relaxed molecular clock process, but this may impose considerable estimation bias if discrete rate variation exists among clades or lineages. For HIV-1 group M, rate variation among subtypes has been shown to result in inconsistencies in time to the most recent common ancestor estimation. Although this calls into question the adequacy of available molecular dating methods, no solution to this problem has been offered so far. Here, we investigate the use of mixed effects molecular clock models, which combine both fixed and random effects in the evolutionary rate, to estimate divergence times. Using simulation, we demonstrate that this model outperforms existing molecular clock models in a Bayesian framework for estimating time-measured phylogenies in the presence of mixed sources of rate variation, while also maintaining good performance in simpler scenarios. By analysing a comprehensive HIV-1 group M complete genome data set we confirm considerable rate variation among subtypes that is not adequately modelled by uncorrelated relaxed clock models. The mixed effects clock model can accommodate this rate variation and results in a time to the most recent common ancestor of HIV-1 group M of 1920 (1915-25), which is only slightly earlier than the uncorrelated relaxed clock estimate for the same data set. The use of complete genome data appears to have a more profound impact than the molecular clock model because it reduces the credible intervals by 50 per cent relative to similar estimates based on short envelope gene sequences.