School of Biological Sciences, Flinders University, PO Box 2100, Adelaide, South Australia 5001; School of Biological Sciences, University of Adelaide, Adelaide, South Australia 5005; and Earth Sciences Section, South Australian Museum, North Terrace, Adelaide 5000, Australia.
School of Biological Sciences, Flinders University, PO Box 2100, Adelaide, South Australia 5001; School of Biological Sciences, University of Adelaide, Adelaide, South Australia 5005; and Earth Sciences Section, South Australian Museum, North Terrace, Adelaide 5000, Australia School of Biological Sciences, Flinders University, PO Box 2100, Adelaide, South Australia 5001; School of Biological Sciences, University of Adelaide, Adelaide, South Australia 5005; and Earth Sciences Section, South Australian Museum, North Terrace, Adelaide 5000, Australia
Syst Biol. 2015 May;64(3):532-44. doi: 10.1093/sysbio/syv005. Epub 2015 Jan 22.
Virtually all models for reconstructing ancestral states for discrete characters make the crucial assumption that the trait of interest evolves at a uniform rate across the entire tree. However, this assumption is unlikely to hold in many situations, particularly as ancestral state reconstructions are being performed on increasingly large phylogenies. Here, we show how failure to account for such variable evolutionary rates can cause highly anomalous (and likely incorrect) results, while three methods that accommodate rate variability yield the opposite, more plausible, and more robust reconstructions. The random local clock method, implemented in BEAST, estimates the position and magnitude of rate changes on the tree; split BiSSE estimates separate rate parameters for pre-specified clades; and the hidden rates model partitions each character state into a number of rate categories. Simulations show the inadequacy of traditional models when characters evolve with both asymmetry (different rates of change between states within a character) and heterotachy (different rates of character evolution across different clades). The importance of accounting for rate heterogeneity in ancestral state reconstruction is highlighted empirically with a new analysis of the evolution of viviparity in squamate reptiles, which reveal a predominance of forward (oviparous-viviparous) transitions and very few reversals.
实际上，所有用于重建离散特征祖先状态的模型都做出了一个关键假设，即感兴趣的特征在整个树中以均匀的速率进化。然而，在许多情况下，这种假设不太可能成立，特别是在对越来越大的系统发育树进行祖先状态重建时。在这里，我们展示了未能解释这种可变进化率如何会导致高度异常（可能是不正确的）的结果，而三种适应率变异性的方法则产生了相反的、更合理和更稳健的重建。随机局部时钟方法（在 BEAST 中实现）估计树中率变化的位置和幅度；分裂 BiSSE 估计为预定义的分支分配单独的速率参数；隐藏速率模型将每个字符状态划分为多个速率类别。模拟表明，当字符以不对称（字符内状态变化的不同速率）和异速（不同分支的字符进化的不同速率）进化时，传统模型是不充分的。通过对蜥蜴类动物中胎生进化的新分析，从经验上强调了在祖先状态重建中考虑速率异质性的重要性，该分析揭示了向前（卵生胎生）转变的优势，以及很少有逆转。
