Variation in gene tree estimates is widely observed in empirical phylogenomic data and is often assumed to be the result of biological processes. However, a recent study using tetrapod mitochondrial genomes to control for biological sources of variation due to their haploid, uniparentally inherited, and non-recombining nature found that levels of discordance among mitochondrial gene trees were comparable to those found in studies that assume only biological sources of variation. Additionally, they found that several of the models of sequence evolution chosen to infer gene trees were doing an inadequate job fitting the sequence data. These results indicated that significant amounts of gene tree discordance in empirical data may be due to poor fit of sequence evolution models, and that more complex and biologically realistic models may be needed. To test how the fit of sequence evolution models relates to gene tree discordance, we analyzed the same mitochondrial datasets as the previous study using two additional, more complex models of sequence evolution that each includes a different biologically realistic aspect of the evolutionary process: a covarion model to incorporate site-specific rate variation across lineages (heterotachy), and a partitioned model to incorporate variable evolutionary patterns by codon position. Our results show that both additional models fit the data better than the models used in the previous study, with the covarion being consistently and strongly preferred as tree size increases. However, even these more preferred models still inferred highly discordant mitochondrial gene trees, thus deepening the mystery around what we label the “Mito-Phylo Paradox” and leading us to ask whether the observed variation could, in fact, be biological in nature after all.

中文翻译：

---

序列进化的复杂模型改善了四足动物有丝分裂基因组的拟合，但没有改善基因树不一致


基因树估计的变异在经验系统发育数据中被广泛观察到，并且通常被认为是生物过程的结果。然而，最近的一项研究使用四足动物线粒体基因组来控制由于其单倍体、单亲遗传和非重组性质而导致的生物变异来源，发现线粒体基因树之间的不一致水平与仅假设生物变异来源的研究中发现的水平相当。此外，他们发现，选择用于推断基因树的几个序列进化模型在拟合序列数据方面做得不够。这些结果表明，经验数据中大量的基因树不一致可能是由于序列进化模型的拟合不佳，并且可能需要更复杂和生物学上真实的模型。为了测试序列进化模型的拟合与基因树不一致的关系，我们使用两个额外的、更复杂的序列进化模型分析了与之前研究相同的线粒体数据集，每个模型都包含进化过程的不同生物学现实方面：一个协变量模型，用于合并跨谱系的位点特异性速率变化（异种）， 以及一个分区模型，用于按密码子位置整合可变的进化模式。我们的结果表明，这两个附加模型都比之前研究中使用的模型更符合数据，随着树木大小的增加，协变量始终受到强烈偏好。 然而，即使是这些更受欢迎的模型仍然推断出高度不一致的线粒体基因树，从而加深了我们称之为“线粒体-叶系悖论”的谜团，并引导我们质疑观察到的变异是否实际上在本质上可能是生物的。