Data from: The identification of the closest living relative(s) of tetrapods: phylogenomic lessons for resolving short ancient internodes

Identifying the closest living relative(s) of tetrapods is an important, yet still contested question in vertebrate phylogenetics. Three hypotheses are possible and ruling out alternatives has proven difficult even with large molecular data sets due to weak phylogenetic signal coupled non-phylogenetic noise resulting from relatively rapid speciation events that occurred a long time ago (>400 Ma.). Here, we revisit the identity of the closest living relative(s) of land vertebrates from a phylogenomic perspective and include new genomic data for all extant lungfish genera. RNA-seq proves to be a great alternative to genomic sequencing, which currently is technically not feasible in lungfishes due to their huge (50-130 Gb) and repetitive genomes. We examined the most important sources of systematic error, namely long-branch attraction, compositional heterogeneity and distribution of missing data and applied different correction techniques. A multispecies coalescent approach is used to account for deep coalescence that might come from the short and deep internodes separating early sarcopterygian splits. Concatenation methods favored lungfishes as the closest living relatives of tetrapods with strong statistical support. Amino acid profile mixture models can unambiguously resolve this difficult internode thanks to their ability of reducing systematic error. We assessed the performance of different site-heterogeneous models and data partitioning and compared the ability of different strategies designed to overcome long-branch attraction, including taxon manipulation, reduction of among-lineage rate heterogeneity and removal of fast-evolving or compositionally heterogeneous positions. The identification of lungfish as sister group of tetrapods is robust regarding the effects of non-stationary composition and distribution of missing data. The multispecies coalescent method reconstructed strongly supported topologies that were congruent with concatenation, despite pervasive gene tree heterogeneity. We reject alternative topologies for early sarcopterygian relationships by increasing the signal-to-noise ratio in our alignments. The analytical pipeline outlined here combines probabilistic phylogenomic inference with methods for evaluating data quality, model adequacy and assessing systematic error, and thus is likely to help resolve similarly difficult internodes in the tree of life.确定四足动物（tetrapods）现存最近亲缘类群是脊椎动物系统发育学中一个重要且至今仍存在争议的问题。目前存在三种假说，而即便借助大型分子数据集，排除其他备选假说也颇具难度——这是因为距今超过4亿年（>400 Ma.）的快速物种形成事件所产生的系统发育信号较弱，同时伴随非系统发育噪声。本研究从系统发育组学视角重新探讨陆生脊椎动物现存最近亲缘类群的身份，并纳入所有现存肺鱼属的全新基因组数据。RNA测序（RNA-seq）已被证明是基因组测序的极佳替代方案：当前由于肺鱼拥有庞大（50~130 Gb）且重复序列丰富的基因组，其基因组测序在技术上仍难以实现。我们考察了最为关键的系统误差来源，即长枝吸引（long-branch attraction）、组成异质性以及缺失数据的分布情况，并应用了多种校正技术。我们采用多物种溯祖（multispecies coalescent）方法，以应对肉鳍鱼类（sarcopterygian）早期分化所形成的短而深的内部节点可能带来的深度溯祖问题。串联法在较强统计支持度下，将肺鱼推断为四足动物的现存最近亲缘类群。氨基酸谱混合模型（amino acid profile mixture models）能够有效降低系统误差，因此可以明确解析这一棘手的内部节点。我们评估了不同位点异质性模型与数据分区策略的表现，并对比了多种旨在克服长枝吸引的策略的效果，包括类群操作、降低支系间速率异质性，以及移除快速进化或组成异质性位点。在非平稳组成与缺失数据分布的影响下，将肺鱼确定为四足动物姊妹群的结果具有稳健性。尽管普遍存在基因树异质性，但多物种溯祖方法重建出了与串联法一致且具有强统计支持度的拓扑结构。我们通过提升序列比对（alignments）中的信噪比，否定了肉鳍鱼类早期分化关系的其他拓扑结构。本研究提出的分析流程将概率性系统发育组学推断与序列质量评估、模型适配性检验以及系统误差评估方法相结合，有望解决生命之树中其他类似的棘手内部节点问题。