Sources of Signal in 62 Protein-Coding Nuclear Genes for Higher-Level Phylogenetics of Arthropods

BackgroundThis study aims to investigate the strength of various sources of phylogenetic information that led to recent seemingly robust conclusions about higher-level arthropod phylogeny and to assess the role of excluding or downweighting synonymous change for arriving at those conclusions.Methodology/Principal FindingsThe current study analyzes DNA sequences from 68 gene segments of 62 distinct protein-coding nuclear genes for 80 species. Gene segments analyzed individually support numerous nodes recovered in combined-gene analyses, but few of the higher-level nodes of greatest current interest. However, neither is there support for conflicting alternatives to these higher-level nodes. Gene segments with higher rates of nonsynonymous change tend to be more informative overall, but those with lower rates tend to provide stronger support for deeper nodes. Higher-level nodes with bootstrap values in the 80% – 99% range for the complete data matrix are markedly more sensitive to substantial drops in their bootstrap percentages after character subsampling than those with 100% bootstrap, suggesting that these nodes are likely not to have been strongly supported with many fewer data than in the full matrix. Data set partitioning of total data by (mostly) synonymous and (mostly) nonsynonymous change improves overall node support, but the result remains much inferior to analysis of (unpartitioned) nonsynonymous change alone. Clusters of genes with similar nonsynonymous rate properties (e.g., faster vs. slower) show some distinct patterns of node support but few conflicts. Synonymous change is shown to contribute little, if any, phylogenetic signal to the support of higher-level nodes, but it does contribute nonphylogenetic signal, probably through its underlying heterogeneous nucleotide composition. Analysis of seemingly conservative indels does not prove useful.ConclusionsGenerating a robust molecular higher-level phylogeny of Arthropoda is currently possible with large amounts of data and an exclusive reliance on nonsynonymous change.

研究背景
本研究旨在探究支撑近期节肢动物高阶系统发育看似稳健结论的各类系统发育信息来源的强度，并评估排除或降低同义突变（synonymous change）权重对获得上述结论的作用。

研究方法与主要结果
本研究对80个物种的62个独立编码核蛋白基因的68个基因片段的DNA序列展开分析。单独分析的基因片段能够支持联合基因分析中得到的诸多分支节点，但鲜有能够支持当前学界最关注的高阶分支节点。不过目前也并无证据支持与这些高阶分支节点相悖的替代假说。整体而言，非同义突变（nonsynonymous change）速率更高的基因片段通常具备更强的系统发育信息价值，而速率较低的基因片段则更能为深层分支节点提供有力支持。完整数据矩阵中自展值（bootstrap）处于80%~99%区间的高阶分支节点，在特征抽采样后其自展百分比出现显著下降的敏感性显著高于自展值为100%的节点，这表明若数据量远少于完整矩阵，此类节点大概率无法获得强有力的支持。将总数据按（以同义突变为主要特征）和（以非同义突变为主要特征）进行数据集分区，可提升整体分支节点的支持度，但该结果仍远不及仅对（未分区的）非同义突变进行分析的效果。具备相似非同义突变速率特征（如高速率与低速率组）的基因簇，在分支节点支持度上呈现出部分独特模式，但几乎不存在冲突。研究表明，同义突变几乎无法为高阶分支节点的支持提供系统发育信号，但却会引入非系统发育信号，这一现象大概率源于其内在的核苷酸组成异质性。对看似保守的插入缺失（indels）位点的分析并未展现出实用价值。

研究结论
当前，借助海量数据并仅依托非同义突变进行分析，即可构建出稳健的节肢动物高阶分子系统发育树。