Ascarrunz Eduardo, Claude Julien, Joyce Walter G
Department of Geosciences, University of Fribourg, Fribourg, Switzerland.
Institut des Sciences de l'Évolution de Montpellier, UMR UM/CNRS/IRD/EPHE, Montpellier, France.
PeerJ. 2019 Aug 22;7:e7476. doi: 10.7717/peerj.7476. eCollection 2019.
In the last 20 years, a general picture of the evolutionary relationships between geoemydid turtles (ca. 70 species distributed over the Northern hemisphere) has emerged from the analysis of molecular data. However, there is a paucity of good traditional morphological characters that correlate with the phylogeny, which are essential for the robust integration of fossil and molecular data. Part of this problem might be due to intrinsic limitations of traditional discrete characters. Here, we explore the use of continuous data in the form of 3D coordinates of homologous landmarks on the turtle shell for phylogenetic inference and the phylogenetic placement of single species on a scaffold molecular tree. We focus on the performance yielded by sampling the carapace and/or plastral lobes and using various phylogenetic methods.
We digitised the landmark coordinates of the carapace and plastron of 42 and 46 extant geoemydid species, respectively. The configurations were superimposed and we estimated the phylogenetic tree of geoemydids with landmark analysis under parsimony, traditional Farris parsimony, unweighted squared-change parsimony, maximum likelihood with a Brownian motion model, and neighbour-joining on a matrix of pairwise Procrustes distances. We assessed the performance of those analyses by comparing the trees against a reference phylogeny obtained from seven molecular markers. For comparisons between trees we used difference measures based on quartets and splits. We used the same reference tree to evaluate phylogenetic placement performance by a leave-one-out validation procedure.
Whatever method we used, similarity to the reference phylogeny was low. The carapace alone gave slightly better results than the plastron or the complete shell. Assessment of the potential for placement of single species on the reference tree with landmark data gave much better results, with similar accuracy and higher precision compared to the performance of discrete characters with parsimony.
在过去20年中，通过对分子数据的分析，地龟科龟类（约70种，分布于北半球）之间的进化关系已大致明了。然而，与系统发育相关的优质传统形态特征却很匮乏，而这些特征对于可靠整合化石和分子数据至关重要。部分问题可能源于传统离散特征的内在局限性。在此，我们探索以龟壳上同源地标三维坐标形式的连续数据用于系统发育推断，以及将单个物种在支架分子树上进行系统发育定位。我们聚焦于对背甲和/或腹甲叶进行采样并使用各种系统发育方法所产生的效果。
我们分别数字化了42种和46种现存地龟科物种背甲和腹甲的地标坐标。将这些构型进行叠加，并通过简约法、传统法里斯简约法、非加权平方变化简约法、布朗运动模型的最大似然法以及基于成对普氏距离矩阵的邻接法，利用地标分析来估计地龟科的系统发育树。我们通过将这些树与从七个分子标记获得的参考系统发育树进行比较，来评估这些分析的效果。对于树之间的比较，我们使用基于四重奏和分裂的差异度量。我们使用相同的参考树，通过留一法验证程序来评估系统发育定位性能。
无论我们使用何种方法，与参考系统发育树的相似度都很低。仅背甲给出的结果略优于腹甲或完整龟壳。用地标数据评估单个物种在参考树上的定位潜力，得到的结果要好得多，与简约法离散特征的性能相比，具有相似的准确性和更高的精确性。
