Premium
Optimal data partitioning, multispecies coalescent and Bayesian concordance analyses resolve early divergences of the grape family (Vitaceae)
Author(s) -
Lu Limin,
Cox Cymon J.,
Mathews Sarah,
Wang Wei,
Wen Jun,
Chen Zhiduan
Publication year - 2018
Publication title -
cladistics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.323
H-Index - 92
eISSN - 1096-0031
pISSN - 0748-3007
DOI - 10.1111/cla.12191
Subject(s) - coalescent theory , vitaceae , tree (set theory) , biology , bayesian probability , supertree , set (abstract data type) , concordance , computer science , data set , supermatrix , artificial intelligence , evolutionary biology , machine learning , mathematics , phylogenetic tree , gene , bioinformatics , genetics , vitis vinifera , combinatorics , current algebra , algebra over a field , affine lie algebra , horticulture , pure mathematics , programming language
Abstract Evolutionary rate heterogeneity and rapid radiations are common phenomena in organismal evolution and represent major challenges for reconstructing deep‐level phylogenies. Here we detected substantial conflicts in and among data sets as well as uncertainty concerning relationships among lineages of Vitaceae from individual gene trees, supernetworks and tree certainty values. Congruent deep‐level relationships of Vitaceae were retrieved by comprehensive comparisons of results from optimal partitioning analyses, multispecies coalescent approaches and the Bayesian concordance method. We found that partitioning schemes selected by PartitionFinder were preferred over those by gene or by codon position, and the unpartitioned model usually performed the worst. For a data set with conflicting signals, however, the unpartitioned model outperformed models that included more partitions, demonstrating some limitations to the effectiveness of concatenation for these data. For a transcriptome data set, fast coalescent methods ( STAR and MP ‐ EST ) and a Bayesian concordance approach yielded congruent topologies with trees from the concatenated analyses and previous studies. Our results highlight that well‐resolved gene trees are critical for the effectiveness of coalescent‐based methods. Future efforts to improve the accuracy of phylogenomic analyses should emphasize the development of new methods that can accommodate multiple biological processes and tolerate missing data while remaining computationally tractable.