Baker R H, DeSalle R
Department of Entomology, American Museum of Natural History, 79th Street at Central Park West, New York, New York 10024, USA.
Syst Biol. 1997 Dec;46(4):654-73. doi: 10.1093/sysbio/46.4.654.
Relationships among representatives of the five major Hawaiian Drosophila species groups were examined using data from eight different gene regions. A simultaneous analysis of these data resulted in a single most-parsimonious tree that (1) places the adiastola picture-winged subgroup as sister taxon to the other picture-winged subgroups, (2) unites the modified-tarsus species group with flies from the Antopocerus species group, and (3) places the white-tip scutellum species group as the most basal taxon. Because of the different gene sources used in this study, numerous process partitions can be erected within this data set. We examined the incongruence among these various partitions and the ramifications of these data for the taxonomic consensus, prior agreement, and simultaneous analysis approaches to phylogenetic reconstruction. Separate analyses and taxonomic consensus appear to be inadequate methods for dealing with the partitions in this study. Although detection of incongruence is possible and helps elucidate particular areas of disagreement among data sets, separation of partitions on the basis of incongruence is problematic for many reasons. First, analyzing all genes separately and then either presenting them all as possible hypotheses or taking their consensus provides virtually no information concerning the relationships among these flies. Second, despite some evidence of incongruence, there are no clear delineations among the various gene partitions that separate only heterogeneous data. Third, to the extent that problematic genes can be identified, these genes have nearly the same information content, within a combined analysis framework, as the remaining nonproblematic genes. Our data suggest that significant incongruence among data partitions may be isolated to specific relationships and the "false" signal creating this incongruence is most likely to be overcome by a simultaneous analysis. We present a new method, partitioned Bremer support, for examining the contribution of a particular data partition to the topological support of the simultaneous analysis tree.
利用来自八个不同基因区域的数据，研究了夏威夷果蝇五个主要物种组代表之间的关系。对这些数据进行同步分析得出了一棵单一的最简约树，该树：(1) 将阿迪亚斯托拉绘翅亚组作为其他绘翅亚组的姐妹分类群；(2) 将跗节变异物种组与触角果蝇物种组的果蝇归为一类；(3) 将白盾片物种组作为最基部的分类群。由于本研究使用了不同的基因来源，因此在该数据集中可以构建许多过程分区。我们研究了这些不同分区之间的不一致性，以及这些数据对系统发育重建的分类学共识、先验一致性和同步分析方法的影响。单独分析和分类学共识似乎不足以处理本研究中的分区。虽然可以检测到不一致性，并且有助于阐明数据集之间存在分歧的特定领域，但基于不一致性对分区进行分离存在诸多问题。首先，分别分析所有基因，然后将它们全部作为可能的假设呈现或获取它们的共识，实际上几乎没有提供关于这些果蝇之间关系的任何信息。其次，尽管有一些不一致性的证据，但在仅分离异质数据的各种基因分区之间没有明确的划分。第三，在联合分析框架内，就能够识别出有问题的基因而言，这些基因与其余无问题的基因具有几乎相同的信息含量。我们的数据表明，数据分区之间的显著不一致性可能仅限于特定关系，并且通过同步分析最有可能克服造成这种不一致性的“错误”信号。我们提出了一种新方法——分区布雷默支持度，用于检验特定数据分区对同步分析树拓扑支持的贡献。
