Species Delimitation Using Dominant and Codominant Multilocus Markers
Species Delimitation Using Dominant and Codominant Multilocus Markers
复制标题
DOI：
10.1093/sysbio/syq039
复制
发表时间：
2010-10-01
影响因子：
6.5
通讯作者：
Hennig, Christian
中科院分区：
文献类型：
--
作者：
Hausdorf, Bernhard;Hennig, Christian
我们提出了一种基于显性或共显性多位点数据使用高斯聚类和异常值噪声分量来界定物种的方法。案例研究表明，基于优势多位点数据使用高斯聚类界定的临时物种与基于其他数据界定的临时物种吻合良好。然而，基于少数共显性标记的高斯聚类在界定物种方面的表现只是中等。由少数个体代表的物种通常包含在噪声成分中，因为用有限的数据很难识别簇。作为替代方法，我们评估了最初提出的两种基于模型的聚类方法，即 STRUCTURE 和 STRUCTURAMA，以及“重组场”方法，用于推断种群结构并根据种群内 Hardy-Weinberg 平衡的假设将个体分配给种群。后者导致将每个数据集中的所有个体与共显性标记集中在一起，而 STRUCTURE 通常不提供有关簇数量的决定，而 STRUCTURAMA 通常会产生正确或几乎正确的簇数量。基于共显性标记的 STRUCTURAMA 分析的分类成功率非常好，但其与显性标记的性能不太一致。基于案例研究中使用显性和共显性多位点标记界定物种的不同方法的成功分类，我们建议对具有显性标记的数据集使用高斯聚类，对具有共显性标记的数据集使用 STRUCTURAMA。
We propose a method for delimiting species based on dominant or codominant multilocus data using Gaussian clustering with a noise component for outliers. Case studies show that provisional species delimited using Gaussian clustering based on dominant multilocus data correspond well with provisional species delimited based on other data. However, the performance of Gaussian clustering in delimiting species based on few codominant markers was only moderate. Species represented by few individuals are usually included in the noise component because clusters are difficult to recognize with limited data. As alternative methods, we evaluated two model-based clustering methods originally proposed to infer population structure and assign individuals to populations based on the assumption of Hardy-Weinberg equilibrium within populations, namely STRUCTURE and STRUCTURAMA, as well as the "fields for recombination" approach. The latter resulted in lumping all individuals of each data set with codominant markers together, and whereas STRUCTURE often provides no decision about the number of clusters, STRUCTURAMA usually yields correct or almost correct numbers of clusters. The classification success of STRUCTURAMA analyses based on codominant markers was very good, but its performance with dominant markers was less consistent. Based on the classification success of the different methods for delimiting species with dominant and codominant multilocus markers in the case studies, we recommend using Gaussian clustering for data sets with dominant markers and STRUCTURAMA for data sets with codominant markers.