Data from: Evaluating summary methods for multi-locus species tree estimation in the presence of incomplete lineage sorting

Species tree estimation is complicated by processes, such as gene duplication and loss and incomplete lineage sorting (ILS), that cause discordance between gene trees and the species tree. Furthermore, while concatenation, a traditional approach to tree estimation, has excellent performance under many conditions, the expectation is that the best accuracy will be obtained through the use of species tree estimation methods that are specifically designed to address gene tree discordance. In this paper, we report on a study to evaluate MP-EST – one of the most popular species tree estimation methods designed to address ILS – as well as concatenation under maximum likelihood, the greedy consensus, and two supertree methods (MRP and MRL). Our study shows that several factors impact the absolute and relative accuracy of methods, including the number of gene trees, the accuracy of the estimated gene trees, and the amount of ILS. Concatenation can be more accurate than the best summary methods in some cases (mostly when the gene trees have poor phylogenetic signal or when the level of ILS is low), but summary methods are generally more accurate than concatenation when there are an adequate number of sufficiently accurate gene trees. Our study suggests that coalescent-based species tree methods may be key to estimating highly accurate species trees from multiple loci.物种树估计常受基因重复、基因丢失与不完全谱系分选（incomplete lineage sorting, ILS）等过程干扰，这些过程会导致基因树与物种树之间产生拓扑冲突。此外，尽管作为传统物种树估计方法的串联法在诸多条件下表现优异，但学界普遍认为，唯有采用专门针对基因树冲突问题设计的物种树估计方法，方能获得最高的估计精度。本文开展了一项评估研究，对MP-EST（目前针对不完全谱系分选设计的最流行的物种树估计方法之一）、最大似然框架下的串联法、贪婪合意树法以及两种超树方法（MRP与MRL）进行了测评。本研究表明，诸多因素会影响各类方法的绝对与相对估计精度，包括基因树的数量、估计所得基因树的精度以及不完全谱系分选的程度。在部分场景下，串联法的精度会优于最优的总结式物种树推断方法（多出现于基因树系统发育信号较弱或不完全谱系分选程度较低的情况）；但当拥有足够数量且精度足够高的基因树时，总结式方法的整体精度通常优于串联法。本研究提示，基于溯祖理论的物种树方法，或为从多位点序列数据中构建高精度物种树的关键手段。