FlipCut Supertrees: Große und akkurate Phylogenien schneller bestimmen

超级树方法将具有重叠分类群集的输入树组合成一个更大、更全面的树。这里的问题是以合理的方式处理不兼容的数据。超树方法可以作为分而治之的系统发育树推断方法的一部分：(1)我们计算，比如说，分类群的几个子集的最大似然（ML）树；(2)我们将得到的树合并成一个超树；(3)最后，我们将该超树作为种子，对整个分类群进行ML分析。当前的超级树方法在分析大型数据集时速度太慢或太不准确。我们最近开发了一种新的超级树，叫做FLIPCUT超级树，它速度非常快，计算出质量很好的超级树。在这个项目中，我们将进一步推进这种方法，使超树的质量与目前的超树“黄金标准”MRP（矩阵表示）相当。与此同时，我们想要保持这种方法非凡的快速性。然后，我们将得到的方法集成到一个分而治之的方法中，用于系统发育重建，并对“纯”ML程序进行评估。

Supertree approaches combine input trees with overlapping taxa sets into one large and more comprehensive tree. Here, the problem is dealing with incompatible data in a reasonable way. Supertree methods can be used as part of a divide-and-conquer approach for phylogenetic tree inference: (1) we compute, say, maximum likelihood (ML) trees for several subsets of taxa; (2) we merge the resulting trees into one supertree; (3) finally, we use this supertree as a seed for an ML analysis on the complete set of taxa. Current supertree methods suffer from being too slow or too inaccurate when analyzing large datasets. We have recently developed a novel supertree called FLIPCUT supertrees, which is very swift and computes supertrees of good quality. In this project, we will further advance this methods so that supertree quality is on a par with the current supertree “gold standard” Matrix Representation with Parsimony (MRP). At the same time, we want to preserve the extraordinary swiftness of the method. We will then integrate the resulting method into a divide-and-conquer approach for phylogenetic reconstruction, and evaluate it against “pure” ML programs.