|
Abstract:
|
In phylogenetics, evolutionary relationships among organisms are depicted by a binary tree structure, in which internal nodes represent ancestral species, and external nodes represent present-day organisms. Usual procedure to estimate a species tree involves the estimation of gene trees from genomic data, and then the estimation of the species tree from the inferred gene trees. The main challenge of this two-step approach is that there are many stochastic processes that create discordance in the gene tree sample, and the models that govern the likelihood of a species tree given a sample of gene trees tend to focus on a single process at a time. Thus, gene trees originated from a different process are outliers difficult to detect. Here we propose RANSANEC (RAndom SAmple NEtwork Consensus), a new method to detect gene tree outliers in the form of random sample consensus. By identifying the outlier gene trees that violate a specific biological process, we can reconstruct the species tree with the inliers with high precision. Furthermore, our method allows to consolidate the inliers species tree and the outlier gene trees in a combinatorial method to reconstruct a phylogenetic network.
|
