SuperFine: Fast and Accurate Supertree Estimation

Abstract

Many research groups are estimating trees containing anywhere from a few thousands to hundreds of thousands of species, toward the eventual goal of the estimation of a Tree of Life, containing perhaps as many as several million leaves. These phylogenetic estimations present enormous computational challenges, and current computational methods are likely to fail to run even on data sets in the low end of this range. One approach to estimate a large species tree is to use phylogenetic estimation methods (such as maximum likelihood) on a supermatrix produced by concatenating multiple sequence alignments for a collection of markers; however, the most accurate of these phylogenetic estimation methods are extremely computationally intensive for data sets with more than a few thousand sequences. Supertree methods, which assemble phylogenetic trees from a collection of trees on subsets of the taxa, are important tools for phylogeny estimation where phylogenetic analyses based upon maximum likelihood (ML) are infeasible. In this paper, we introduce SuperFine, a meta-method that utilizes a novel two-step procedure in order to improve the accuracy and scalability of supertree methods. Our study, using both simulated and empirical data, shows that SuperFine-boosted supertree methods produce more accurate trees than standard supertree methods, and run quickly on very large data sets with thousands of sequences. Furthermore, SuperFine-boosted matrix representation with parsimony (MRP, the most well-known supertree method) approaches the accuracy of ML methods on supermatrix data sets under realistic conditions.

Parallelizing SuperFine

“…Clade-based trees by themselves cannot be used to assemble a tree on the entire data-set, but the inclusion of the scaffold tree provides the overlap that makes this possible. As observed in [31], supertrees are typically more accurate when scaffold trees are more densely sampled, but typical practice in systematics produces scaffolds containing only a sparse subset of the taxa. We present results for 1000-taxon model trees with 20% and 100% scaffolds, in order to consider both extremes (the case which is more typical of biological practice, and the case where the best accuracy is obtained, respectively).…”

“…Recently, Swenson et al developed a new supertree method called "SuperFine" [31], which is both faster than MRP and also more accurate (as determined on simulated data-sets, described below). SuperFine has three phases.…”

“…Swenson et al studied SuperFine in comparison to other supertree methods (including MRP) on a collection of previously studied biological supertree data-sets and simulated data-sets [31]. SuperFine was generally faster than MRP: it never took more than 2.5 minutes on any simulated supertree data-set, and under 35 minutes on the most difficult biological data-set, while MRP heuristics took up to 2 hours on the simulated data-sets and 14 hours on the most difficult biological supertree data-set.…”

“…In this study, we parallelized SuperFine and compared performance of these parallel versions to that of sequential SuperFine, using biological and simulated data-sets obtained from [31]. These biological data-sets consist of:…”

MRL and SuperFine MRL, New Supertree Methods

“…Next every leaf in each of d subtrees is relabeled by the label (from 1...d) assigned to the root of its subtree. At this point, SuperFine creates a new set of source trees, by modifying each of the input source trees so that each contains at most d leaves, as follows: If x is an internal node in a source tree that is adjacent to two leaves, each of which has the same label l, then we remove its neighboring leaves and relabel x by l. In [24], it was proven that this modification produces source trees that have at most one leaf with each label. SuperFine then applies its base supertree method (such as MRP) to compute a supertree on the modified source trees, each of which has at most d leaves.…”