TreeFix: Statistically Informed Gene Tree Error Correction Using Species Trees

Abstract

Accurate gene tree reconstruction is a fundamental problem in phylogenetics, with many important applications. However, sequence data alone often lack enough information to confidently support one gene tree topology over many competing alternatives. Here, we present a novel framework for combining sequence data and species tree information, and we describe an implementation of this framework in TreeFix, a new phylogenetic program for improving gene tree reconstructions. Given a gene tree (preferably computed using a maximum-likelihood phylogenetic program), TreeFix finds a “statistically equivalent” gene tree that minimizes a species tree-based cost function. We have applied TreeFix to 2 clades of 12 Drosophila and 16 fungal genomes, as well as to simulated phylogenies and show that it dramatically improves reconstructions compared with current state-of-the-art programs. Given its accuracy, speed, and simplicity, TreeFix should be applicable to a wide range of analyses and have many important implications for future investigations of gene evolution. The source code and a sample data set are available at http://compbio.mit.edu/treefix.

Genome of the pitcher plant Cephalotus reveals genetic changes associated with carnivory

“…To evaluate different tree reconstruction methods, simulated gene sequences were generated using the R package 'Phylosim' 92 . We used publicly available simulated data sets for 16 fungi species 93,94 . These data sets contain 1,000 simulated tree topologies of gene families, each of which was generated under observed gene duplication…”

“…5), they were considered to represent the same gene, whereafter we retained our own sequences to circumvent incorrect inference of gene duplication events in phylogeny reconciliation. A maximum-likelihood tree was reconstructed using third codon position sequences of the trimAl-processed alignments, and it was subsequently reconciled with a species tree prepared from a dated large-scale plastid phylogeny of flowering plants 54 using Treefix 94 with default parameters, except with the number of iterations increased to 1,000. Although the plastid-based topology 54 is partly different from nuclear-based topology 52,53 ( Supplementary Fig.…”

Rapid expansion of immune-related gene families in the house fly,Musca domestica

“…To both increase the computational efficiency of Treefix (see below), and improve the accuracy of our rate estimation, we used a custom Perl script (treesplit.pl on Github) to split trees where the deepest node was inferred to be a duplication rather than a speciation event. After the first round of tree splitting, we used the programs Treefix (v. 1.1.8; default options except -m PROTGAMMAWAG, -niter ¼ 1,000, and -maxtime) and treeannotate (part of the treefix package) to reconcile the species tree with each gene tree and compute the likely number of gains and losses on the tree (Wu et al 2012). Treefix attempts to produce the most parsimonious tree with respect to duplications and losses while remaining consistent with the maximum likelihood gene tree.…”

Duplication, Rearrangement and Reconciliation: A Follow-Up 13 Years Later

“…As reconciled gene trees also yield orthology relationships, it is possible to compare the results from both independent methods. This is what is often done to assess the quality of gene trees [73,74]. Orthologs obtained from synteny are assumed to represent the truth, and can be compared with orthologs obtained from reconciliations.…”