Home - zwickl/terraphy GitHub Wiki

Table of Contents Terraphy What are terraces? Why should I care? Input files and formats Getting started

Various analyses and utilities related to phylogenetic terraces

A phylogenetic terrace is a sometimes very large set of phylogenetic trees that have identical scores with respect to a some data set according to a criterion like parsimony or likelihood. These can occur for many reasons, but the one that we consider here has entirely to do with patterns of missing data. In a multilocus sequence alignment, for example, we can describe the pattern of data presence/absence by a taxon coverage matrix, and the structure of this matrix determines many properties of terraces. Terrace properties have been most fully explored for maximum parsimony and maximum likelihood optimality scores on sequence data sets. However, terraces can also occur in species tree inference based on gene trees (rather than the underlying sequence data). See "Background" for more information.

Large terraces can impede searches for optimal trees, bias estimates of confidence in resulting phylogenies, and alter inferences about ancestral states or other comparative analyses (because of incorrect assessment of the true tree diversity). See "Background".

Only two pieces of information must be provided to the software. The first is the pattern of taxon coverage, i.e, which loci are sampled for each taxon. The second is the input "parent" tree. The taxon coverage matrix is generally inferred from a sequence alignment, but the sequences themselves are irrelevant to the concept of terraces.

We recommend jumping to the "Tutorial" page for information about installation and getting started quickly.