Algebraic techniques for phylogeny reconstruction from DNA sequence data.
by
Elizabeth S. Allman
University of Southern Maine
Coauthors: John A. Rhodes

A phylogenetic invariant for a model of biological sequence evolution along a phylogenetic tree is a polynomial that vanishes on the expected frequencies of base patterns at the terminal taxa. While the use of these invariants for phylogenetic inference has long been of interest, explicitly constructing such invariants has been problematic.

We construct invariants for the general Markov model of k-base sequence evolution on an n-taxon tree, for any k and n. The method depends primarily on the observation that certain matrices defined in terms of expected pattern frequencies must commute, and yields many invariants of degree k+1, regardless of the value of n. We define strong and parameter-strong sets of invariants, and prove several theorems indicating that the set of invariants produced here has these properties on certain sets of possible pattern frequencies. Thus our invariants may be sufficient for phylogenetic applications.

We show that on a set of biologically interesting arrays that phylogenetic invariants may be used to test that data is consistent with a general Markov model of sequence evolution.

Date received: August 8, 2003