Most methods for inferring phylogenies from sequence data assume that patterns of substitution have been stationary over time. Changes in evolutionary constraint can result in nonstationary substitution patterns that are phylogenetically misleading unless modeled appropriately. Here we present a multiple-alignment-based method to identify regions that are likely to contain misleading phylogenetic signals due to changes in evolutionary constraints. The method uses a moving window approach to identify regions with a statistically significant deviation from stationarity in the physicochemical properties of amino acids among taxa. The protocol has been implemented in the software package DRUIDS (Detecting Regions of Unexpected Internal Deviation from Stationarity), available from the first author upon request.
Copyright 2005 Wiley-Liss, Inc.