Identifying associations between amino acid changes and meta information in alignments.

Abstract

We present a method that identifies associations between amino acid changes in potentially significant sites in an alignment (taking into account several amino acid properties) with phenotypic data, through the phylogenetic mixed model. The latter accounts for the dependency of the observations (organisms). It is known from previous studies that the pathogenic aspect of many organisms may be associated with a single or just few changes in amino acids, which have a strong structural and/or functional impact on the protein. Discovering these sites is a big step toward understanding pathogenicity. Our method is able to discover such sites in proteins responsible for the pathogenic character of a group of bacteria. We use our method to predict potentially significant sites in the RpoS protein from a set of 209 bacteria. Several sites with significant differences in biological relevant regions were found. Our tool is publicly available on the CRAN network at http://cran.r-project.org/ naya@pasteur.edu.uy Supplementary data are available at Bioinformatics online.