USING SUBSTITUTION PROBABILITIES TO IMPROVE POSITION-SPECIFIC SCORINGMATRICES

Each column of amino acids in a multiple alignment of protein sequence s can be represented as a vector of 20 amino acid counts. Ebv alignmen t and searching applications, the count vector is an imperfect represe ntation of a position, because the observed sequences are an incomplet e sample of the full set of related sequences. One general solution to this problem is to model unobserved sequences by adding artificial 'p seudo-counts' to the observed counts. We introduce a simple method for computing pseudo-counts that combines the diversity observed in each alignment position with amino acid substitution probabilities. In exte nsive empirical tests, this position-based method outperformed other p seudo-count methods and was a substantial improvement over the traditi onal average score method used for constructing profiles.