Probabilistic analysis of the frequencies of amino acid pairs within characterized protein sequences.

Here, we describe a unique probabilistic evaluation of the 20, naturally occurring, amino acids and their distributions within the Swiss-Prot and Complete Human Genebank databases. We have developed a computational technique that imparts both directionality and length constraints into searches for unique combinations of amino acids within protein sequences. Using statistical approaches, we have carried out searches of all possible two- and three-residue motifs contained within these databases. This technique is based on the unusually high occurrence of a small number of these motifs when compared to the expected probability of finding a specific residue grouping within a given database. Subsequent filtering of this search to identify such unique combinations has provided several examples that can be used as markers to identify particular proteins within or across databases. We focus on three of these motifs, which were found to be of greatest interest to us. The CC, CM and a combination of the two, CCM motifs all occur either more or less frequently than would be predicted based on standard amino acid distributions within the entire human proteome.