Poisson, compound Poisson and process approximations for testing statistical significance in sequence comparisons (Q1194425)

Approximate statistical distributions under the null hypothesis of sequence independence are obtained for certain scores when DNA or protein sequence comparisons are performed. The comparison between two sequences is made on a diagonal and deletions and insertions are not considered. The scores studied are the total number of quality \(q/k\) \(k\)-word matches on a diagonal (number of \(k\)-words with \(q\) or more letter agreements), number of clumps on a diagonal (clumps of the above matches are conveniently defined in the paper), and maximum clump size for all diagonals. The approximate distributions are Poisson, compound Poisson, and integerized extreme value. The theorems also present bounds on the approximations. The paper concludes with some data analysis (including significance tests) and references to open problems.