Extreme value theory in some statistical analysis of genomic sequences (Q2463683)

The authors consider the problem of multiple alignments of protein-coding DNA sequences. A concept of multiple alignment profiles for a domain of protein sequences is introduced. It is assumed that the profiles are random and possibly with gaps. Statistical significance of the profile scores is estimated by deriving the distribution of their maximum. It is shown that the tail distributions of the scores behave like a mixed normal distribution. An application to the Immunoglobulin domain (Ig) and results of simulations are presented.