Statistics Of Sequence Comparison

点击翻译按钮获取中文摘要

Work this year included the publication of a new measure of sequence similarity that unifies a traditional measure of alignment similarity with a new measure of compositional similarity. In brief, protein sequence database search programs may be evaluated both for their retrieval accuracy - the ability to separate meaningful from chance similarities - and for the accuracy of their statistical assessments of reported alignments. However, methods for improving statistical accuracy can degrade retrieval accuracy by discarding compositional evidence of sequence relatedness. This evidence may be preserved by combining essentially independent measures of alignment and compositional similarity into a unified measure of sequence similarity. We have studied two measures of compositional similarity, and found that one, when combined with alignment similarity, improves the statistical accuracy of blastp, as well as its retrieval accuracy measured using a SCOP-based test set. Further work this year focussed on developing scoring systems for recognizing correlated positions in multiple sequence alignments. We have built on other published work to confirm that in alignments of simulated sequences with correlated mutations, a form of normalized mutual information (nmi) appears to be the most effective measure. We have studied the mean and standard deviation of nmi for uncorrelated multiple alignment columns as a function of sequence number N and average per position relative entropy h.