PSSMs: Not just roadkill on the information superhighway.

摘要

I have developed two computational scoring tools that can be applied to biological problems. The first is the PHAT substitution matrix designed for transmembrane regions of proteins. The PHAT scoring matrix may improve database searching, tree phylogenies, and alignment for transembrane proteins. The second is a procedure that predicts whether an amino acid substitution affects protein function. This method, called SIFT which sorts intolerant from tolerant amino acid substitutions, bases its predictions on sequence homology. Prediction performance on substitutions in three proteins for which the affected phenotype is shown. When SIFT is applied to human variant databases, a majority of substitutions annotated to cause disease are predicted to be damaging whereas a majority of putative single nucleotide polymorphisms (SNPs) are predicted to be neutral. Performance results indicate that SIFT can be used as a guide to predict which mutations should be studied in more detail.