Automated sequence homology : using empirical correlations to create graph-based networks for the elucidation of protein relationships.

Abstract

Identification of sequence homology has presented a formidable obstacle despite significant increases in both technological capability and detailed knowledge of genomes and proteomes. While PSI-BLAST remains the popular tool for the job, it often returns inaccurate results with unacceptable levels of false positives. In order to increase the sensitivity and accuracy of homology finding, we have developed a software application called Automated Sequence Homology that bypasses these shortcomings and provides reliable and precise results. The system presented here is based upon the creation of a graph-based network highlighting the relational connections between proteins using empirical correlations. It takes a step back from PSI-BLAST to the acclaimed BLAST algorithm to create a sampling of the protein relational network.