Code optimization of the subroutine to remove near identical matches in the sequence database homology search tool PSI-BLAST.

A central task in protein sequence characterization is the use of a sequence database homology search tool to find similar protein sequences in other individuals or species. PSI-BLAST is a widely used module of the BLAST package that calculates a position-specific score matrix from the best matching sequences and performs iterated searches using a method to avoid many similar sequences for the score. For some queries and parameter settings, PSI-BLAST may find many similar high-scoring matches, and therefore up to 80% of the total run time may be spent in this procedure. In this article, we present code optimizations that improve the cache utilization and the overall performance of this procedure. Measurements show that, for queries where the number of similar matches is high, the optimized PSI-BLAST program may be as much as 2.9 times faster than the original program.