Improved Hit Detection Algorithm for FSA Protein BLAST

Abstract

Abstract-Basic Local Alignment Search Tool (BLAST) is one of the most widely used bioinformatics tools to determine similarities between genomic sequences. Ever since its inception several algorithmic improvements have been made to improve speed and runtime memory requirements without affecting the sensitivity and selectivity of the tool. Fast search algorithm (FSA) BLAST has been the most successful among such improvements with 20-30% faster processing rate. In this work a modified data structure is used for the hit detection process. Codes have been developed separately for use of existing and modified data structures and tested on sample database. It has been found that the use of new data structure results in up to 81% reduction in run time memory requirement and about 20% reduction in hit detection time without affecting sensitivity and selectivity of the algorithm.

BLOSUM Trie for Faster Hit Detection in FSA Protein BLAST

“…In the modified structure, it is made dependent on length of query sequence and the number of neighborhood words to each query word. This reduces the run time space of algorithm by a considerable amount, by initializing only the necessary query pointers [19]. This paper is organized into seven sections.…”

“…The main functional differences between NCBI BLAST and FSA BLAST are, one is the structure used for finding hits between a query sequence and database sequence during the hit detection process and the other is using semi-gapped and restricted insertion alignments during alignment stage of the algorithm [6,7]. A modified DFA proposed in the earlier work, reduced the runtime space significantly during the hit detection stage of the FSA protein BLAST algorithm [19]. But reduction in hit detection time is not significant.…”