Parallel Algorithms for Computational Biology

Bioinformatics algorithms are computationally demanding and most of them are compute intensive. Hence we require huge amount of computing power. High Performance Computing as been traditionally believed to be an expensive affair. In this paper we investigate techniques that have been used to parallelize conventional algorithms, coding patterns and architectural enhancements of platforms for better performance on Shared Memory and Distributed Memory architectures. 1 Sequence Comparison Sequence database searching is on of the most important tasks in Bioinformatics [RS00]. The Smith-Waterman Algorithm (SMA) is the best known algorithm and the only algorithm which guarantees the finding of the local alignment amongst given two sequences. But it is also known that it is one of the slowest of all algorithms too [Far07]. Since the actual algorithm is very slow, many heuristic based approaches like BLAST [AGM90] have been tried. But these approaches being based on heuristics, do not give exact results. Hence, there has been considerable effort amongst the researchers to find alternate ways of achieving faster yet accurate results. It is known that the size of the GenBank/EMBL/DDBJ nucleotide database doubles every 15 months [BKML00]. If this continues, there will be need for extraordinary amount of computing power for searching the databases. Several special purpose hardware