MpiBLAST: Nucleotide/Protein Searching
From Debian Clusters
About mpiBLAST
mpiBLAST is a parallel implementation of the Blast algorithm. BLAST, or the Basic Local Alignment Search Tool, is used to search large databases. "Local alignment" refers to the way BLAST compares sequences. When dealing with a potentially large sequences, rather than looking for a match in entirety, it looks for smaller patches that match. When it finds a match, it evaluates the probability that this match would occur by chance. If it's highly likely the match occurred by chance, it probably isn't very statistically significant. These matches can be searched for in either a string of nucleotides (which make up DNA) or amino acids (which make up proteins). When running BLAST, you select which type your sequence is and choose the databases of known sequences you're interested in comparing it against.
mpiBlast divides the work of searching these databases up amongst multiple processes (and processors, if available) by using MPI. First, the database to be searched is split into as many pieces are there will be processes. Then, the processes communicate amongst themselves using MPI. You'll need to have MPICH or another implementation of MPI installed in order for mpiblast to work.
There are two pages about mpiblast:
- Installing mpiBLAST - all the setup and configuration
- Using mpiBLAST - running mpiblast
References
- Blast: Basic Local Alignment Search Tool
- The Design, Implementation, and Evaluation of mpiBLAST. A. Darling, L. Carey, and W. Feng. 4th International Conference on Linux Clusters: The HPC Revolution 2003 in conjunction with ClusterWorld Conference & Expo, June 2003. Available online at http://www.mpiblast.org/downloads/pubs/cwce03.pdf

