NEW TECHNOLOGIES   

Minerva Biotecnologica 2003 September;15(3):191-3

Copyright © 2004 EDIZIONI MINERVA MEDICA

language: English

DNCBLAST: a parallel BLAST for computer clusters

Rossi I., Marchignoli D., Fariselli P., Beltrami R., Medini D., Casadio R.

CIRB Biocomputing Unit and Laboratory of Biophysics, Department of Biology, University of Bologna, Bologna, Italy

The ­BLAST ­suite is ­designed to effi­cient­ly per­form­ing ­sequence sim­i­lar­ity search­es in bio­log­i­cal data­bas­es, and in ­this ­task it is con­sid­er­ably fast­er ­than oth­er heur­is­tic algo­rithms. However, espe­cial­ly for ­large ­genomes, ­BLAST per­for­manc­es can be optim­ised. We ­describe a sim­ple and effec­tive way to intro­duce par­allel­ism in the ­BLAST ­code by apply­ing a “­divide and con­quer” (DNC) ­approach ­based on split­ting and dis­trib­ut­ing the orig­i­nal data­base. In ­this way ­DNCBLAST can effi­cient­ly run on com­put­er clus­ters and over­comes the mem­o­ry lim­i­ta­tion of a sin­gle ­machine.