User Tools

Site Tools


mpiblast

This is an old revision of the document!


mpiBlast

mpiBlast is a freely available, opensource, parallel implementation of NCBI Blast. mpiBlast takes advantage of shared parallel computign resources, i.e. a cluster this gives it access to more avaliable resources unlike NCBI blast which only can take advantage of shared-memory multi-processors(SMP's).

More information is available here.

Environment Setup

Make a directory to contain the FASTA database that will be fragmented. Download the database and decompress it.

razor-l1:jokinsey:~$ mkdir db
razor-l1:jokinsey:~$ cd db
razor-l1:jokinsey:~/db$ wget ftp://ftp.ncbi.nlm.nih.gov/blast/db/FASTA/mito.nt.gz
razor-l1:jokinsey:~/db$ gunzip mito.nt.gz

Create a $HOME/.nbirc file with these values. The shared path tells mpiBlast where to access the FASTA database.

[mpiBLAST]
Shared=/home/YourUserName/db
Local=/local_scratch/YourUserName

Format the database for parallel use, by fragmenting the database for each processor. We will be using a node with 12 processors so we will include the option - -nfrags=12.

razor-l1:jokinsey:~$ mpiformatdb -i ~/db/mito.nt --nfrags=12
Reading input file
Done, read 2605891 lines
Database type unspecified, assuming nucleotide
Breaking mito.nt into 12 fragments
Executing: formatdb -i /home/jokinsey/db/mito.nt -p F -N 12 -o T 
Created 12 fragments.
<<< Please make sure the formatted database fragments are placed in /home/jokinsey/db/ before executing mpiblast. >>> 
mpiblast.1505249881.txt.gz · Last modified: 2017/09/12 20:58 by jokinsey