mpiBLAST 1.3.0 has (finally) been released. It incorporates several major design changes that improve reliablilty, performance, and accuracy.
Key enhancements are:
- Utilizes latest NCBI Toolbox (Oct 20, 2004)
- E-value scores are exact and identical to scores reported by NCBI BLAST by using both the effective query and effective database lengths. (It is necessary/required to patch the NCBI Toolbox before compilation)
- Database pipelining speeds up database distribution by limiting access to shared filesystem. Number of concurrent accesses is a run-time option (--concurrent).
- Results pipelining reduces memory requirements for large query sets.
- Database distribution can occur through cp, scp, or rcp system calls, through MPI function calls (MPI_Send/MPI_Recv), or not at all if you choose to use the shared filesystem as local storage. It is a runtime option (--copy-via).
mpiBLAST version 1.3.0 was a long time in coming, but is the fastest, most exact, and most stable version of mpiBLAST yet.
Download it here:
http://mpiblast.lanl.gov
RPMs of 1.3.0 (source and binary) are available from Scalable Informatics download pages:
http://downloads.scalableinformatics.com/downloads/mpiblast/
Working mpich-1.2.6 RPMs (source and binaries) are also available:
http://downloads.scalableinformatics.com/downloads/mpich/
![[Next-Generation Sequencing Book]](https://www.bioinformatics.org/images/ack_banners/NGS-Book-1-728x90.png)
