Obviously BLAST is going to have better on-average performance on a smaller database than on a larger database. But for a particular (large) database (let's say 10s of GB), how does splitting up the database into smaller files affect performance of subsequent BLAST searches? The default is to split the database into files no larger than 1GB, but is performance significantly affected if I decides to split the same database into chunks of 0.5GB or 5GB? It doesn't seem like this would make much difference...
...which leads to the second part of my question: what is the purpose of the max_file_sz option? Does file size indeed effect performance? Or is this perhaps a holdover from the days when 32-bit architectures placed constraints on file size?
•
written
by
Daniel Standage
1 answer
One purpose for max_file_sz (or use; I don't know if it was included for this particular point) is for using the GPU to BLAST. GPUs don't have as much global memory as CPUs do so this might sometimes be necessary to lower the max file size(see GPU_BLAST README for more details). I haven't seen it used as a performance improvement directly.
Don't know about the splitting of databases (you could test it and post the results here;)
•
written
by
Niek De Klein
Log in to answer this question.
-
Bowtie1 error outputting data
written by Sbrillo •
-
makeblastdb file file size question.
written by Julian.dekker •
-
Need help in BLASTP parameters (dbsize & searchsp)
written by shiv •
-
Splitting a fasta file based on the number of headers.
written by sicat.paolo20 •
-
Error with makeblastdb on TREMBL data FASTA file ?
written by DavidK •
-
Convert 10gb .bam file into smaller 1gb .bam files
written by kavitarege •
-
Strategy For Dealing With Reference Files (Genomes) That Are Larger Than The Software Limits
written by Pawel Szczesny
-
Blast E-Value To Database Size
written by Hranjeev
-
Problems With Blast And Nr Database
written by Daniel Standage
-
Makeblastdb Error
written by Daniel Standage
