Blastx speed issue...?

[lcdog]

Hey all! First time poster here.
I am blasting 122,000 transcripts against the NCBI NR database (using NCBI-blast 2.6.0). Have downloaded the database and formatted it and am using the following command

/blastx -query ***.fasta -db nr -outfmt "6 qseqid sseqid salltitles pident length slen mismatch gapopen qstart qend sstart send evalue bitscore" -num_alignments 1 -max_hsps 1 -num_threads 18 -out ***.txt

now the issue i'm having is that each set of 16 transcripts are taking about 10min to blast before the next 16 begin, meaning at the rate i'm going i'll be blasting for 60 days before I have this complete. Is there anything I can do to speed this process up? Or have I done something incorrectly? Or is this just the normal speed and process? (I'm running on a cluster with 20 cores and 120gb RAM)

Thanks

[GenoMax]

NR is a large databases and this may be the normal speed for your particular cluster.

People sometimes prefer using DIAMOND or PLAST as a speedier alternative to BLAST. You would need to make your own nr reference indexes for these packages.

[jonfoox]

Keep in mind that only certain phases of the BLAST algorithm can be hyperthreaded. So even though you are requesting -num_threads 18, the majority of the search is being done using just one. There's evidence that you get diminishing returns with the more threads you request – see http://voorloopnul.com/blog/how-to-c...g-num_threads/

Using another algorithm like GenoMax suggested is likely a good idea, but if you are stuck with BLASTX, the more you can parallelize the better. Do you have access to a cluster, where you can BLASTX many sets of 16 transcripts at once?