Optimization for big sequences

[ArthurPERE]

Hello,

I have a file with 73 sequences that I would like to blastx against NR (~350 GB). But 10 of those sequences are big, the biggest is > 27 Mpb and the smallest of those 10 sequences is > 16 Mpb. The rest of the sequences is < 500 000 pb.

I am on a cluster, I put 150 cpu and 760 G of RAM. With the following options : -k 5 -b 6 -c 1 -f 6 -e 1e-25 --sensitive. But even with those options, the whole blastx would take ~31 days of calculs.

Is there another options I can use to speed up the calcul ? I can go higher on CPU and RAM if needed.

Thanks

[bbuchfink]

Sequences this long will only work efficiently in frameshift alignment mode, so you should set -F 15 --range-culling. Also there's a known performance issue with this in the current version, so please run v2.0.15 until this is fixed.

The other option would be to extract ORFs and run blastp.

[ArthurPERE]

Thank you for the fast answer. I will test those options.