Time frame for BLASTN against NT

[jackfrost]

Hi Everyone,

I'm trying to find out a time frame I should expect for running blastn. I have about 50k sequences that I want to run against NT. I realize there are many factors and so I'm trying to get a rough idea of whether I should be expecting hours, days, weeks, months(!), etc; but here are a few important ones:

The computer is an i7 processor ~1.7 GHz and I'm running four threads (using almost 400% in %CPU in top). Several hours into it I'm using just .5 GB of RAM (I have 8 on the computer) and I don't think I'm swapping. I'm running dc-megablast as the type.

Anyone with more experience running blastn able to give me a general time frame I should be expecting?

Thanks in advance.

[GenoMax]

Hopefully you are doing a test with just 5 sequences instead of full 50K

How long are the query sequences? What parameters are you using? What format are you saving the output as? Has anything been written to the output in the time the job has been running?

Realistically .. it could take days (to week(s)) for 50K sequences on a standalone machine with 4 cores. You really out to be doing this on a compute cluster with input split into multiple parallel jobs to speed things up.

[jackfrost]

Sequences vary but probably average around 1000-2000 bases. -outfmt 10; -evalue 0.0001 -lcase_masking -task dc-megablast

I'm guessing none of those would drastically change the time frame; in my output I have ~43,000 entries so far (after ~4 hours) although to be perfectly honest I'm not sure how many I expect.

I'm not sure that I have access to a server for this job sadly; but allowing it to run for a week isn't a big deal (the computer will be idle otherwise anyway). I'm not used to working with NT so its very hard for me to judge; but given the above does ~ 1-2 weeks sound too optimistic?

[GenoMax]

How many query sequences has the job gone through to produce those 43,000 entries (count unique query ID's in column 1 if you are using tabular output)? I assume the job hasn't gone through 43K queries (though it could possibly have, if it is not finding hits).

[jackfrost]

133 unique transcripts ... which I think answers my question as closer to 2 months by a rough calculation ! I'm going to give either megablast a try or just have to find a smarter way to do this.

Thanks for the insight though; I hadn't realized I was asking the computer to do such a big task.

[GenoMax]

If you don't need full NT you could use a subset (which may be better a choice depending on what you are trying to do, annotate?). That would speed things up.