BLASTn produces empty output + output is 'Killed'

Question

[Shraddha]

Hello all,

I'm trying to run blastn using an own fasta file, but each time I try, I get an empty output file with the response 'Killed' on the terminal, and no other information. Here's the command I'm giving:

blastn -query sus_DEseqs.txt -subject mastertranscriptome.fasta -outfmt 6 -out sus_v_master.csv

I've also tried adding -task blastn to this command, but it produces the same output. The reference fasta file is about 2.1 Gb. Running makeblastdb produced the db files too. Also, it's not like blastn crashes immediately. It runs for roughly 30 minutes before outputting the empty file and the 'Killed' message.

My guesses on the issue at hand are:

1. improperly formatted fasta file (but this seems unlikely because this file doesn't give me issues with other tools)
2. we had memory issues on the HPC recently, so maybe carried over from that? Although if I run htop, I see it using 50-95% CPU and memory. It's a lot, but at least it's working.

Any ideas what's wrong/how I can fix it? Thanks!

[jkim]

I am not sure that my experience was helpful, but I think I had a sort of similar issue. My query files had 50,000 reads and when I was running it againt nt, they stopped after running for hours. So I increased the memory usage up to 100G, I finally was able to get the blast results.

[colindaven]

Sounds like you're running out of RAM. Try just doing a couple of query sequences instead of 2.1 GB. Do you get results back ? If it's going over 95% memory, linux is most likely going to kill the job to avoid going down.

If using a HPC, run on a bigger mem machine, or request more RAM (was the killed from linux, or from the job scheduler?)

[Shraddha]

Yes it looks like that's the issue, we've been asked not to use the HPC for a while so I guess it was an IT problem, not the command itself. If it continues when the HPC is up and running again, I'll give it a shot, thanks!

[GenoMax]

What is the length of the input query sequences? Have you tried to take just a few and run blast to make sure database is kosher and all else is working.

Answer 1

[JC]

your database needs to be formatted with makeblastdb first, somethig like:

makeblastdb -in mastertranscriptome.fasta -dbtype nucl -out mastertranscriptome
blastn -query sus_DEseqs.txt -db mastertranscriptome  -outfmt 6 -out sus_v_master.tsv

[Shraddha]

I did run makeblastdb, and the .nhr, .nin, and .nsq files are all there. Still gives me the killed error though.

[JC]

are you using -db param?

otherwise could be the memory limit as pointed by jkim

[Shraddha]

No I wasn't, but I'll try that out, thanks!