Samtools problem on a "rocks" cluster

[GenoMax]

I am having some problems getting samtools to run on a machine that is part of a rocks (v. 4.3) cluster (it happens to be the head-node which has 16GB RAM).

Even if I use the pre-compiled binary available from the samtools site (or compile samtools locally from source) I am getting a segmentation fault running samtools ("import"). The job is *not* being submitted to cluster queue. I want to make sure samtools actually runs and completes a job on this OS.

Has anyone else ran into this before?

BTW: How may of you are using "rocks" based cluster(s) for processing/analyzing next gen sequence data?

-- Hemant

[jnfass]

I often use maq or bwa on a rocks cluster, but haven't used samtools yet. But if the job is not being submitted to the queue, then how do you know that samtools is segfaulting? Some examples of commands, output, your shell script you're submitting to the queue, etc, would be helpful ...

[GenoMax]

I often use maq or bwa on a rocks cluster, but haven't used samtools yet. But if the job is not being submitted to the queue, then how do you know that samtools is segfaulting? Some examples of commands, output, your shell script you're submitting to the queue, etc, would be helpful ...

Let us set aside the fact that this is a cluster for a moment. I am trying to make sure the program can run successfully on the head node.

I am logged into the head node and then run a basic --> samtools "import" database *.sam *.bam command. samtools runs for sometime (e.g. I get this: [sam_header_read2] 28203826 sequences loaded) and then the program seg faults without producing any additional error messages.

Any ideas?

[jnfass]

sorry - seems like a straight samtools problem, unrelated to cluster usage (if you've used the command successfully on other data, you could verify that) ...

I can't help, but I'd suggest you post the exact command, and the exact output, so others might be able to help you. Good luck!

[nilshomer]

I am having some problems getting samtools to run on a machine that is part of a rocks (v. 4.3) cluster (it happens to be the head-node which has 16GB RAM).

Even if I use the pre-compiled binary available from the samtools site (or compile samtools locally from source) I am getting a segmentation fault running samtools ("import"). The job is *not* being submitted to cluster queue. I want to make sure samtools actually runs and completes a job on this OS.

Has anyone else ran into this before?

BTW: How may of you are using "rocks" based cluster(s) for processing/analyzing next gen sequence data?

-- Hemant

Can you give me the command you are using? What version of rocks and zlib are you using? We have rocks 4 and 5 running with samtools.

[GenoMax]

Can you give me the command you are using? What version of rocks and zlib are you using? We have rocks 4 and 5 running with samtools.

This is the command line:

samtools import ~/index/hg18.fa ~/single/single_end_mod.txt.part10.sam ~/single/single_end_mod.txt.part10.bam

Rocks v.4.3 and the stock zlib that came with 4.3 (which is 1.2.1 I think).

[nilshomer]

This is the command line:

samtools import ~/index/hg18.fa ~/single/single_end_mod.txt.part10.sam ~/single/single_end_mod.txt.part10.bam

Rocks v.4.3 and the stock zlib that came with 4.3 (which is 1.2.1 I think).

Try running

samtools faidx ~/index/hg18.fa

before the import. You need to index the reference.
Then run

samtools import ~/index/hg18.fa.fai ~/single/single_end_mod.txt.part10.sam ~/single/single_end_mod.txt.part10.bam

(notice the *fa.fai instead of *fa since the argument for import is a reference list, not the reference).

[jfshao1984]

samtools import <in.ref_list> <in.sam> <out.bam>
Since 0.1.4, this command is an alias of:
samtools view -bt <in.ref_list> -o <out.bam> <in.sam>

so you can use "samtools view -bt" instead. Before "view" ,you should run "samtools faidx" first. I am using rocks 4.3 cluster, and have not met some problem with samtools. Good luck to you

[GenoMax]

Thanks "nilshomer" and "jfshao1984" for pointing out the solution.