[Question]: What does htvc stand for in haplotypecaller - Parabricks 4.2.0-1

Hi

This is with reference to Parabricks 4.2.0-1 - nvcr.io/nvidia/clara/clara-parabricks:4.2.0-1

I want to know what htvc means in haplotypecaller and what does the associated binary at /usr/local/parabricks/binaries//bin/htvc do within the program. I tried searching for documentation around this but could not find any useful information

/usr/local/parabricks/run_pb.py haplotypecaller <.....snipped.....> --verbose --x3 --num-gpus 4

/usr/local/parabricks/scheduler.py /usr/local/parabricks/binaries//bin/htvc

The man page for haplotypecaller from 4.2.0 has these mentions of htvc

https://docs.nvidia.com/clara/parabricks/4.2.0/documentation/tooldocs/man_haplotypecaller.html#man-haplotypecaller

--htvc-bam-output HTVC_BAM_OUTPUT
File to which assembled haplotypes should be written. (default: None)

--num-htvc-threads NUM_HTVC_THREADS
Number of CPU threads to use per GPU. (default: 5)

--htvc-low-memory
Use low memory mode in htvc. (default: None)

Looks like this was introduced in 4.2.0 as the option is missing in 4.1.1
https://docs.nvidia.com/clara/parabricks/4.1.1/documentation/tooldocs/man_haplotypecaller.html#man-haplotypecaller

It would be beneficial if more information is made available on this.

Thanks in advance.

Hello @avenkatraman,

You are correct to notice that htvc-bam-output is present in 4.2 but not in 4.1. This option lets you output the assembled BAM file. In previous versions, you could only get out the variants as a VCF, but now it let’s you also output the assembled BAM if you need that as well.

Thanks @gburnett. What does htvc stand for - high throughput variant calling ?

Also is there any other documentation around this? Thanks in advance.

Anytime. htvc stands for haplotype variant caller. We do not have any other documentation around this, it’s just an additional flag that you can now use. Is there something that you’re looking for specifically?

Hi @gburnett

Is there something that you’re looking for specifically?

Yes - the issue mentioned in Parabricks 4.2.0-1 haplotypecaller error - cudaSafeCall() failed - out of memory

do I need to change/enable --htvc-low-memory and/or --num-htvc-threads 

Yes for a g4dn.12xlarge I would recommend that you use the htvc-low-memory flag. you should be alright to leave the num-htvc-threads flag alone.

1 Like

Thanks @gburnett

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.