Inconsistent mapping result

Healthcare & Life Sciences Parabricks
1

Hello, I am trying to compare the mapping result between parabricks v4.3.2-1 and vanilla minimap2 v2.26-r1175. But I found significant different numbers of mapped reads.

Parabricks:
docker run --rm --gpus all --volume ./:/workdir --volume ./:/outputdir --workdir /workdir nvcr.io/nvidia/clara/clara-parabricks:4.3.2-1 pbrun minimap2 --preset map-ont --low-memory --ref /workdir/chr20.fa --in-fq /workdir/ulk.chr20.fq.gz --out-bam /workdir/gpu.ulk.chr20.bam

Minimap2:
minimap2 -x map-ont chr20.fa ulk.chr20.fq.gz -t32 -a | samtools view -bhS -@32 > cpu.ulk.chr20.bam

image
image549×775 15.8 KB

May I know why this happens and how to fix this? Thanks!

2

Hi @sh0rt2l0ng,

Parabricks minimap2 is compatible with minimap2 version 2.24, not 2.26 so I would compare it against that version and see if you’re facing the same issue.

You can see which version the parabricks tools are compatible with by running:

docker run --gpus all --rm nvcr.io/nvidia/clara/clara-parabricks:4.3.2-1 pbrun minimap2 --version
Please visit NVIDIA Clara - NVIDIA Docs for detailed documentation

Version Information

Tool: minimap2
pbrun: 4.3.2-1

Compatible With:
minimap2: 2.24

Thank you

3

with minimap v2.24-r1122
image
Is this related to that parabricks can not deal with ultra long reads correctly?

4

Hi @sh0rt2l0ng,

That might be causing an issue. Can you tell me more about the data you’re using?

Thank you

5

Hi @sh0rt2l0ng ,

There used to be an issue in our minimap2 when handling ultra-long reads. And it got fixed in our 4.4 release. Please try that and see if the mismatches are still there.

Thanks!

6

Thanks. I tried this version with similar result compared to minimap2 now. But the problem is that it seems this v4.4 is not accelerated by GPU (tried 4090m and A800). Any comments?