How to speed up clair3?

[ChillarAnand]

For sample of ~3GB, clair3 is taking ~3 hours to complete on a machine with 40 cores, 512GB RAM.

How to speed it up? I am planning to run thousands of samples and the speed is becoming a bottleneck.

MODEL_NAME="r941_prom_sup_g5014"

time ${clair3_sh_filepath} --bam_fn ${bam_file} --ref_fn ${ref_genome} --threads 30 --platform ont --model_path ${model_dir}/${MODEL_NAME} --output ${folder_path} --sample_name=${i} --qual=8 --ctg_name=chr1,chr2,chr3,chr4,chr5,chr6,chr7,chr8,chr9,chr10,chr11,chr12,chr13,chr14,chr15,chr16,chr17,chr18,chr19,chr20,chr21,chr22,chrX,chrY,chrM --enable_phasing --enable_long_indel --min_mq=5 --chunk_size=5000000 > ${log_file} 2>&1

[aquaskyline]

have you got access to GPU? if yes, we have a version soon to be released that leverage GPU and reduces per sample time to below an hour.

[ChillarAnand]

have you got access to GPU? if yes, we have a version soon to be released that leverage GPU and reduces per sample time to below an hour.

That would be great. Does it work with apple silicon M series as well?

[aquaskyline]

ya, on a M3 Ultra finished whole genome 50x in like 20 minutes. Stay tuned, gonna release by early next month by @zhengzhenxian.

[ChillarAnand]

Excellent, @aquaskyline. I have bit of experience with python/bio-info/gpu. If I can help with development and/or testing, let me know.

[zhengzhenxian]

@ChillarAnand

GPU version released in v1.2.0.