I’m interested in secondary alignments but I don’t see any reported and parabricks doesn’t appear to accept the -a parameter from bwa. Can that be added please?
Thank you for letting us know that you would like this feature. I just let the engineering team know about this request and they might be able to accommodate it in a later release.
Can add more parameter ? eg: -O -d -r -E -k
Thank you @hehua0226, Can you elaborate more on your use case and why those flags are important for you?
According to testing, certain parameters may be more suitable for our sequencing data. In addition, with~120G WGS data, hard disk, CPU, and other similar conditions, I used a T4 16GB test to test fq2bam(~120min), but the running speed was not as good as bwa mem2(~100min). why? T4 is too poor?
docker run --rm --gpus all --volume $INPUT_DIR:/workdir --volume $OUTPUT_DIR:/outputdir --volume $REF_DIR:/refdir
–env TCMALLOC_MAX_TOTAL_THREAD_CACHE_BYTES=268435456
–workdir /outputdir
nvcr.io/nvidia/clara/clara-parabricks:4.3.2-1
pbrun fq2bam
–ref /refdir/${REFERENCE_FILE}
–in-fq /workdir/${INPUT_FASTQ_1} /workdir/${INPUT_FASTQ_2}
–out-bam /outputdir/${OUTPUT_BAM}
–out-duplicate-metrics /outputdir/${OUTPUT_BAM}.rmdup.metrics
–out-qc-metrics-dir /outputdir/qc1
–bwa-options “-Y -K 10000000”
–tmp-dir /outputdir
–markdups-picard-version-2182
–read-group-sm GT7
–read-group-lb LIB
–read-group-pl MGI2000
–read-group-id-prefix GT7
–bwa-cpu-thread-pool 48
–gpuwrite
–gpusort
–low-memory
Thank you for elaborating. For Parabricks, the storage will have a huge impact on performance. For the best results, you should put all the input data, output data, and tmp files onto an SSD local to the machine, usually on /raid. For more information check out the performance section of the documentation.