Handling BAMs aligned with BWA-MEM -m flag

[tjbencomo]

Hi - I have tumor BAMs that have already been fully processed, including reference genome alignment with bwa mem using the -m flag. I saw in the documentation that reads should be aligned without the -m flag.

Would there be any issue if I used Picard's SamToFastq to convert the BAM to a FASTQ and then used the FASTQ as input to the pipeline? I can access the original FASTQs if needed but it would be easier to use the BAM.

Thanks

[jluebeck]

Hi Tomas, thanks for this question - indeed the reads will need to be realigned and you can provide the two fastq files as input to the pipeline. For unmapping bams, we provide this script for users who might need it, which uses samtools to do the unmapping. This method is supposedly faster than Picard's SamToFastq but I have not done any benchmarking myself - just an option to consider.

I recommend two stages for this workflow when running from fastqs:

Stage 1: Run AmpliconSuite-pipeline.py for alignment and CNV seed generation (multithreaded as needed), but do not set --run_AA --run_AC
Stage 2: Run AmpliconSuite-pipeline.py again but now for focal amplification identification (--run_AA --run_AC set) using the bam file and [sample]_AA_CNV_SEEDS.bed file produced from stage 1. This stage only uses a single thread.

Thanks,
Jens

[tjbencomo]

Hi Jens - thanks for the quick response! I will try that script instead of using Picard and then follow the 2 step process you suggested. Thanks!