Re: [Samtools-help] samtools calmd behavior and MD tag

[Keiran R. <kr...@sa...>]

Hi Martin,

Adding 'calmdnmrecompindetonly=1' will increase performance further as it will only recompute the MD/NM values if the reference section has ambiguity/N within the span of the reads.

Regards,

Keiran Raine
Principal Bioinformatician
Cancer Genome Project
Wellcome Trust Sanger Institute

kr...@sa...
Tel:+44 (0)1223 834244 Ext: 7703
Office: H104

> On 2 Aug 2016, at 14:47, Martin MOKREJŠ <mmo...@gm...> wrote:
> 
> Keiran Raine wrote:
>> For BAM in/out yes:
>> 
>> inputthreads=<[1]>            : input helper threads (for inputformat=bam only, default: 1)
>> outputthreads=<[1]>           : output helper threads (for outputformat=bam only, default: 1)
> 
> bamsort fixmates=1 calmdnm=1 calmdnmreference="$reference" blockmb=40960 inputthreads=8 outputthreads=8 level=9 I="$sample".realignedtogether.BQSR.namesorted.bam O="$sample".realignedtogether.BQSR.namesorted.fixmate.calmd.bam
> 
> The above takes about 3 cores during input and since much later it starts writing output it takes 8 cores. Maybe just because of the extreme output compression only. But definitely, it outperformed "samtools clamd" step doing half of the work (just MD: tag calculations). Actually, processing the whole file took maybe 2 minutes in total? "samtools calmd" ran out of wallclock time limit at 12hrs on a cluster node (running on a single core).
> 
> Thank you for pointing me to bamsort, I added biobambam2 with libmaus2 to my Gentoo Linux recently (is in science overlay now), so it was simple to call it.
> 
> Martin




-- 
 The Wellcome Trust Sanger Institute is operated by Genome Research 
 Limited, a charity registered in England with number 1021457 and a 
 company registered in England with number 2742969, whose registered 
 office is 215 Euston Road, London, NW1 2BE.