James Bonfield wrote:
> On Tue, Aug 02, 2016 at 01:44:01PM +0200, Martin MOKREJ? wrote:
>>> The warnings are there because it is correcting the aligner output.
>>> The proper fix is to fix the aligners to produce the correct MD in the
>>> first place, not to break calmd to be buggy in the same manner.
>>
>> Hmm, but bwa mem did not introduce the MD: tag, see the lines from
>>
>> samtools view mysample-PB.bam | grep 
>> HWI-xxxxx:xxx:xxxxxxxxx:2:1101:1110:65038
>>
>> in a previous email.
>
> Ok fair enough, consider my maligning of the aligner rescinded!

Pardon me? I am not a native speaker. ;-)

>
> However the point is still valid; somewhere some tool wrote incorrect
> MD tags.  It is not appropriate to make Samtools gloss over these
> errors, or worse, change the algorithm to compute things in the same
> broken manner.

"samtools calmd" is the first tool in my pipeline writing out the MD: tag,
but as it turned out this is more about NM: tag being mistaken, and that
was bwa mem's job. (I don't even know how to tell bwa mem to output also MD:
tag at the first sweep.

>
> It may be worth studying the @PG lines to figure out where the duff MD
> tags came from and send in a bug report.
>
>> So, I still think "samtools calmd" could be less verbose.
>
> It's a valid warning, but maybe there should be a silent mode.

Most people use bwa mem it seems, and I tried to use stampy on top of it to 
improve
mapping but I am not certain it is a good idea if I use GRCH38DH due to the alt 
locations
which would need to be supported in the stampy code. I assume it is not coping
with alt loci, and I do not have the alt file because bwa mem is said to infer
alternative mapping from the reference contignames containing _alt, _unknown, 
_random,
whatever.

> However with non-broken data you'll find it is very quiet so
> collectively I think the effort should be placed at fixing the
> tool that produced data. :-)

Indeed, join https://github.com/lh3/bwa/issues/82  .

------------------------------------------------------------------------------
_______________________________________________
Samtools-help mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/samtools-help

Reply via email to