Feature request: Collect* only for specified chromosomes/contigs about picard HOT 4 CLOSED

You could also stream the data into the various metrics collection tools. For example, something like samtools view -b <in.bam> <region> | java -jar CollectWgsMetrics.jar I=/dev/stdin .... This would allow you to select only the regions in which you are interested.

If you are up for it, we would be happy to take a look at a pull request for this tool.

from picard.

Fair enough. Streaming is a good approach, though I'm not sure if the reference will also need modifying to get sensible output. That is something I will play with when I get the time and try to adjust/fix if needed.
Thanks for the quick response.

from picard.

Sorry, I don't thin your suggestion (streaming from samtools view) will actually work.
The problem is the header and reference.

To get the proper GENOME_TERRITORY (and serveral other things I suspect), the reference needs to be modified to just include the region/chrom of interest. That isn't difficult (samtools faidx), but does remove the possibility of a single command without intermediate files.

However, a critical problem occurs when picard is looking for regions in the reference base upon the bam file. Samtools doesn't remove entries outside of the selected region from the bam header. I've tried modifying the bam header (removing all chroms but the one of interest), but just can't get it to work. CollectWgsMetrics keeps looking for entries which aren't in the sequence dictionary... I have no idea why, since everything but a single chromosome has been removed from the bam header and alignments as well as the reference.

Anyway, even if I could get this to work (I've put several hours of trying into it already), the complexity required makes it far from a viable option IMO.
It seems like it would be much simpler to just add functionality to the Collect*Metrics commands to enable selecting a region. I imagine that the readers (BAMFileReader and such) already have most of the required functionality, but I'm not familiar enough with the code (nor have the time) to do it myself.

from picard.

I've added the functionality needed (and fixed a minor bug).
Please see my existing pull request, which was originally for a different (very minor) issue.
#87
Specifically, see this diff.
travc@2b039dc

At least a couple of people on BioStars and elsewhere have asked for this functionality, so it would probably be good if it could be merged soon. The changes are rather simple.

from picard.