Re: [Samtools-devel] Propose to add the backward CIGAR operation 'B'

[Heng Li <lh...@sa...>]

Hi Srinka,

On Dec 5, 2011, at 3:01 PM, Srinka Ghosh wrote:
> There are issues w/ GATK indel calling. I'm curious as to how you are
> dealing with quality sores of the overlapping bases?

Currently samtools does this way: if two overlapping bases agree, the base quality (baseQ) equals the higher one of the two; if the two disagree, the consensus base is taken as the one having the higher baseQ and the consensus baseQ equals the subtraction. I think how to compute baseQ does not matter too much especially given deep coverage.


As to SNP/INDEL calling, could you provide more details about how CG are doing this? I could think of three ways:

1) Call SNPs and INDELs from the read-to-evidence alignment and then map the coordinates of called variants back to the reference using the evidence-to-reference alignment.

2) Call variants from the evidence-to-reference alignment without looking back at the supporting reads.

3) Project the read-to-evidence alignments to the reference coordinate and then call variants from the resulting read-to-reference alignments. (I guess this is less likely to be happening?)

My question is actually more related to the other topic we have not discussed in samtools-devel. But let's hijack this thread for a minute.

Thanks,

Heng

> Thanks,
> Srinka



-- 
 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.