Inonsistent base correction using CallOverlappingConsensusBases

[jackvwbacon]

Hello!

I'm having some difficulties reverse-engineering the logic behind CallOverlappingConsensusBases. It seems that sometimes, bases are corrected, and other times, they are not, within the same sample.

In this read, the low quality "A" base (BQ=9) in the right alignment is successfully consensus-corrected by the high quality "G' base (BQ=40) in the overlapping left alignment:

However, just a few bases to the right, this low quality 'T' base (BQ=9) in the right alignment fails to be consensus-corrected by the high quality 'A' base (BQ=40) in the overlapping left alignment:

I don't understand the difference between the two situations, and why the first example was consensus-corected, while the second was not.

Is there something I am missing about the functionality of the tool?

Thanks in advance!

[tfenne]

@jackvwbacon I'm not sure why that'd be happening. Would you be willing to generate a BAM file with just the overlapping reads shown here in it and share that? If so, I can extract a test case from it and see what's going on. I think your intuition is right that both should be corrected.