Non-discordant overlapping reads don't get their phred scores tweaked about methyldackel HOT 8 CLOSED

the flags are 97 and 145 for the example bam, where does the discordant from BWA come in?

from methyldackel.

Exactly, those are flags for discordant alignments. 99 and 147 would be concordant (that's what bison produces, though one needs to convert the fastq files to be all uppercase for it to work (I'll add that to my list of things to fix for the next release)).

from methyldackel.

ah right.
I made up that data. does bison produce the same alignments with proper pair for cg100 and cg_R{1,2}.fq?

from methyldackel.

I think skipping unpaired should be the default. or we could just accept rf, ff like samtools

from methyldackel.

Yeah, bison produces the correct flags (well, once you make the reads upper case, otherwise it produces complete gibberish).

We can certainly default to skipping singletons (and discordant for that matter). The more I think about it, the more I think the discordant issue in the test dataset is really a bwa bug, since those aren't reasonably discordant alignments. I'll try to send a message to the bwa-help email list and see what Heng has to say (especially since this could be an issue for bwa mem/mpileup-based variant calling).

Let me know your opinion on defaulting to keeping or skipping discordants and I'll set the defaults accordingly.

from methyldackel.

Yes, I think requiring properly paired should be default

from methyldackel.

Those are now the defaults.

from methyldackel.

The most recent merges should take care of this issue. The only time pairs won't be tested for overlap now is if they're on incompatible strands (e.g., OT and OB).

from methyldackel.