Should we handle VDDJ rearrangements?

[psathyrella]

Might not be too hard, and in cases where there's really two Ds it'll of course make a huge difference to annotation accuracy.

[scharch]

If you decide to implement this, the program should be aware of the order of various D segments on the chromosome to cut down on false positives. That is, you could have D1-1 followed by D1-7, but not vice versa. See Figure 5 of Briney et al Immunology 2012 for more details.

…

On Tue, Feb 14, 2017 at 2:47 PM, Duncan Ralph ***@***.***> wrote: Might not be too hard, and in cases where there's really two Ds it'll of course make a huge difference to annotation accuracy. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#226>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGZe4MGx404sPTJwMOfzx2eg4qtnREtBks5rcgTAgaJpZM4MA6X-> .

[psathyrella]

ooh, excellent, thanks for the tip.

I also just realized this should be easy to implement -- we just need smith-waterman to look for them during parameter caching, and if it finds any it just adds the smooshed-together double D to the germline set that gets passed to the hmm.

[scharch]

Would that work? There would typically be N-nucleotides inserted between the two D genes...

[psathyrella]

huh, no, not if there's insertions between the Ds. I don't think there's much chance of adding two Ds to the hmm, that'd be crazy complicated. But the sw annotations are only a little less accurate than the hmm ones, anyway, so we could just look for double Ds there.