gistfile1.txt

Do proportion of SVs plots
See what BAF and logR are of my BP clusters that don't have SVs -- are they subclonal? Do they have smaller deviations?
  We should report these nulls back to PCAWG-6 -- now?
  Can we use our results to help PCAWG-6 -- ask them to look for non-consensus SVs that one or two methods called around our wll-supported BPs

Send list of BPs to PCAWG6, along with CN states upstream and downstream
  Send this list to group, which can check BAF and logR
  Include TRAs, but throw out small SVs

Geoff: Anything < 500 kb is often missed by CN methods for duplications or deletions
  Quaid: it may be that the CN methods are missing small ones -- I need to condition on size to figure out how bad this problem is

Plan for getting consensus:
  Include all SVs above 10 kb

How do we deal with "every start needs end" and "every end needs start"?
  Just progresssively relax threshold -- e.g., you might end up putting in five ends if we relax threshold all the way to only "one method"

What if we have small SV? <10 kb
  These should be included, but then we agree not to call CN on the <10 kb region between these SV start/end points

Consensus BPs:
  Include all SVs, including small <10 kb ones
  All consensus BPs
  And extra points to "end starts" and "start ends"

Only time we use the 10 kb threshold on SVs: asking whether there's a nearby SV relative to the consensus BP cluster
We have BB breakpoints even on the 1000 (of 2700) samples for which they've not released CN info because Stefan isn't happy it's correct

Stefan's plots:
  In one sample, if a cluster is on the diagonal and another is on the axis, suc hat their sum exceesd 1.0 in CCF space, you can't build a tree
    David: the diagonal one i often artifactual

TODO:
  Consensus BPs
  Send co-clustering matrices on three multisample cases to Jonas via Jamboree for constructing trees
    From then, we build the trees by hand