To do list for the v1.1b panel #56
Description
-
1. Can we run the same HPRC 1.1 w/o TRGT as a baseline? Less priority would be HPRC 1.0 w/o TRGT (if we don't already have that run somewhere, but I'd do a fresh run anyway).
-
2. Can one run HiPhase with short+TRGT only? Is there some distinction between how HiPhase handles the SV and TRGT inputs?
-
3. What is the overlap between the TRGT catalog and 1.1? What AFs does the TRGT catalog cover?
-
4. If the overlap is substantial, can we do some quick Vcfdist evals of the appropriate HiPhase SV outputs, stratified by TR/non-TR regions to get an idea of whether 1.1 or TRGT is more reliable in each region type? For example, if HPRC 1.1 w/o TRGT metrics in TR regions look not so great, maybe we should just prefer the TRGT genotypes in those regions. One might imagine that if we end up preferring TRGT genotypes in TR regions and end up with disjoint sets of 1.1 non-TR and TRGT TR, HiPhase and bcftools concat will have an easier time.
-
5. Have we made any progress looking at TRGT merge (on pure TRGT outputs, not HiPhase outputs)? Is this just a trivial operation on squared genotypes?
-
6. I think we are using HiPhase 1.3.0. Have we tried more recent versions? In particular, it seems like 1.4.0 might introduce some noteworthy changes.
-
7. See where the raw per-sample TRGT calls lie on Fabio's latest ROC plots. This will of course just be a point rather than a curve for each sample.