Might be nice, especially if the FDR estimates used for fixing were based on a decoy context. (For example, if the FDR estimates used were based on rare CP2 mutations in the decoy contig, then we are making the assumption that all rare CP2 mutations in the decoy contig are false -- but we have not said anything about rare non-CP2 mutations in the decoy contig.)
There are various ways we could try to "rescue" these mutations, although most of these would probably start to cause problems with the validity of the FDR estimation logic in the first place:
Easy option = output all rare mutations that weren't included in the context that are above a user-specified "indisputable except not really" p or r threshold. (At this point, just set that p or r threshold to the "indisputable" one in the first place, right?)
Harder option = construct a "fake" target contig consisting of the not-included-in-the-decoy-context positions in the decoy contig, and treat it as if it were a target contig. This sounds kind of appealing on the surface: however, I feel like this is a risky approach (e.g. the composition of the original decoy contig can influence both the "real" decoy contig we use, and this "fake" target contig we create).
Might be nice, especially if the FDR estimates used for fixing were based on a decoy context. (For example, if the FDR estimates used were based on rare CP2 mutations in the decoy contig, then we are making the assumption that all rare CP2 mutations in the decoy contig are false -- but we have not said anything about rare non-CP2 mutations in the decoy contig.)
There are various ways we could try to "rescue" these mutations, although most of these would probably start to cause problems with the validity of the FDR estimation logic in the first place:
Easy option = output all rare mutations that weren't included in the context that are above a user-specified "indisputable except not really" p or r threshold. (At this point, just set that p or r threshold to the "indisputable" one in the first place, right?)
Harder option = construct a "fake" target contig consisting of the not-included-in-the-decoy-context positions in the decoy contig, and treat it as if it were a target contig. This sounds kind of appealing on the surface: however, I feel like this is a risky approach (e.g. the composition of the original decoy contig can influence both the "real" decoy contig we use, and this "fake" target contig we create).
Proobably not worth worrying about for now.