Should VRS-Python Translators have an annotation mode?

[korikuzma]

I was looking through the Variation Normalizer issues and noticed we had Add moleculeType to SequenceReference (#594). The AnyVar developers recently had a discussion about molecule type, which would likely be needed for Write functions for forward and reverse transcription and translation (#146).

I think this raises the question on whether VRS-Python Translators should have some kind of annotation mode. Currently, the translate_from methods generate the minimal value object. Should we consider having an annotation mode to allow for information such as moleculeType to be included in the generated VRS object?

We should make sure that annotations do not make any additional policy decisions (these belong in the Variation Normalizer) and should stick with the current nomenclature's policies.

[jsstevenson]

We talked about this a bit on the call, but how do we envision that we'd get this from input? Would we have specific rules based on mode (e.g. gnomad/vcf is always genomic, hgvs gives you the type in the string, etc?)

[korikuzma]

Yes. We could also include residueAlphabet too as another annotation field we populate.

[SalemBajjali]

I think this is a great idea — in my work translating VRS Allele to FHIR AlleleProfiles, I’ve had to extract moleculeType and residueAlphabet myself when they weren’t already present. Having this supported directly in vrs-python would definitely streamline that process and improve consistency.

[SalemBajjali]

One thing I’m wondering: if these fields can be reliably derived from the input, shouldn’t we just include them by default in the translated output? Rather than needing a separate annotation mode, maybe they could be considered part of the normal translation process when that information is clearly inferable.