Standardize CAF output for CAF Generator Function

[quinnwai]

User Story
As a CAF Generator user, I want standardized fields in the outputted CAF so that I can reliably process the genotype and phenotype information when using CAFs for analysis.

Resources

• Our generated CAFs: Terra notebook
• CAF schema from va-spec

Description
Scope might include...

• Formalizing storage of phenotype information and what phenotype information is being stored in the CAF ()
• Standardizing to a CAF schema
• Enabling provenance of data, eg allow user input to specify the data release in the upload process by passing a CAF for users to pre-populate

Questions

• For the CAF, for a specified phenotype should the derivedFrom field include the phenotype(s) it was filtered by?
• How should we store the phenotype information?

[ahwagner]

I took a look at the Terra notebook, and see the issue. Phenotypes are analogous to genetic ancestry with regard to how they should be specified with respect to CAF statements. Namely, this information should be transmitted in the cohort field, not ancillary results.

Grant-writing focus today, but will follow-up soon with proposals on how to manage this.

Re: CAF, we have a complete, implemented JSON Schema used in gnomAD. To avoid getting too deep into the internals of the GKS standards development process, suffice it to say that locking this into VA-Spec will take a little time due to the numerous parties and personalities involved. I will do what I can to move that along but request that we hold off on sharing this with downstream implementers until this issue is resolved.