Constraints on molecule type of Variant Path Statement subject

[mbrush]

As for other Statement types, we need to consider what constraints to put on the subject variation molecule type (genomic, transcript, protein) for Variant Pathogenicity statements. (e.g. for Molecular Consequence we constrained to genomic variation only).

Since sources like CIViC hold many annotations where a protein-level variation is the subject, or even a non-sequence variation such as "EGFR Exon18 overexpression", I assume we will need to allow for a broad range of variation levels as statement subjects.

If, however, we end up separating germline pathogenicity / predisposing statements from somatic oncogenicity statements (as debated here) - it may be that the former can be constrained to genomic variation subjects, but not the latter. But even in this case, I would opt to be more permissive.

[mbrush]

Note this relevant comment from @larrybabb (made from the perspective of a germline path resource):

The subject of these interpretations are most notably the genomic DNA change, not the RNA or AA change. While those downstream impact(s) are directly related to the cause, the interpretation is an assessment of the genomic dna change.

[ahwagner]

I would also opt for permissiveness in genomic, transcript, protein for pathogenicity (both predisposition + oncogenicity) statements. See, for example, this study of a predisposing protein consequence.

[dsonkin]

Genetics community spend years to make sure DNA changes are captured (there are special considerations for large genetic alterations). For germline entries not requiring DNA changes capture would be going against best practices and will push field back into dark ages. Somatic field is lacking behind in consistently capturing DNA changes, at a minimum we should only allow AA changes if submitter certifies that all genomic changes leading to particular AA change have exactly the same biological consequences. Pathogenicity statements refer to genetic alterations and should not include mRNA over-expression.

[larrybabb]

I would concur with @dsonkin that we should make strong recommendations that the genomic DNA representation of variants be used for pathogenicity statements whenever reasonably possible. And I also agree that if an AA variant is used that has no single genomic change then the protein change can be used. However, I do understand that folks will generate statements as they wish, but if they submit to knowledge bases and repos like ClinVar they will be reduced to the genomic change whenever possible. This transformation to the genomic change would then be the "reference variant" for these statements making the ability to aggregate the statements with the same variant possible.

[mbrush]

Decisions from 2-24-21 VA Call:

1. Allow for variation at genomic, transcript, or protein level in subject slot. But provide recommendations. a. Germline variant (Mendelian path) - genomic subject strongly recommended but not formally required. b. Somatic variant / cancer oncogenicity - genomic preferred, protein allowed (when annotation made explicitly at this level - often because annotation covers >1 genomic variation that leads to the same protein change)
2. Spec will provide recommendations (perhaps in a policy-type doc?) a. e.g. that data creator specifies genomic changes as well, when an as variation is subject, in the VOD . . . but where/how exactly??? b. e.g. allow AA changes as subjects only if submitter certifies that all genomic changes leading to particular AA change have exactly the same biological consequence (Dmitriy)? c. perhaps recommendations around checking if variant is near intron exon boundary - to be sure all have same tx change
3. Question - what needed w.r.t. a recommendation / policy doc for v0? Who will write recommendations / policies?