Improve Clinvar display for various issues

[holtgrewe]

Is your feature request related to a problem? Please describe.
The now-improved ClinVar filter makes certain issues visible, see below.

Describe the solution you'd like
We need to address these issues in one way or another.

Describe alternatives you've considered
N/A

Additional context

• Follow-Up: Incorporate Clinvar haplotype information #653
• clinvar-tsv: Group output by VCV bihealth/clinvar-tsv#6

Issue 1: We currently display the interpretation with (impact, point-rating). This does not map to ClinVar.

In contrast, ClinVar merges this as follows because it trusts records with assertion criteria better than those without.

Even more pronounced:

Here, ClinVar creates three "clusters" and it is not obvious which one is "correct":

Issues 2: Gene families create many entries

ClinVar is smarter, see here:

[holtgrewe]

Here are the replies from NCBI/ClinVar helpdesk.

Point 1

The VCV page includes all submitted records for the variant, regardless of the interpreted condition. Aggregation/clustering of records is by the variation, not the variation-condition pair, which is represented by ClinVar's RCV accessions.

Point 2

The NM_006894.4(FMO3):c.472G>A variant is associated with three distinct Variation IDs in ClinVar, two haplotypes and one SNV. That may well be the reason, depending on your query, that you retrieved three results from the ClinVar database.

Point 3

For the third example, annotation of the UGT1A cluster on human chromosome 2 is somewhat unique. The Reference Sequence group at NCBI annotates the UGT1A isoforms as distinct genes rather than transcript variants that encode different protein products. ClinVar uses the Reference Sequence groups annotation, and so any of the UGT1A genes that include this variation are listed in the "Gene(s)" column. You can see how these genes are annotated here: https://www.ncbi.nlm.nih.gov/gene/54658 (in the "Genomic context" section).

[holtgrewe]

Resolution Proposal
We most likely should not touch point 3 for now. We could later take care of gene groups but that would have to start on the level of Jannovar. Rather, we should finalize the improvement in bihealth/clinvar-tsv#6 first. Then, we can adjust the model and query and UI for VarFish.

Affected Components

• clinvar-tsv
• VarFish

Affected Modules/Files

• variants/*

Required Architectural Changes
None

Required Database Changes
Adjust Clinvar model to the new data generated by the merge output step of clinvar-tsv.

Backport Possible?
Yes. Great care must be taken that the clinvar module's migrations are consistent.

Resolution Sketch

• Adjust form to select the clinvar summary mode between "clinvar" and "paranoid"
• Adjust model and queries and make tests run through
• Adjust REST API code
• Document both ClinVar modes
• Adjust the tests and check for consistency of results with real-world data
• Create backport