Mismatched variant consequences between OT Genetics VEP data and the OT Platform evidence page?

[fedde]

The VEP (Ensembl) tab on https://genetics.opentargets.org/variant/1_169681903_G_A lists that, in the gene C1orf112 (ENSG00000000460), that variant’s most severe consequence is that it is intronic. The GWAS lead variants table on the same page shows that the variant came up in (among others) a study titled “Acne (severe) [fixed effect model]”. This association does appear in the platform, but listing it as being intergenic near that gene. gnomAD itself agrees with OT Genetics.

The discrepancy suggests to me that the ‘variant annotation’ (referenced on https://github.com/opentargets/evidence_datasource_parsers#processing-genetics-portal-evidence) used to generate https://ftp.ebi.ac.uk/pub/databases/opentargets/platform/22.04/input/evidence-files/genetics-portal-evidences-2022-04-12.json.gz was less complete/up-to-date than https://ftp.ebi.ac.uk/pub/databases/opentargets/genetics/22.02.01/v2g/; the input files I linked currently match what their respective applications say.

This is only one example of many similar cases. Am I correct to be surprised by this difference or mistaken about how to read the data model?

[dsuveges]

Hi @fedde , you are right, this information should be consistent across the Genetics Portal and the Platform. We are taking a look at the logic how the Platform evidence is generated.

[ochoa]

Issue opened

Incorrect variant consequence in Platform L2G evidence

[dsuveges]

Hi @fedde , We have identified the source of this discrepancy: for the Platform disease/target evidence we were picking the consequence on the canonical transcript of the gene. While other Genetics Portal (as well as GnomAD) uses the most severe consequence term across all transcripts.

In your example, the canonical transcript of this gene is the third from top.

Thanks for raising this issue. We are discussing how to go about this. We certainly make the two products consistent.