Variant-class score distribution changed several hours after publishing score set urn:mavedb:00001283-a-1

[kennethngkq]

The displayed score distribution by variant class appears to have changed after the score set was uploaded and published.

Score set: https://www.mavedb.org/score-sets/urn:mavedb:00001283-a-1

Approximately 2–3 hours after publishing, the distributions for the different variant classes appeared correct and consistent with the uploaded data.

However, approximately 12 hours after publishing, the distributions changed and began to look incorrect. They have continued to appear incorrect since then.

For example, the current visualization appears to show too few synonymous variants. Other variant classes may also be affected.

Before:

After:

Could this be related to delayed variant annotation, reprocessing, caching, or a change in how variants are assigned to consequence classes? Would appreciate any help on this!

[bencap]

Hi @kennethngkq, thanks for the report and sorry for the delay getting back to you.

This stems from our variant annotation pipeline. When variants are first uploaded, the protein effect view is inferred based on the provided HGVS strings (for instance, we might infer NM_000023.4:c.36G>A/NP_000014.1:p.Val12= as a synonymous variant).

After we map variants to the human reference, we query Ensembl VEP to fetch protein consequences and take the most severe consequence. VEP annotates every overlapping transcript, not just the one we mapped to. In this case, it looks like that included transcripts with different exon structures which produce different reading frames, causing VEP to report a missense consequence as the most severe, even though this doesn't apply to our transcript of interest.

For now, I'm going to clear out the VEP consequences on this dataset so we can get back to the original histogram. We'll work on a fix internally so we can ensure VEP always just gives us the most severe consequence as it relates to the transcript of interest and re-annotate the score set when it's ready. I'll keep you updated once that happens.