ClinVar.

A shared record of what the field thinks a human variant means, and, just as important, how confident it is.

ClinVar, from the NIH, is the public archive where labs and expert panels deposit their interpretations of human genetic variants: pathogenic, benign, uncertain, and everything between. Its quiet genius is the second number it attaches to every classification: a review status, from zero to four stars, that says how much scrutiny the call has had. A single submitter with no stated criteria is very different from a variant reviewed by an expert panel or written into a practice guideline.

That confidence signal is what lets ClinVar be used honestly. The vast majority of the archive is one and two star. The expert-reviewed tier, three stars and up, is genuinely small, and it is the only tier we let make an affirmative claim.

The human clinical verdict, on the gene the dog shares, modeled the way the field models it.

When a dog gene resolves to a single human counterpart through Ensembl Compara orthology, Sniff reads ClinVar for that human gene and states, on the dog gene page, what human medicine has concluded. We adopt the Monarch Initiative's representation wholesale: the classification is an edge to a Mondo disease, pathogenicity is carried in the relationship itself, and only the three-star-plus, expert-reviewed tier drives a claim. Matching how Monarch models it keeps Sniff interoperable with the wider biomedical knowledge graph, not a private dialect.

Then we go one honest step further than a silent filter. Where the human evidence is not expert-reviewed, Sniff does not go quiet, which could be misread as reassurance. It abstains out loud: it says the classifications conflict, or that a pathogenic report exists but below expert-review confidence, and it says plainly that this is unsettled evidence, not that variants there are benign. It is GENE-level and about the human counterpart, never a verdict on the dog's own variant. "We do not know" is never allowed to read as "it is fine."