Skip to main content
snıff

Research instrument · Spectra

Where do cited evidence layers agree?

Pick a scientific job below. Sniff will show genes where dog and human evidence both ring, how many cited channels are lit (some share provenance), where tumor biology diverges honestly, and where our coverage is uneven (not a verdict about importance).

… genes with both a canine signal and human evidence. Every lit layer is a cited count you can open. A flat layer means we have not looked there yet.

What do you want to find?

Start here · a clear finding

ABCA4

Loading the live spectrum for ABCA4…

Require a layer lit (stack filters)

… genes

Tuning fork · pre-registered

loading…

Does channel agreement beat fame?

Loading the leave-source-out MVP…

Temporal probe pending. The clean proof is foresight: did consonance rank genes that only later became ClinVar ≥2★ P/LP, before they were curated? That run needs the box and has not been done. Until it lands, do not read the MVP as proof of the instrument.

Pre-registered before results. Fame baseline mandatory. Leave-source-out anti-circularity. No free per-edge weights (INV-84). Explore mode: we tune the measurement, not the graph or the waterline.

What this is. A recount of cited evidence, never a score of our own. Ranking uses two numbers from a pre-registered stream table (INV-96): evidence streams that agree (cross-stream, after shared-curator / shared-literature deflation) and deepest stream (within-stream channel richness). Raw field count is a receipt only. Each lit channel traces to its source. We render and cite others' classifications; we do not model variants ourselves.

Study depth is secondary. PubMed gene–PMID counts (NCBI gene2pubmed) label well-studied vs less-studied genes so high agreement is not mistaken for novelty. Sort by “surprisingly dense for PubMed fame” to put high-agreement, lower-citation genes first. Not a quality score.

Gene-level, candidate-not-verdict. A lit human layer means the human gene carries that evidence, never that a dog allele equals a human one. A cross-species match is a candidate model to test.

Dissonance (INV-81). Disagreement only when both sides are cited on a comparable lesion. See also comparative oncology.

Darkness is a result. A flat layer is unexplored or absent in our join, not “nothing in biology.”

Sources: OMIA · Dog10K · ClinVar · gnomAD · Mondo/Monarch · orthology methods · AVCG (Boeykens 2024) · oncology cohorts. Each lit segment cites its own.