Built for the VUS problem

A sequencing study surfaces candidate variants, and most of them come back as variants of uncertain significance. The evidence that bears on each one is scattered across records, and assembling it by hand, variant by variant, is the slow part.

Builds an evidence pack for each candidate variant: a calibrated probability where the evidence supports one, a citation behind every claim, and an explicit decline where it does not — so a researcher starts from the evidence, not from a blank search page.

It does not classify variants clinically, and it does not replace ACMG/AMP review by qualified professionals. Its reports are evidence summaries for research.

A cohort accumulates hundreds of uncertain variants over time. Re-checking them one by one means most are never re-checked at all, and the ones the evidence cannot support quietly disappear into a spreadsheet.

Scores a batch list in one run, surfaces the borderline band (0.40–0.60) for human review first, and keeps declines visible in the dashboard and the exports instead of burying them — so the whole cohort is triaged, including the honest “not enough evidence” answers.

It does not silently drop what it cannot score, and it does not auto-classify. Every variant in the batch is accounted for: summarised, borderline, or declined.

Before a team commits to a gene as a target, the human-genetics evidence for it sits spread across variant records, and the pipeline runs behind an analysis are hard to reproduce exactly when a result is challenged.

Produces gene-level evidence summaries from the variant record, and reproducible pipeline runs through the Atlas — the same code, run again, returning byte-identical results that a reviewer can verify.

It does not run clinical-grade pipelines, and it does not validate a target’s biology. It organises the genetic evidence and makes the computation reproducible; the science judgment stays with the team.