How the Symptom Finder Ranks Rare Diseases

This tool searches the Human Phenotype Ontology (HPO), the standard vocabulary clinical geneticists use to describe symptoms, against Orphanet's expert-curated database of rare disease phenotypes. Results are ranked with an information-theoretic scoring model informed by the methods behind clinical computational phenotyping tools such as Phenomizer and LIRICAL, which are used in genome diagnostics.

In practice that means rare, specific symptoms count for far more than common ones: a finding like corneal verticillata, seen in only a handful of diseases, moves a match far more than a headache, which appears in thousands. The model also weighs how frequently each symptom occurs in each disease using Orphanet's published frequencies, treats a symptom the literature explicitly excludes from a disease as evidence against that match, and caps every score by how informative your selected symptoms actually are, so a single vague symptom can never produce a misleadingly confident result.

Two further refinements mirror how clinical tools reason. Related symptoms earn partial credit through the HPO hierarchy, so selecting a specific seizure type still counts toward a disease annotated with the broader term. And each score incorporates a pretest probability weight from Orphanet's published prevalence classes, the same principle LIRICAL applies: when two diseases match your symptoms equally well, one affecting 1 in 10,000 people ranks above one documented in a handful of families worldwide. Results are listed strictly in score order.

This is an educational research tool, not a diagnosis. Rare disease diagnosis requires clinical evaluation and often genetic testing; bring promising matches to your doctor or a genetics professional.