Methodology

How we determine what we show

Every number in Skinceptor is decomposable. If a component cannot be traced to a public database record or a published study, it does not appear.

Evidence Index components (experimental)

Study design quality — randomised and controlled human work counts most
28%
Proportion of evidence generated in human skin or hair
22%
Number of independent curated studies
14%
Largest reported sample size
12%
Independent replication across separate studies
12%
How recent the body of evidence is
7%
Traceability — verified PMIDs or DOIs
5%

Confidence tiers

HIGH means the relationship is documented in a curated public database (UniProt, Reactome, Gene Ontology) or replicated in human studies. MEDIUM means mechanistic support exists but human replication is limited. LOW means the relationship is proposed, in-vitro only, or inconsistently reported. When we cannot place a relationship in any tier, we omit it rather than guess.

Network construction

Networks are generated from the curated dataset in a fixed order: ingredient → molecule → biological target → gene/protein → pathway → tissue process → studied outcome. A tier appears only when curated entities exist for it. Missing links are shown as gaps, not bridged with inference.

What the Evidence Index is not

It is not a safety rating, an efficacy guarantee, a product score, or a clinical recommendation. It summarises how much and what kind of published evidence our dataset holds for an ingredient. A low index frequently means 'little studied', not 'ineffective'.

Quantitative synthesis

We only describe pooled effects when at least four comparable human or randomised studies exist. Otherwise the interface states: insufficient comparable evidence for quantitative synthesis.

Medical boundary

Skinceptor is educational biological intelligence and research navigation. It does not diagnose, treat, or prescribe, and it is not a substitute for a dermatologist or physician.

See the data sources page for every database we draw on.