How a marker gets into the panel.
Rule-based, documented and repeatable. Every decision can be traced back to a source and a written rule.
Six steps, in this order.
The pipeline is reproducible code. When a source is updated, the panel can be rebuilt and every change is visible.
Screen
All genome-wide associations in the GWAS Catalog, filtered by a 984-term keyword list across 47 categories and P ≤ 5×10⁻⁸.
Annotate
ClinVar, Open Targets, gnomAD allele frequencies, Europe PMC citations, errata and retractions are joined to every variant.
Grade
Each association gets an evidence grade from Strong to Limited, using fixed thresholds rather than judgement.
Ask: can you act on it?
Traits a person cannot change, such as height or eye colour, are removed however well studied they are.
Separate the medical
Diagnostic, pharmacogenomic and disease-named findings are classified and kept out of lifestyle advice.
Expert review
Each recommendation is reviewed and signed off by a nutrition professional before any user sees it.
How strong is the science?
Strong
Replicated, highly significant and consistent in direction.
≥ 3 publications · P ≤ 10⁻²⁰Moderate
Consistent across studies, or one very large study.
≥ 2 studies · or P ≤ 10⁻¹⁵, N ≥ 50,000Emerging
A single genome-wide significant study. Shown in research views only.
1 study · P ≤ 5×10⁻⁸Limited
Studies disagree on the direction of the effect. Not reported.
direction mixed · < 80 % agreementCan a change in daily life move it?
Our nutrition reviewer pointed out that an earlier selection favoured traits that had been studied the most. We rebuilt the score so that modifiability multiplies the evidence instead of competing with it.
| Class | What it means | Example | Weight |
|---|---|---|---|
| 1 | Changes a measurable biomarker | Vitamin D, homocysteine, LDL cholesterol | × 1.00 |
| 2 | Changes behaviour or intake | Caffeine metabolism, bitter taste | × 0.85 |
| 3 | Changes body composition | Response to physical activity | × 0.90 |
| 4 | Partially modifiable | Traits with small lifestyle effects | × 0.45 |
| 5 | Fixed or structural | Height, hair colour, age at menopause | × 0.10 |
Result: 525 variants in 445 genes, no fixed or structural traits, 194 graded Strong, 99 with documented gene × lifestyle interaction.
Where wellness ends.
Every row carries a safety class and the reason it was assigned. The class decides what a user may ever see.
- LifestyleReported normally, with its source.
- BorderlineThe action is kept; disease names and risk figures are removed.
- DiagnosticNever returned. These belong with a physician.
- PharmacogenomicInformation for prescribers only.
- RefusedThe market claims it; the evidence does not support it, and we say so.
We read 65 competitors before writing a line of product.
North America, Europe and Asia: websites, sample reports, validation papers and regulator lists.
Public, citable databases.
GWAS Catalog
Genome-wide associations and study ancestry (EMBL-EBI / NHGRI).
ClinVar
Clinical significance, used to separate diagnostic variants from lifestyle variants.
Open Targets
Variant effect prediction, allele frequencies, credible sets and locus-to-gene scores.
Europe PMC
Citations, errata and retractions for every paper we rely on.
CPIC & ClinPGx
Drug–gene guidance, used to route pharmacogenomic variants away from wellness advice.
PGS Catalog
Polygenic score metadata and performance, kept for research views.
Known limits: most genome-wide studies are of European ancestry; journal impact metrics are not used; key papers still need manual full-text appraisal. We list these openly because they affect what we can claim.