bioArc
Begin
MMethodology

The concordance score in plain language.

A finding is just a hypothesis until enough independent signals point at it. The concordance score is bioArc's way of putting a number on "enough."

01 — Inputs

Five independent signals.

For any candidate finding (e.g., methylation-cycle dysfunction), we extract a per-domain probability that the signal supports the hypothesis.

Genetics
p_gene = posterior given variant prior + GWAS β + PRS percentile
Blood
p_blood = z-score of biomarker against reference + functional range
Methylation
p_meth = age-residual on Horvath / Hannum / PhenoAge ensemble
Family
p_family = inheritance-weighted disease prior (Mendelian + complex)
Lifestyle
p_lifestyle = mediator from logged behaviour + supplement compliance
02 — Combine

Weighted likelihood, not vote-counting.

We combine the five domain probabilities with a Bayesian weighted geometric mean, where each weight reflects evidence quality (sample size, validation status, recency).

Concordance score · simplified
C(F) = (∏ pᵢ^wᵢ) ^ (1 / Σwᵢ)

where  pᵢ = posterior probability from domain i
       wᵢ = evidence-quality weight (0..1)
       i  ∈ {gene, blood, meth, family, lifestyle}

Geometric — not arithmetic — mean: a single weak signal can still pull the score down. Five strong signals must agree.

03 — Threshold

What counts as a finding.

We use two thresholds — calibrated against published prior distributions for each finding class — to gate when the system surfaces and when it auto-applies.

Surfacing
C ≥ 0.70
Finding shown to user; advisor may discuss.
Auto-protocol
C ≥ 0.85
Advisor may propose protocol; user confirms; 5-min undo TTL.
04 — Validation

How we keep ourselves honest.

  • 01Every domain probability is reproducible from the raw inputs — no black-box closed weights.
  • 02Thresholds are re-calibrated quarterly against the closed-beta cohort outcomes.
  • 03Each finding the user sees links to the citations that contributed (ClinVar IDs, GWAS β, clock papers).
  • 04False-discovery rate is reported per finding class on a public dashboard (in development).
05 — Limits

What this is not.

  • 01Not a clinical decision tool. It informs you and your physician — it doesn't replace them.
  • 02Not a diagnostic. The score quantifies signal-agreement, not pathology.
  • 03Not a substitute for monogenic testing. For Mendelian conditions, see a clinical geneticist.
  • 04Limited to the data you provide. Fewer signals → lower weight, lower confidence.
← bioArcBegin a reading