Research

Validated polygenic risk, calibrated across ancestries.

ARC-omix scores are validated in two ancestrally varied U.S. cohorts — eMERGE IV (n=19,348) and All of Us (n=239,645) — with per-cluster performance reported across European-, African-, and Hispanic/Latino-ancestry participants.

View publicationsRead the methods
Methods

How ARC-omix scores are trained, calibrated, and controlled.

Training
Multi-ancestry meta-analysis

Scores draw on large genetic studies spanning UK Biobank, All of Us, FinnGen, Million Veteran Program, and PAGE, covering ancestrally varied populations.

Scoring
Ensemble per-condition models

Each condition is scored with a dedicated PRSMix ensemble integrating PRS-CS, LDpred2, lassosum2, and clumping-and-thresholding through penalized regression on a held-out tuning set.

Calibration
Ancestry-stratified reclassification

Each score is interpreted alongside standard clinical risk tools (the Pooled Cohort Equations) and family history, with discrimination and calibration evaluated separately by ancestry cluster.

Quality
CLIA-ready production pipeline

Testing runs on Blended Genome-Exome sequencing performed by our partner lab — a CLIA-certified, CAP-accredited facility.

Validation — Coronary Heart Disease

Polygenic risk and family history are independently and additively informative.

The C-statistic of the Pooled Cohort Equations rises from 0.719 to 0.753 when polygenic risk and family history are added together.

C-Statistic · CHD Risk Prediction · eMERGE IV (n=19,348)
Pooled Cohort Equations (PCE)0.719
95% CI: 0.700–0.737
PCE + Polygenic Risk Score
0.735+0.016
95% CI: 0.717–0.753
PCE + PRS + Family History
0.753+0.034 ←
95% CI: 0.735–0.770
eMERGE IV (n=19,348). In adults <40 yr, c-statistic rises from 0.757 to 0.827.
Reclassification at 7.5% 10-Year Threshold
PCE + PRSUp 3.5% · Down 4.7%
PCE + PRS + Family HistoryUp 4.6% · Down 14.1%
Net benefit: ~4 additional true positives per 1,000 screened at the 7.5% threshold.
Additive CHD Hazard Ratio · All of Us (n=239,645)

Patients in the top 5% of PRS with a positive family history have 3.79× the hazard of CHD vs. neither. Cox model adjusted for age, sex, 5 PCs.

Neither high PRS nor positive FamHx
HR 1.00Reference
High PRS only (top 5%)
HR 1.8595% CI 1.41–2.43
Positive family history only
HR 1.7895% CI 1.59–1.99
High PRS + Positive Family History
HR 3.79 ←95% CI 2.98–4.82
All of Us (n=239,645). Top 5% PRS threshold. Cox proportional hazards model.
Clinical interpretation

PRS and family history capture partially non-overlapping genetic risk. A patient with both carries nearly 4× the baseline risk — a signal that standard clinical assessments miss.

Reference cohorts

Validated across large, ancestrally varied populations.

eMERGE IV
19,348
U.S. multi-site · Primary validation cohort
All of Us
239,645
U.S. NIH · Prospective replication cohort
UK Biobank
~500k
UK · Training data source
FinnGen
~520k
Finland · Training data source
MVP
~650k
U.S. Veterans · Training data source
PAGE
~50k
Multi-ancestry · Training data source
Relevant articles

Publications

2026

Review

Clinical use of polygenic risk scores: current status, barriers and future directions

Kullo IJ, et al.

Nature Reviews Genetics · 2026

View on PubMed →

2025

RCT

Effect of disclosing a polygenic risk score for coronary heart disease on adverse cardiovascular events

Kullo IJ, et al.

Circulation: Genomic and Precision Medicine · 2025

View on PubMed →

2024

Cohort

Associations of self-reported race, social context, and polygenic risk with coronary heart disease

Kullo IJ, et al.

Journal of the American College of Cardiology · 2024

View on PubMed →

2022

Clinical

Polygenic risk score for peripheral artery disease

Kullo IJ, et al.

Vascular Medicine · 2022

View on PubMed →

2016

RCT

Incorporating a genetic risk score into coronary heart disease risk estimates: effect on LDL cholesterol levels

Kullo IJ, Jouni H, Austin EE, et al.

Circulation · 2016

View on PubMed →