Polygenic Risk Scores

What Varia publishes on PRS

A polygenic risk score (PRS) combines genotype information at hundreds or thousands of common variants into a single population-level estimate of disease risk. Each variant contributes a small effect; the score sums those effects using weights derived from large GWAS cohorts. PRS describe patterns in research populations. They are not diagnoses and they are not substitutes for cholesterol panels, blood pressure, family history, or the clinical risk tools physicians already use.

Varia's PRS surface is editorial only. It sits alongside The Genome, the Citation Library, and Editorial Standards as a public reference layer. Each domain entry is VEGS-graded and VEAS-cited at the same evidence discipline as the variant catalog. Varia does not aggregate genotypes, apply effect weights, impute missing positions, or emit a percentile, tier label, or any other per-user risk output from a reader's file.

The reputational stake is rigor, not a number. Consumer genomics products that print a heart-disease percentile from a SNP chip overstate what the file can support. Varia's posture is the opposite: explain what validated scores exist, what the authorities say about them, and where a clinically validated result comes from.

Coronary artery disease

Heart & Lipids area

What a PRS is

Coronary artery disease (CAD) is the most studied PRS phenotype in cardiovascular genetics. A CAD PRS counts how many risk-raising alleles a person carries across a defined set of variants, each weighted by its association with CAD in training cohorts. Nearby variants on the same chromosome are often inherited together (linkage disequilibrium), so PRS models use statistical methods that account for correlation among positions rather than treating every variant as independent.

In cohort studies, researchers often report a hazard ratio comparing event rates between people in the highest and lowest PRS groups. That ratio describes the study population. It is not a personal risk output and Varia does not apply it to a reader.

Evidence tier and validation status

CAD is the best-validated PRS application that exists today. The American Heart Association's 2022 scientific statement concludes that a CAD PRS performs similarly to conventional risk factors for prognosis in middle-aged adults and adds prognostic information beyond standard factors alone (O'Sullivan et al., Circulation 2022; PMID 35862132). The ESC Council on Cardiovascular Genomics 2025 clinical consensus documents the strongest cross-cohort validation for CAD PRS among cardiovascular traits (ESC 2025, Eur Heart J).

The current state-of-the-art multi-ancestry score is GPS_Mult (Patel et al., Nat Med 2023; DOI 10.1038/s41591-023-02429-x), registered in the PGS Catalog as PGS003725. It improves risk prediction across ancestries relative to earlier European-centric scores in the published validation cohorts.

VEAS annotations for the anchor citations are available in the Citation Library (PMID 35862132, DOI 10.1093/eurheartj/ehae649, DOI 10.1038/s41591-023-02429-x).

What it adds over traditional risk factors

The honest clinical-utility picture is nuanced. PRS add the most value when conventional risk stratification lands in a borderline or intermediate zone: enough uncertainty that genetic information might shift preventive planning, but not so clear that standard factors already dictate maximal treatment. In people already at high risk by cholesterol, blood pressure, diabetes, or smoking, or already on intensive statin therapy, incremental PRS value is marginal in the published reanalyses.

Statin benefit tracks with genetic risk in post-hoc analyses of large trials. Mega et al. (Lancet 2015; PMID 25748612) reported relative risk reduction from statin therapy of 13%, 29%, and 48% across low, intermediate, and high genetic-risk tertiles in a 27-variant genetic risk score. Natarajan et al. (Circulation 2017; PMID 28223407) found a 44% relative risk reduction in the high genetic-risk group versus 24% in others in the WOSCOPS primary-prevention statin trial using a 57-variant score. These are population-level trial reanalyses. They support the borderline-risk framing. They do not produce a treatment directive for any individual reader.

The ESC 2025 consensus is explicit: European guidelines do not yet advocate routine PRS use in clinical practice. In validation cohorts, people in the highest PRS percentiles carried roughly 3 to 5 times the CAD risk of the lowest percentiles and may derive greater relative benefit from statins in modeled analyses. That framing tempers over-claiming. It is not a recommendation to obtain a score.

Ancestry transferability caveat

GPS_Mult (Patel 2023) pools multi-ancestry GWAS discovery data, but the score itself was trained with LDpred2 on European-ancestry UK Biobank participants. Performance gains in non-European groups in the published validation work still depend on ancestry-matched calibration that consumer files and browser-side tools do not perform.

Martin et al. (Nat Genet 2019; DOI 10.1038/s41588-019-0379-x) showed that European-derived PRS underperform in non-European populations, with the largest gaps in African ancestry groups. Among GWAS published from 2008 to 2017, roughly 67% used European-only discovery cohorts. Any displayed CAD score must be read with that training-data composition in mind.

Where to obtain a clinically validated score

A clinically validated CAD PRS is obtained through the healthcare system, not through a consumer SNP export interpreted in a browser. The pathway runs through an ordering clinician, a laboratory or health-system program with CLIA-validated genotyping or imputation, ancestry-matched calibration, and counseling on what the result means in the context of lipids, blood pressure, family history, and existing treatment.

Authoritative starting points:

• PGS Catalog (pgscatalog.org): provenance, performance metrics, and training-data ancestry for published scores including PGS003725 (GPS_Mult) and smaller validated CAD scores such as PGS000058.
• AHA 2022 scientific statement (PMID 35862132): framework for when PRS evidence is mature enough to discuss with a cardiology team.
• ESC 2025 consensus (DOI 10.1093/eurheartj/ehae649): current European professional posture on implementation.
• Genetic counseling: the National Society of Genetic Counselors directory lists counselors who can place a PRS result in clinical context.

Varia stops at sourced information and these authority pathways. Varia does not offer a PRS product, does not direct dosing or screening intervals, and does not state what a reader should do with a score.

Alzheimer's disease

Brain & Cognition area

What an AD-PRS is

Late-onset Alzheimer's disease (LOAD) PRS combine genotype data at many common variants beyond the APOE locus into a population-level genetic risk estimate. Some published scores include the APOE region in the variant set; others exclude it and model APOE genotype separately. In either design, the APOE ε4 allele remains by far the largest common-variant effect in LOAD genetics. The remaining PRS captures smaller contributions spread across the genome, each weighted from large GWAS cohorts.

AD-PRS describe patterns in research populations. They are not diagnoses, not substitutes for amyloid or tau biomarkers, and not treatment-decision tools. Varia does not apply any AD-PRS to a reader or emit a percentile, tier label, or other per-user genetic risk output.

Evidence tier and validation status

Bellou et al. (Alzheimers Res Ther 2025; PMID 39762974) benchmarked multiple AD prediction approaches across methodologies and cohorts. The best-performing models combined APOE genotype with a remaining PRS, reaching area-under-the-curve values of roughly 0.72 to 0.76 in the published validation work. Adding the PRS beyond APOE alone improved prediction modestly. The gain is real in cohort analyses but small in absolute terms.

No major neurology guideline body endorses routine clinical use of an AD-PRS for diagnosis, screening, or treatment selection. Published scores are research and risk-stratification tools. They are not validated individual clinical results without laboratory calibration, ancestry-matched reference data, and counseling in a care setting.

VEAS annotations for the anchor citations are available in the Citation Library (PMID 39762974, PMID 34386572, DOI 10.1038/s41588-019-0379-x).

What it adds beyond APOE

The honest framing next to Varia's APOE flagship: a PRS carries information beyond APOE genotype alone, including in people who carry APOE ε4 alleles. That incremental signal is modest and population-level. It does not overturn APOE as the dominant common-variant factor and it does not produce an individual prediction on this page.

Huq et al. (Alzheimers Dement (Amst) 2021; PMID 34386572) studied APOE ε4/ε4 homozygotes at phenotypic extremes: people with LOAD versus older cognitively healthy ε4/ε4 controls. AD-PRS were higher in the LOAD cases (odds ratio 8.39 in the published comparison, P = .003). The PRS further stratified risk within the ε4/ε4 genotype at the extremes of cognitive phenotype in that cohort. PRS also carry signal in ε4 non-carriers in broader LOAD genetics literature; Bellou 2025 treats APOE and the remaining PRS as complementary predictors rather than substitutes.

These statistics describe study populations and published model performance. They are not applied to any reader here. Many ε4/ε4 carriers never develop LOAD; a PRS does not change that clinical reality and Varia does not present one as a personal forecast.

Ancestry transferability caveat

Martin et al. (Nat Genet 2019; DOI 10.1038/s41588-019-0379-x) showed that European-derived PRS underperform in non-European populations, with the largest gaps in African ancestry groups. Among GWAS published from 2008 to 2017, roughly 67% used European-only discovery cohorts. LOAD PRS inherit the same transferability limits as cardiovascular PRS because the underlying effect-size estimates are ancestry-specific.

Where to obtain a clinically validated assessment

LOAD risk genetics is discussed in clinical care through APOE genotyping, biomarker testing (amyloid, tau), cognitive evaluation, and family-history review. A research-grade AD-PRS from a consumer SNP export is not a substitute for that pathway. AD genetics informs risk context; it does not direct drug choice, supplement protocols, or screening intervals on its own.

Authoritative starting points:

• PGS Catalog (pgscatalog.org): provenance, performance metrics, and training-data ancestry for published LOAD scores.
• Bellou 2025 benchmarking study (PMID 39762974): current methodological comparison of APOE-plus-PRS prediction approaches.
• Varia APOE catalog: single-variant diplotype interpretation at APOE & Alzheimer's Risk and the APOE guide.
• Genetic counseling: the National Society of Genetic Counselors directory lists counselors who can place genetic risk information in clinical context.

Varia stops at sourced information and these authority pathways. Varia does not offer an AD-PRS product, does not direct treatment or screening, and does not state what a reader should do with genetic risk information.

References

Every claim on this page traces to a verified primary source from the PRS citation register (2026-06-10). VEAS slot summaries below follow the Citation Library annotation shape.

• O'Sullivan JW, et al. Polygenic Risk Scores for Cardiovascular Disease: A Scientific Statement From the American Heart Association. Circulation 2022;146(8):e93-e118. PMID 35862132; DOI 10.1161/CIR.0000000000001077. VEAS: CAD is the most-studied PRS phenotype; CAD PRS perform similarly to conventional risk factors and add prognostic information in middle-aged adults.
• ESC Council on Cardiovascular Genomics, et al. Clinical utility and implementation of polygenic risk scores for predicting cardiovascular disease. Eur Heart J 2025;46(15):1372-1383. DOI 10.1093/eurheartj/ehae649. VEAS: ESC guidelines do not yet advocate routine PRS use; highest PRS percentiles carry 3-5x CAD risk in validation cohorts.
• Patel AP, et al. A multi-ancestry polygenic risk score improves risk prediction for coronary artery disease (GPS_Mult). Nat Med 2023;29:1793-1803. DOI 10.1038/s41591-023-02429-x; PGS Catalog PGS003725. VEAS: Multi-ancestry GWAS discovery; LDpred2 training on European-ancestry UK Biobank.
• Mega JL, et al. Genetic risk, coronary heart disease events, and the clinical benefit of statin therapy. Lancet 2015;385:2264-2271. PMID 25748612. VEAS: Statin relative risk reduction 13%, 29%, and 48% across low, intermediate, and high genetic-risk tertiles.
• Natarajan P, et al. Polygenic risk score identifies subgroup with higher burden of atherosclerosis and greater relative benefit from statin therapy in the primary prevention setting. Circulation 2017;135(22):2091-2101. PMID 28223407. VEAS: WOSCOPS primary prevention; statin RRR 44% in high genetic risk versus 24% in others.
• Martin AR, et al. Clinical use of current polygenic risk scores may exacerbate health disparities. Nat Genet 2019;51:584-591. DOI 10.1038/s41588-019-0379-x. VEAS: European-derived PRS transfer poorly to non-European ancestries; ~67% of 2008-2017 GWAS were European-only.
• Bellou E, et al. Benchmarking Alzheimer's disease prediction: personalised risk assessment using polygenic risk scores across various methodologies and genome-wide studies. Alzheimers Res Ther 2025 Jan 6. PMID 39762974; DOI 10.1186/s13195-024-01664-9. VEAS: Best prediction from APOE plus remaining PRS (AUC 0.72-0.76); PRS adds modest accuracy beyond APOE.
• Huq AJ, et al. Polygenic score modifies risk for Alzheimer's disease in APOE ε4 homozygotes at phenotypic extremes. Alzheimers Dement (Amst) 2021. PMID 34386572; DOI 10.1002/dad2.12226. VEAS: AD-PRS higher in ε4/ε4 LOAD cases vs older cognitively healthy ε4/ε4 controls (OR 8.39, P = .003); PRS stratifies within ε4/ε4 at phenotypic extremes.

Editorial discipline for this surface follows Editorial Standards and Methodology. Single-variant APOE interpretation lives in the Brain & Cognition APOE catalog entry. Pharmacogenomic variant interpretation for statin response (SLCO1B1, CYP2C19) lives on the Medication Response page and in the Heart & Lipids domain catalog.