Research program

Evaluating Polygenic Risk Scores at Scale

How well does a polygenic risk score work outside the study where it was developed?

We compare published scores in large biobanks and release the results, methods, and score files so that other researchers can inspect the evaluations.

Why external evaluation matters

Polygenic risk scores often perform best in the data used to develop them. Testing them in another population or health system, or with a different phenotype definition, shows which results replicate, which differ across datasets, and which need further study.

What we evaluate

Start with the published score

Each evaluation records the score definition, trait, and source study so readers can understand the comparison.

Test it in independent data

PRSweb reports evaluations from the Michigan Genomics Initiative and UK Biobank side by side.

Report more than one metric

Calibration, discrimination, and phenome-wide associations answer different questions about a score's performance.

Keep each release tied to its paper

Each archived PRSweb release includes the score files, mappings, results, and citation for its published analysis.

Selected work

These papers compare disease and exposure scores across biobanks, assess performance across ancestry groups, and test whether scores distinguish clinically important disease. The portal cards link to all three PRSweb releases.

Current work

Projects and collaborations

Public research resource

Current collection with archived releases

PRSweb Research Portals

PRSweb provides polygenic risk score evaluations, phenome-wide association results, and downloadable score files.

Related resources

Tools & data

Polygenic risk resource

Public collection

PRSweb Research Portals

Compare published polygenic risk scores evaluated in the Michigan Genomics Initiative (MGI) and UK Biobank, and explore phenome-wide associations and downloadable files.

View resource details

Polygenic risk archive

Archived release

Cancer PRSweb

The archived June 2020 release reports evaluations of cancer polygenic risk scores in the Michigan Genomics Initiative (MGI) and UK Biobank.

More than 305 filtered scores across more than 39 retained traits.

View resource details

Polygenic risk archive

Archived release

ExPRSweb

The archived November 2021 release reports evaluations of polygenic risk scores for common health-related exposures.

More than 336 filtered scores across more than 28 retained traits.

View resource details

Polygenic risk archive

Archived release

Skin Cancer PRSweb

The original June 2019 release evaluates 20 polygenic risk scores for melanoma, basal cell carcinoma, squamous cell carcinoma, and overall skin cancer.

Twenty scores across four traits in the June 2019 release.

View resource details

PheWAS software

Public

phewasFlow

An R package for reproducible phenome-wide association studies in both directions, with local and cluster workflows, multiple-testing correction, and Manhattan and volcano plots.

View resource details

Publications

Selected papers

All publications →
2020

Cancer PRSweb: An Online Repository with Polygenic Risk Scores for Major Cancer Traits and Their Evaluation in Two Independent Biobanks

American Journal of Human Genetics

Why it matters: Cancer PRSweb evaluated published cancer scores with the same methods in two independent biobanks. The side-by-side results show how performance differs by source GWAS, score-building method, phenotype definition, and cohort.

2022

ExPRSweb: An online repository with polygenic risk scores for common health-related exposures

American Journal of Human Genetics

Why it matters: Smoking, body mass, lipid levels, and other exposures that shape disease risk are absent or unevenly recorded in clinical data. ExPRSweb evaluates whether genetic predisposition to these exposures adds information to phenome-wide analyses and prediction. These genetic scores can complement measured exposure data, but they do not replace it.

2021

On cross-ancestry cancer polygenic risk scores

PLOS Genetics

Why it matters: Breast and prostate cancer scores built from genome-wide association studies in people of European ancestry performed differently across ancestry groups in UK Biobank, so the same absolute cutoff did not apply to every group. Within each group, rankings still separated people by risk. The paper shows the difference between a score that ranks risk within a group and one that can be interpreted the same way across groups.

2025

Assessing the Clinical Utility of Published Prostate Cancer Polygenic Risk Scores in a Large Biobank Data Set

European Urology Oncology

Why it matters: Researchers tested 16 published prostate cancer scores in the Michigan Genomics Initiative (MGI) and used detailed biopsy data to evaluate their clinical utility. Even the best-performing score separated cases from controls only modestly, and none distinguished aggressive from indolent disease. Predicting a diagnosis is not the same as identifying the cancers that most need treatment.

Updates

News and talks

· Video

MGI 2024: Lars Fritsche

In this U-M symposium talk, Lars Fritsche reviews a decade of EHR-linked genetics research, including PheWeb and PRSweb.