Open-source · R package

End-to-end polygenic risk score analysis

From GWAS retrieval to cohort association in a single reproducible R workflow.

R>install.packages("PolyGenius")

Get started View on GitHub

Holstege Lab · Amsterdam UMC

One workflow, end to end

Specify what you want — let the framework handle how.

generate

Build PRS models from GWAS sources

compute

Score cohort genotypes via PLINK

evaluate

Discrimination, calibration, incremental R²

associate

Regression, survival, mediation, MR

visualize

Publication-ready figures

Two questions, one framework

Once you have models — generated from GWAS, loaded from disk, or downloaded from the PGS Catalog — the next step is always the same: apply them to cohort data.

compute$scores(data, models)→scores applied to your cohort

Evaluation

How well does this PRS predict the outcome?

Assess discrimination (AUC, C-statistic), calibration, risk stratification, and incremental R² over clinical variables. Compare model candidates systematically.

evaluate$performanceevaluate$calibrationevaluate$stratificationevaluate$incremental

Explore evaluation

Association

What does this PRS associate with?

Run inferential models — linear/logistic regression, Cox survival, Fine-Gray competing risks, mediation analysis, Mendelian randomization, and cross-cohort meta-analysis.

associate$regressionassociate$survivalassociate$mediationassociate$mr

Explore association

Up and running in minutes

A full PRS-to-association workflow in three calls.

01Generate a model

models <- generate$models(
  source = generate$sources$opengwas("ieu-b-5067"),
  algorithm = generate$algorithms$ldpred2()
)

02Score your cohort

data <- compute$scores(
  data   = PolyGeniusData(geno, pheno),
  models = models
)

03Associate & evaluate

assoc <- associate$regression(
  data    = data,
  outcome = outcome("AD", type = "binary")
)
eval <- evaluate$performance(data, outcome = "AD")