Skip to content

REGENIE

Summary

REGENIE is a C++ software package designed for genome-wide association studies (GWAS) and exome-wide association studies (ExWAS) in large biobanks. It uses a two-step machine learning framework to account for population structure and sample relatedness, offering dramatic speedups and memory efficiency compared to traditional linear mixed models (LMMs).

Overview

REGENIE fits a whole-genome regression model to control for population stratification and cryptic relatedness. - Step 1: Runs a ridge regression on blocks of common variants to construct a predictor of the individual trait values. - Step 2: Tests association between individual genetic variants (both common and rare coding variants) and the trait, adjusting for the predictor constructed in Step 1.

In the multi-ancestry lipid study by Koyama et al. (2026), REGENIE (version 3.1.3) was used for exome-wide association analysis of total cholesterol (TC), low-density lipoprotein cholesterol (LDLC), high-density lipoprotein cholesterol (HDLC), and triglycerides (TG) across MVP, UK Biobank, and All of Us cohorts.

Features

  • Handles both quantitative and binary traits (using Firth logistic regression in Step 2).
  • Highly scalable, supporting sample sizes of over 1 million individuals.
  • Efficiently processes rare variants, including predicted loss of function (pLoF) and missense alleles.

Citations

  • [1] Mbatchou, J. et al. (2021). Computationally efficient whole-genome regression for quantitative and binary traits. Nature Genetics, 53(7), 1097–1103. DOI: 10.1038/s41588-021-00870-7
  • [2] Koyama, S. et al. (2026). Exome-wide association study of blood lipids in 1,158,017 individuals from diverse populations. Nature Genetics, 58(6), 1268-1279. DOI: 10.1038/s41588-026-02613-y. Source paper: s41588-026-02613-y.pdf