Regenie.QRS: computationally efficient whole-genome quantile regression at biobank scale

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genotype-phenotype associations are often dynamic and context-dependent, leading to heterogeneous genetic effects across a phenotype's distribution.
• Traditional linear regression may not fully capture these complex genetic effects, necessitating advanced analytical methods.

Purpose of the Study:

• To introduce Regenie.QRS, a computationally efficient whole-genome quantile regression technique for analyzing biobank-scale genome-wide association studies (GWAS) data.
• To detect and characterize heterogeneous genotype-phenotype associations more effectively than existing methods.

Main Methods:

• Proposed a novel whole-genome quantile regression method, Regenie.QRS, designed for large-scale GWAS data incorporating genetic structure.
• The method estimates polygenic effects and integrates them as an offset within a non-mixed quantile regression model.
• Validated through simulations and applications on UK Biobank and ProgeNIA/SardiNIA datasets.

Main Results:

• Simulations confirmed robust control of type I error and increased power for detecting heterogeneous associations compared to linear regression.
• Regenie.QRS demonstrated improved power over marginal quantile regression tests.
• Real-world applications highlighted the method's advantage in identifying and characterizing heterogeneous genetic effects, exemplified by the G6PC2 locus's role in glucose regulation.

Conclusions:

• Regenie.QRS is a powerful and efficient tool for uncovering complex genotype-phenotype relationships in large-scale GWAS.
• The method successfully identifies genetic variants with effects that vary across the phenotype distribution, offering deeper biological insights.
• Findings illustrate the utility of quantile regression in understanding genetic contributions to complex traits, such as the protective role of G6PC2 against hypoglycemia.