Inferences from genomic models in stratified populations

Area of Science:

• Genetics
• Bioinformatics
• Statistical Genomics

Background:

• Population stratification can cause spurious associations in genome-wide association studies (GWAS).
• Whole-genome random regression (WGRR) models offer an alternative for fitting all genetic markers simultaneously.
• Existing methods for accounting for stratification in WGRR models are insufficient.

Purpose of the Study:

• To develop a reparameterized WGRR model for simultaneous inference of genetic parameters and population structure.
• To provide a robust method for estimating genomic parameters while accounting for stratification.
• To compare the proposed method with existing approaches for handling population structure in WGRR.

Main Methods:

• Reparameterization of WGRR models using eigenvalue decomposition.
• Simultaneous inference of model parameters and unobserved population structure.
• Application to wheat yield and human complex traits (height, HDL, blood pressure) from the British 1958 cohort.

Main Results:

• The proposed method effectively estimates genomic parameters with and without accounting for stratification.
• Population structure was detected in both wheat and human datasets, impacting inferences differently.
• The novel approach demonstrates advantages over including eigenvectors as fixed effects in WGRR models.

Conclusions:

• The developed method offers a unified, computationally efficient, and easy-to-implement approach for studying population structure and inferring genetic parameters.
• This technique improves the accuracy of genomic analyses by properly addressing population stratification.
• The findings have implications for genetic risk prediction and understanding complex traits.