Fast and flexible linear mixed models for genome-wide genetics.

Area of Science:

• Quantitative genetics
• Bioinformatics
• Statistical genomics

Background:

• Linear mixed effect models are crucial for genome-wide association studies (GWASs) but can be computationally intensive.
• Simplifications of these models may reduce statistical power or introduce bias in genetic analyses.
• Accounting for population structure and multiple sources of variation is essential for accurate genetic architecture estimation.

Purpose of the Study:

• To introduce Grid-LMM, an efficient and extendable algorithm for fitting complex linear mixed models.
• To enable the analysis of multiple sources of heterogeneity, including genetic and environmental factors, in large-scale genetic studies.
• To improve the speed and accuracy of genome-wide association studies and quantitative trait analyses.

Main Methods:

• Developed Grid-LMM, an algorithm designed for the rapid fitting of complex linear mixed models.
• Implemented methods to account for additive and non-additive genetic variance, spatial heterogeneity, and genotype-environment interactions.
• Utilized Grid-LMM to compute approximate frequentist test statistics and Bayesian posterior summaries for genome-wide analyses.

Main Results:

• Grid-LMM significantly reduces computation time compared to existing general-purpose methods while maintaining high accuracy.
• Application to quantitative genetic analyses revealed new discoveries by modeling multiple sources of heterogeneity.
• Identified gene expression traits affected by non-additive genetic variation and QTL by accounting for spatial variability.

Conclusions:

• Grid-LMM offers a computationally efficient solution for complex genetic analyses, overcoming limitations of traditional methods.
• Modeling multiple sources of heterogeneity using Grid-LMM leads to enhanced power and novel discoveries in genetic studies.
• The algorithm facilitates more comprehensive understanding of the genetic architecture of complex traits and gene expression.