Linear mixed models for association analysis of quantitative traits with next-generation sequencing data

Area of Science:

• Genetics
• Biostatistics
• Statistical Genetics

Background:

• Genetic association studies are crucial for understanding complex traits.
• Pedigree data presents unique challenges for statistical analysis due to relatedness.
• Existing methods may have limitations in controlling type I error rates and power.

Purpose of the Study:

• To develop and evaluate novel statistical models for gene-based association tests in pedigrees.
• To assess the performance of linear mixed models (LMMs) and functional linear mixed models (FLMMs) in controlling type I error rates and maximizing statistical power.
• To provide robust tools for whole genome and whole exome association studies.

Main Methods:

• Development of linear mixed models (LMMs) and functional linear mixed models (FLMMs).
• Incorporation of fixed effects for major genes and random effects for polygenes.
• Utilizing inbreeding/kinship coefficients to model correlations among pedigree members.
• Construction of F-statistics and chi-squared likelihood ratio test (LRT) statistics for association testing.

Main Results:

• F-distributed statistics demonstrate good control of type I error rates.
• LMM F-test statistics show comparable or superior power to FLMMs, famSKAT, and famBT.
• FLMM F-statistics are effective for analyzing combined rare and common variants.
• FLMM LRT statistics maintain type I error control across various sample sizes, unlike LMM LRT statistics which can be inflated.

Conclusions:

• The proposed LMMs and FLMMs provide reliable methods for gene-based association testing in pedigrees.
• These models offer advantages in type I error control and statistical power for complex trait studies.
• The developed statistical frameworks are valuable for whole genome and whole exome association studies.