BAYESIAN GROUP LASSO FOR NONPARAMETRIC VARYING-COEFFICIENT MODELS WITH APPLICATION TO FUNCTIONAL GENOME-WIDE...

Area of Science:

• Genetics
• Biostatistics
• Complex Trait Analysis

Background:

• Genome-Wide Association Studies (GWAS) are powerful but face challenges with high-dimensional genetic data, environmental factors, and longitudinal trait analysis.
• Existing GWAS methods struggle to fully capture the dynamic genetic influences on complex traits over time.

Purpose of the Study:

• To develop a novel statistical framework, functional Genome-Wide Association Studies (fGWAS), for integrating functional phenotypic data into GWAS.
• To address the limitations of traditional GWAS in analyzing complex traits with longitudinal or functional characteristics.

Main Methods:

• Proposed a high-dimensional varying-coefficient model incorporating functional aspects of phenotypic traits.
• Developed Bayesian group lasso and Markov Chain Monte Carlo (MCMC) algorithms for SNP identification and estimation of time-varying genetic effects.
• Generalized the model for subject-specific sparse longitudinal data analysis.

Main Results:

• Simulation studies demonstrated the statistical properties of the proposed fGWAS model.
• Applied fGWAS to Framingham Heart Study data, identifying significant single-nucleotide polymorphisms (SNPs) associated with age-specific body mass index changes.
• The model effectively estimates time-varying genetic effects on longitudinal traits.

Conclusions:

• The functional Genome-Wide Association Studies (fGWAS) model, enhanced with Bayesian group lasso, offers a robust approach for genetic and developmental analyses of complex traits.
• This methodology provides a valuable tool for understanding the dynamic genetic architecture of diseases and traits over time.