GW-SEM

GW-SEM integrates structural equation modeling (SEM) into genome-wide association studies (GWAS) to test associations between single nucleotide polymorphisms (SNPs) and multivariate phenotypes or latent constructs.

Key Features:

• Efficient SEM integration: Fits SEMs across the genome to enable testing of complex relationships between SNPs and phenotypic traits.
• Diagonally Weighted Least Squares (DWLS) estimator: Employs a DWLS estimator to fit models with robust parameter estimation and p-value calculation that align closely with full information maximum likelihood methods.
• Supported SEM models: Implements four common SEMs: one-factor model, one-factor residuals model, two-factor model, and latent growth model.
• Comprehensive phenotypic analysis: Accommodates multivariate phenotypes and latent constructs to address stochastic variance in complex traits.
• Performance evaluation: Demonstrated computational efficiency and statistical power through simulations, timing analyses, and comparisons with existing multivariate GWAS software.

Scientific Applications:

• Psychiatric and substance use research: Enables investigation of comorbid traits and complex phenotypic architectures in psychiatric disorders and substance use studies.
• Behavioral genetics: Supports exploration of genetic influences on behavioral traits via multivariate and latent-variable modeling.

Methodology:

Associations between SNPs and multiple phenotypes or latent constructs are tested using SEMs fit genome-wide with a diagonally weighted least squares (DWLS) estimator, and performance is evaluated via simulations and power analyses.