Structural equation modeling with latent variables for longitudinal blood pressure traits using general pedigrees

Area of Science:

• Genetics
• Biostatistics
• Computational Biology

Background:

• Structural Equation Modeling (SEM) is a valuable tool in applied sciences, including genetic analysis.
• The R package 'strum' provides a novel framework for SEM specifically designed for general pedigree data.
• Analyzing complex multivariate longitudinal data, such as blood pressure, requires advanced statistical methods.

Purpose of the Study:

• To explore SEM techniques using the 'strum' package for analyzing multivariate longitudinal data.
• To test the association between single nucleotide polymorphism (SNP) genotypes and blood pressure (BP) traits.
• To apply a novel score test within the SEM framework for genotype association analysis.

Main Methods:

• Utilized the 'strum' R package for SEM on general pedigree data.
• Analyzed quantitative blood pressure traits (systolic BP and diastolic BP) with covariates (age, sex, smoking status).
• Employed autoregressive and latent growth curve models for longitudinal BP data, incorporating censored regression for hypertension treatment adjustment.

Main Results:

• Identified 10 SNPs at a suggestive genome-wide association study (GWAS) P-value level.
• The top 3 most significant SNPs showed consistent ranking across both autoregressive and latent growth curve models.
• Demonstrated the utility of the 'strum' SEM framework for analyzing massive genotype data and complex phenotypes.

Conclusions:

• The 'strum' package offers a robust SEM framework for genetic analysis with pedigree data.
• The approach effectively models longitudinal phenotypes and tests genotype associations.
• This method is highly applicable to large-scale genetic studies involving complex traits and family structures.