MGREML

MGREML estimates SNP-based heritability and genetic correlations across multiple traits using genomic-relatedness-based restricted maximum likelihood for multivariate analysis of complex trait architecture.

Key Features:

• Multivariate Estimation: Implements genomic-relatedness-based restricted maximum likelihood (GREML/REML) to estimate SNP-based heritability and genetic correlations simultaneously for multiple traits.
• Factor Model Fitting: Fits and tests user-specified factor models to characterize the underlying genetic architecture of complex traits.
• Computational Efficiency: Maintains low computational complexity, enabling analyses on large datasets (e.g., a saturated model with multiple factors and fixed effects can be estimated in under an hour on two 2.7 GHz cores with 16 GB RAM).
• Simulation Validation: Demonstrates consistent estimates and valid statistical inference for factor models in simulation studies.
• Structural Equation Modeling: Supports simple structural equation modeling to specify, estimate, and compare genetic factor models using SNP data.

Scientific Applications:

• Estimation of Heritability: Quantifies the proportion of phenotypic variance attributable to SNPs.
• Genetic Correlation Analysis: Assesses the shared genetic basis between different traits via SNP-based genetic correlations.
• Factor Model Testing: Evaluates whether a specified genetic factor model fits the data better than alternative models for hypothesis testing in genetic studies.

Methodology:

Leverages SNP-derived genomic relationships and applies restricted maximum likelihood (GREML/REML) for multivariate estimation of heritability and genetic correlations while fitting user-defined factor models.