bWGR

bWGR implements Bayesian and likelihood-based whole-genome regression methods in R to perform genome-wide prediction, cross-validation, and association analysis of complex traits.

Key Features:

• Bayesian methods with flexible priors: Provides a suite of Bayesian whole-genome regression models parameterized with various priors to accommodate different genetic architectures.
• Likelihood-based fitting and cross-validation: Supports likelihood-based whole-genome regression model fitting and cross-validation procedures to evaluate predictive performance.
• Gibbs sampling and optimized EM algorithms: Implements traditional Gibbs sampling and optimized Expectation-Maximization methods for parameter estimation.
• Multivariate and hierarchical models: Enables fitting of multivariate models and hierarchical model structures for analyses with multiple traits or nested effects.

Scientific Applications:

• Genome-wide prediction: Estimating genetic merit or trait values from genome-wide marker data.
• Cross-validation studies: Assessing model predictive accuracy and generalizability using cross-validation frameworks.
• Association analysis: Detecting genetic variants associated with traits through whole-genome regression approaches.

Methodology:

Implements Bayesian and likelihood-based whole-genome regression estimation using Gibbs sampling and optimized Expectation-Maximization algorithms, and supports cross-validation as well as multivariate and hierarchical model fitting.