BGLR

BGLR implements single-trait and multitrait shrinkage and variable-selection Bayesian regressions for genomic analyses.

Key Features:

• Single-Trait and Multitrait Models: Supports both single-trait and multitrait regression models for joint analysis of multiple phenotypes.
• Shrinkage and Variable Selection: Implements shrinkage and variable-selection Bayesian regression approaches for high-dimensional predictor sets.
• Random-Effects Terms: Allows inclusion of an arbitrary number of random-effects terms to model multiple sources of variability.
• Prior Distributions: Provides multivariate prior choices per predictor set, including diffuse, Gaussian, and Gaussian-spike-slab priors.
• Covariance Parameter Specifications: Supports unstructured, diagonal, factor-analytic, and recursive specifications for (co)variance parameters in multitrait regressions.
• Posterior Sampling via Gibbs Sampler: Generates posterior samples using a Gibbs sampler with computation implemented in R and C.

Scientific Applications:

• Genomic Studies: Applied to analysis of complex traits in genomic datasets.
• Genome-Wide Association Studies (GWAS): Usable for modeling trait associations across genome-wide markers.
• Quantitative Trait Loci (QTL) Mapping: Applicable to mapping and estimation of QTL effects under multivariate models.
• Genetic Prediction: Employed for genomic prediction of phenotypes using shrinkage and variable-selection priors.

Methodology:

BGLR performs Bayesian regression with shrinkage and variable selection using multivariate (diffuse, Gaussian, Gaussian-spike-slab) priors and user-specified (unstructured, diagonal, factor-analytic, recursive) covariance parameterizations, obtaining posterior samples via a Gibbs sampler implemented in R and C.