VIGoR

VIGoR performs variational Bayesian inference to jointly estimate multiple linear learners for genome-wide regression and variable selection in high-dimensional multimodal datasets.

Key Features:

• R package implementation: Implemented as an R package for execution within R-based analysis environments.
• Multiple linear learners: Supports joint estimation of diverse linear learners within a unified model.
• Penalized regression and spike and slab priors: Integrates penalized regression methods and spike and slab priors for variable selection and shrinkage.
• Variational Bayesian inference: Uses variational Bayesian inference to approximate posterior distributions for scalable Bayesian estimation.
• Fast minorize-maximization algorithms: Employs fast minorize-maximization algorithms to accelerate convergence of variational updates.
• Handling multimodal and high-dimensional explanatory variables: Accommodates multimodal and high-dimensional covariates typical of genomic datasets.

Scientific Applications:

• Genome-wide regression: Applied to genome-wide regression analyses to model associations between genotypes and phenotypes.
• Variable selection in genomic studies: Used for variable selection and shrinkage in high-dimensional genetic data.
• Multimodal data integration: Facilitates integration of multimodal explanatory variables in a single regression framework.
• Genetic research and personalized medicine: Supports uncovering complex relationships in genomic data relevant to genetic research and personalized medicine.

Methodology:

Integrates penalized regression and spike and slab priors into a joint-model framework and estimates parameters via variational Bayesian inference using fast minorize-maximization algorithms.