netReg

netReg applies graph-penalized regression to integrate biological network priors into linear models for improved estimation of regression coefficients in genomic studies.

Key Features:

• Graph-Penalized Regression Models: Integrates graph-based prior information into the loss function of linear models to improve estimation of regression coefficients in network-structured genomic data.
• Sparse and Smooth Solutions: Uses biological network-informed penalties to derive sparse or smooth regression coefficient estimates, enhancing interpretability and robustness.
• Computational Efficiency: Implemented in R and C++ with C++ core computations using Armadillo for matrix calculations and Dlib for optimization to enable analysis of large networks and thousands of variables.

Scientific Applications:

• Modeling Multivariate Genomic Responses: Models complex interactions among genomic variables to estimate multivariate genomic responses aligned with known biological pathways.
• Network-Guided Analysis of Molecular Interactions: Incorporates biological networks such as protein-protein interactions and gene regulatory networks into regression analyses to reflect known dependencies among variables.

Methodology:

Employs penalized regression techniques that integrate graph-based prior knowledge into the model loss function; implemented in R and C++ with Armadillo for matrix algebra and Dlib for optimization to address high-dimensional genomic data.