glmSparseNet

glmSparseNet generalizes sparse regression models when features such as genes have a graph structure such as protein-protein interactions, by including network-based regularizers with centrality measures as penalty weights.

Key Features:

• Network-based regularization: Incorporates graph structure into sparse regression by using network centrality measures as penalty weights in the regularization.
• Node degree regularization: Implements regularization based on node degree, i.e. the strength and/or number of associated edges, either promoting hubs or orphan genes in the solution.
• Multiple distribution families: Supports all glmnet distribution families: gaussian, poisson, binomial, multinomial, cox, and mgaussian.

Scientific Applications:

• Genomic feature selection: Applies sparse regression to genes or other genomic features with known network relationships such as protein-protein interactions.
• Hub gene identification: Identifies highly connected hub genes in biological networks through degree-based regularization.
• Orphan gene analysis: Identifies poorly connected orphan genes in biological networks through alternative degree-based regularization.

Methodology:

glmSparseNet uses the glmnet R-package, incorporating centrality measures of the network as penalty weights in the regularization.