GenePEN

GenePEN applies penalized logistic regression with a Pairwise Elastic Net penalty via convex optimization to identify coordinated alterations in genes and proteins within molecular interaction graphs for analysis of functional omics data.

Key Features:

• Pairwise Elastic Net penalty: A novel penalty that prioritizes selection of discriminative genes and proteins based on their interconnectedness within a molecular interaction graph.
• Penalized logistic regression via convex optimization: Solves a convex penalized logistic regression objective and provides provably optimal solutions with an efficient implementation.
• Network-aware selection: Integrates a given molecular interaction network to favor connected groups of genes/proteins during feature selection.
• High-dimensional omics support: Designed to handle high-dimensional functional omics data typical of genomics and proteomics studies.
• Sample classification: Performs sample classification using functional omics data integrated with molecular interaction networks.

Scientific Applications:

• Network-based biomarker identification: Identification of coordinated gene and protein alterations that form connected modules in molecular interaction networks.
• Complex disease analysis: Detection of jointly altered gene/protein groups relevant to understanding disease pathogenesis.
• Therapeutic target prioritization: Prioritization of candidate genes, proteins, and network modules for follow-up in target discovery.
• Genomics and proteomics data analysis: Application to high-dimensional genomics and proteomics datasets for classification and network-informed feature selection.

Methodology:

Optimizes a convex penalized logistic regression objective with a Pairwise Elastic Net penalty using convex optimization on functional omics data integrated with a molecular interaction network to produce provably optimal solutions.