Biomarker

Biomarker applies graph Laplacian-regularized logistic regression to integrate biological network information for disease classification and pathway association.

Key Features:

• Graph Laplacian Regularization: Uses graph Laplacian regularization within a logistic regression framework to incorporate prior biological network knowledge, including curated biological networks from the literature.
• Superior Performance: Outperforms elastic net and lasso regression in simulation studies for disease classification accuracy and reliability.
• Application Validation: Validated on a large breast cancer dataset from The Cancer Genome Atlas (TCGA) to differentiate breast cancer subtypes.
• Pathway and Module Identification: Identifies key pathways and protein-protein interaction modules associated with disease subtypes, with many modules corroborated by existing literature.

Scientific Applications:

• Disease Classification: Enables disease classification and subtype discrimination through selection of genes and pathways indicative of disease.
• Pathway Association: Associates gene-level signals with pathways and protein-protein interaction modules to identify disease-associated biological processes.
• Omics and GWAS Analyses: Applies to mining transcriptomic, epigenomic, and other genome-wide association study datasets while integrating biological networks.

Methodology:

Implements graph Laplacian-regularized logistic regression that incorporates curated biological networks; evaluated via simulation studies against elastic net and lasso and validated on TCGA breast cancer data; identifies pathways and protein-protein interaction modules.