iGlioSub

iGlioSub classifies glioblastoma molecular subtypes by integrating gene expression and DNA methylation (DNAm) profiles to stratify tumors into Classical, Mesenchymal, and Proneural subtypes.

Key Features:

• Data integration: Integrates transcriptomic (gene expression) and epigenomic (DNA methylation, DNAm) profiles for combined analysis.
• Algorithms: Employs Random Forest and Nearest Shrunken Centroid algorithms for subtype classification.
• Training data: Developed using 304 GBM patient profiles from The Cancer Genome Atlas (TCGA), the Human Glioblastoma Cell Culture resource (HGCC), and other publicly available databases.
• Performance: Achieves a mean Area Under the Curve (AUC) of 95.9% for the integrated classifier versus 91.9% for gene expression and 93.6% for DNAm classifiers.
• Molecular signatures: Reveals subtype-specific DNA methylation patterns and gene pathway activations associated with each GBM subtype.

Scientific Applications:

• Molecular stratification: Stratifying GBM tumors into Classical, Mesenchymal, and Proneural subtypes for research and translational studies.
• Biomarker discovery: Identifying subtype-specific DNA methylation markers and associated pathway activations.
• Comparative evaluation: Benchmarking integrative classifier performance against expression-only and DNAm-only classifiers using AUC metrics.

Methodology:

Classification used Random Forest and Nearest Shrunken Centroid algorithms on integrated gene expression and DNA methylation (DNAm) profiles, trained on 304 GBM samples from TCGA, HGCC, and other public datasets, with performance assessed by Area Under the Curve (AUC).