A cellular epigenetic classification system for glioblastoma.

Area of Science:

• Oncology
• Genomics
• Bioinformatics

Background:

• Glioblastoma (GBM) exhibits significant cellular heterogeneity, impacting tumor progression and treatment outcomes.
• Single-cell profiling offers detailed insights but is not feasible for large studies or clinical use.
• Current DNA methylation diagnostics lack cellular resolution.

Purpose of the Study:

• To develop a computational method for deconvoluting bulk DNA methylation profiles in GBM.
• To infer cellular composition, including microenvironment cells and malignant cell differentiation states.
• To reconcile different GBM classification systems and enhance diagnostic accuracy.

Main Methods:

• Utilized a hierarchical non-negative matrix factorization approach (ITHresolveGBM).
• Applied the method to deconvolve bulk DNA methylation data.
• Integrated multi-omic single-cell data for validation.

Main Results:

• ITHresolveGBM identified low tumor cell content as a factor affecting methylation-based classification, particularly linking the mesenchymal subtype with high immune infiltration.
• Epigenetic deconvolution revealed a malignant cell differentiation continuum, correlating with molecular drivers (e.g., PDGFRA, TP53, EGFR) and survival.
• The inferred continuum aligns with existing GBM classification systems.

Conclusions:

• The ITHresolveGBM framework successfully reconciles DNA methylation and RNA-based classification systems.
• It provides a method to integrate bulk tumor profiles with single-cell biology insights.
• This approach refines molecular stratification and enhances GBM diagnostics.