A MAGIBU-based model for pediatric and juvenile CNS tumors: an in-house epigenetic decision-support framework compared with online DNA methylation classifiers.
AI interpretation is pending for this paper.
Open original publication →What the AI sees
Not AI summarized yet.
Research significance
Pending deeper interpretation.
Source abstract
Background: DNA methylation profiling is a tool that provides key support for central nervous system (CNS) tumor classification. However, diagnostically ambiguous pediatric cases may result in discordant outputs across classifiers. We developed MAGIBU, a cross-platform, projection-based framework that embeds individual methylomes into a fixed CNS reference landscape, ranking diagnostic entities by local epigenetic proximity to support clinician-led integrative diagnosis. Methods: As a proof-of-concept, we evaluated MAGIBU in eight morphologically challenging pediatric/juvenile CNS tumors with unresolved diagnoses after institutional and central pathology review. To establish a benchmark in the absence of a definitive histopathological ground truth, a consensus epigenetic reference was defined a priori for cases showing concordant results between the Heidelberg CNS Tumor Methylation Classifier and Methylscape Analysis. Comparisons were also performed with Epigenomic Digital Pathology (EpiDiP). To validate MAGIBU beyond this discovery cohort, performance was assessed at the family level across the CNS methylation spectrum (n = 678, 28 methylation families), on non-array platforms (whole-genome bisulfite sequencing and Oxford Nanopore), and in a focused analysis of the low-grade glioma and diffuse midline glioma compartment across four independent cohorts (n = 670). Results: In the discovery cohort, MAGIBU achieved high concordance with the consensus reference (Cohen's κ = 0.855), outperforming EpiDiP (κ = 0.278), which frequently placed low-grade tumors in proximity to higher-grade reference regions. Conclusions: MAGIBU provides a stable, quantitative differential diagnosis framework that mitigates the limitations of rigid categorical assignments. By leveraging a distance-based proximity metric, it offers a transparent decision-support tool that integrates effectively with clinical, radiological, and molecular data. While performance is inherently dependent on reference atlas composition, MAGIBU represents a robust complementary approach for the diagnostic workup of ambiguous CNS tumors.