Diagnostic concordance of WHO 2021 standards, WGS, and DNA methylation classification in diffuse gliomas: A single-center cohort study

Background: Diffuse gliomas exhibit significant clinical and molecular heterogeneity. However, the real-world concordance among WHO 2021 clinical diagnoses, whole-genome sequencing (WGS), and DNA methylation-based classification remains to be fully characterized. Methods: We retrospectively analyzed 111 glioma patients treated at Ajou University Hospital (2013-2022). Following strict quality control (excluding low tumor fraction <20%, sample swaps, and recurrent-only cases), 93 patients (104 tumors) underwent WGS, enzymatic methyl-seq (EM-seq), and bulk RNA sequencing. Clinical diagnoses were updated to WHO 2021 standards using routine immunohistochemistry and targeted panel sequencing (astrocytoma, n=14; glioblastoma, n=76; oligodendroglioma, n=14). WGS was utilized to refine classifications based on IDH mutation, TERT promoter mutation, EGFR amplification, and chromosome 7 gain/10 loss. Methylation classes (e.g., GBM RTK I/II, Mesenchymal, MID, HGNET-BCOR) were assigned using an AI-driven classifier trained on Illumina array data (~20,000 CpGs). We evaluated the diagnostic concordance across these three layers. Results: Diagnostic concordance was 100% for IDH -mutant astrocytomas and oligodendrogliomas across clinical, WGS, and methylation layers. In contrast, discrepancies emerged in cases clinically diagnosed as glioblastoma. One case was reclassified as pediatric-type diffuse hemispheric glioma, H3 G34-mutant . Another tumor in a 13-year-old patient, harboring complex alterations ( ATRX, PTEN, TP53, PIK3R1, PTPRD, CDKN2A ), was assigned non-GBM methylation labels (CONTR/INFLAM), precluding definitive subclassification. Among tumors classified as GBM RTK I (9/10), GBM RTK II (18/18), or Mesenchymal (15/16) by methylation, nearly all fulfilled molecular glioblastoma criteria ( IDH -wildtype, TERTp mutation, EGFR amp, chr7+/10-). Conversely, tumors in the MID and non-GBM methylation classes frequently lacked these canonical features (17/21). Notably, WGS identified a BCOR::EP300 translocation in a sample with an HGNET-BCOR methylation profile-a fusion undetected by routine clinical testing-expanding the spectrum of BCOR -altered CNS tumors beyond canonical internal tandem duplications. Conclusions: Integrated WGS and methylation profiling in this single-center cohort demonstrate that DNA methylation-based classification refines WHO 2021 diagnoses, particularly for tumors routinely diagnosed as glioblastoma. This approach uncovers rare entities, such as pediatric-type gliomas and BCOR -altered tumors, that may be missed by targeted panels alone. Our findings support the implementation of a multi-omics framework to capture biological heterogeneity obscured by limited molecular markers.