IMPLEMENTATION OF DNA METHYL-ATION PROFILING FOR DIAGNOSIS OF CENTRAL NERVOUS SYSTEM TUMORS: THE FIRST BRAZILIAN EXPERIENCE

Introduction: Central Nervous System (CNS) tumors are currently defined by the WHO 2021 classification. It presents a wide spectrum of clinically and biologically heterogeneous tumors, which makes the reproducibility of the histopathological diagnosis challenging. DNA methylation profiling has revolutionized diagnostic neuropathology since its implementation. It associates a high throughput methylation array with a machine learning classifier to predict the diagnosis of brain tumors. This approach allows appropriate classification and molecular subclassification within tumor classes. DNA methylation-based classification is a novel and powerful tool in diagnostic neuropathology, enabling precise diagnosis and adequate patient treatment. Objectives: Implement and validate DNA methylation profiling for CNS tumor diagnosis in a brazilian neuropathological routine, comparing its results with standard histopathological analysis. Methodology: 44 samples from 40 formalin-fixed and paraffin-embedded (FFPE) CNS samples were submitted to the Infinium Methylation EPIC array. The raw data obtained were submitted to the Brain Tumor Classifier, available online. The classifier provides a calibrated score (CS) for the most compatible methylation class (MC). The CS cut-off for a MC family match was ≥ 0.84. For molecular subclasses the match was considered with a CS ≥ 0.5. Cases below these thresholds were considered as no match. For cases with a “match”, the MC was compared to the original diagnosis. The results were categorized as concordant, refined or discordant and a descriptive analysis was performed. Results: 35 samples (79,5%) received a match CS. Of the 9 cases with no match, 3 were considered unsolved, 3 had low DNA quality, 2 showed low cellularity and 1 was a new entity (Intracranial mesenchymal tumor, FET::CREB positive). For cases with a match, 18 (51,4%) had a concordant diagnosis. A refined classification was achieved in 12 (34,3%) cases and a discordance between MC and original diagnosis occurred in 5 (14,3%) cases. For three of the latter, the methylation diagnosis was considered more appropriated. In 2 cases, the immunomorphological diagnosis was kept. In total, the initial diagnosis was changed or refined in 42,8% of cases. Conclusion: DNA methylation profiling is a robust and feasible tool to be implemented in a neuropathology routine, with potential to refine and provide more accurate diagnosis in a significant proportion of cases.