[Contribution of DNA methylation profiling to the diagnosis of primary central nervous system tumors].
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
INTRODUCTION: The 2021 WHO classification of central nervous system (CNS) tumors has reshaped the diagnostic approach in Neuro-Oncology. The development of increasingly advanced molecular techniques has allowed a histomolecular, or "integrated," diagnosis. DNA methylation profiling is key to some diagnoses, particularly in pediatrics. Although the contribution of this technique is undeniable, results generated by artificial intelligence-based algorithms must be interpreted with caution. The aim of this study was to evaluate the contribution of DNA methylation profiling to CNS tumor diagnostics since its implementation in the department in 2021. MATERIALS AND METHODS: The retrospective cohort included 256 CNS tumors that underwent integrated diagnosis with DNA methylation profiling (EPICv1 arrays, Illumina). Methylation data were analyzed using the classifier of Heidelberg University, Germany. The diagnosis proposed by the classifier was compared with the pre-methylation integrated diagnosis. The cases were divided into five categories: category I: concordant diagnoses; category II: concordant diagnoses with additional diagnostic refinement provided by methylation profiling; category III: discordant diagnoses within the same tumor family; category IV: discordant diagnoses involving a change in tumor family/superfamily; category V: non-contributory methylation results (calibrated score (CS)<0.5 or control tissue classification). Discordant cases were reviewed to establish a final diagnosis either supporting or not the diagnosis proposed by the Heidelberg classifier. RESULTS: In 19% cases (48/256), the methylation-based diagnosis was perfectly concordant with the pre-methylation diagnosis (category I), and in 41% cases (104/256), it provided additional diagnostic refinement (category II). In 9% cases (23/256), the methylation-based diagnosis was partially discordant (category III), and in 17% cases (45/256), it was discordant (category IV). Among these 45 cases, only 26 were truly discordant (with CS>0.9, 10% [26/256]). In 15 cases (5.8%; 15/256), methylation results led to a revision of the initial integrated diagnosis. In 14% cases (36/256), the methylation-based diagnosis was non-contributory (category V). CONCLUSION: In 71% cases with CS>0.9, histomolecular and methylation-based diagnoses were concordant. In more than 80% of these cases, methylation profiling provided additional information. In only 5.8% cases did methylation results lead to a change in the final diagnosis. DNA methylation profiling is a valuable tool for confirming and refining CNS tumor diagnosis. However, its results must always be interpreted in conjunction with clinical, radiological, histopathological, and other molecular data.