No abstract

Current cIMPACT-NOW and WHO guidelines state that the accurate diagnosis of many central nervous system (CNS) tumors must not only incorporate traditional histological findings but also include additional molecular data derived from sequence and methylation profiling techniques. To address this clinical need, our group has adapted and clinically validated the CNS tumor classifier that was pioneered and developed by the German Cancer Research Network (DKFZ) and University Hospital Heidelberg for both adult and pediatric CNS tumors. The classifier analyzes whole genome methylation data derived from the Illumina EPIC array system with a machine learning algorithm trained on a large reference set to provide a tumor family and class (a more granular classification than family) along with calibrated scores to aid in determining confidence in the assigned classifications. Around this core algorithm, we have generated a computational and reporting pipeline to generate a mid-throughput clinical diagnostic assay. We have validated this assay using a combination of adult and pediatric CNS tumor samples and have established reliable run-level and sample-level quality control metrics with empirically defined thresholds. With these thresholds, approximately 90% of samples with a neoplastic content of 70% or greater were “classifiable”, such that a result could be returned by the CNS tumor diagnostic assay. The performance of the assay was exceptional, with a sensitivity, specificity and accuracy each greater than 98% as determined with a validation sample set of 105 specimens. Further, 4 discordant calls between the original diagnosis and the output of the new methylation-based diagnostic were identified and underwent further histopathologic review. This additional analysis led to the re-classification of 2 specimens due to molecular analysis, highlighting the clinical utility of the assay. The new assay is currently offered for clinical testing under the name JAX OncoMethyl ArrayTM and addresses a clear need in the field of CNS tumor diagnostics. Citation Format: Melissa Soucy, Nick Renzette, Xinming Zhuo, Prasanti Nuni, Kristi Herlth, Kevin Kelly, Gregory Omerza, Lei Li. Clinical validation of a methylation array-based diagnostic assay for improved classification of CNS tumors [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 2233.

BACKGROUND Meningiomas are the most common intracranial tumors. According to the World Health Organization’s (WHO) classification guidelines, they can be classified into 15 subtypes with three grades of malignancy based on histology alone. In recent years, analysis of tumor methylomes that reflects both cell-origin methylation signatures and somatically acquired DNA methylation changes has been utilized to better classify meningiomas with great success. Method: In our case series, we report DNA methylation profiling on meningiomas from 17 patients. Formalin fixed paraffin embedded (FFPE) meningioma tumor samples were processed, loaded onto the Infinium Methylation EPIC array, and scanned using the Illumina IScan system. Raw IDAT files were then processed through the v12b6 version of the CNS tumor classifier developed by the Molecular Neuropathology group at the University of Heidelberg. Classifier output consisted of assigned methylation classifications along with confidence scores, with scores ≥0.84 classified as “high confidence.” Result: Among our meningioma samples from 17 patients (18 samples total), 14 samples are classified as “benign,” one sample is classified as “intermediate,” and the remaining three samples from two patients are classified as “malignant.” In addition to tumor methylation profiling, we also present information that includes patient demographics, clinical presentations, tumor characteristics (size, location), surgical approaches, and mutational analysis. Genomic analysis has revealed variants in NF2, SLX4, ARID1A, KDM4C, ERCC2, EP300, ERBB2, KMT2D, LRP1B, FANCE, CRKL, and SPEN in multiple patients. The three meningioma samples from the two patients that are classified as “malignant” all have a high MIB1 indexes consistent with WHO Grade II meningiomas. Both patients had tumor recurrence reflecting the aggressiveness of the tumor in their disease course. CONCLUSION In accordance with prior reports, our case series provides support that tumor DNA methylation profiling can provide useful classification information and can be correlated to both tumor pathology and patient clinical presentation.