Abstract B004: Whole genome nanopore DNA methylation sequencing for rapid molecular classification of CNS tumors

Abstract While central nervous system (CNS) tumors are the leading cause of cancer-related deaths in the United States pediatric population, 5-year survival rates vary drastically depending on the tumor subtype. To optimize outcomes for patients, the tumor type and subtype must be identified quickly and accurately, with development of risk-stratified treatment plans often relying on a combination of radiological, histopathological, and molecular diagnoses. Since subtypes of CNS tumors have distinct DNA methylation (DNAm) patterns, the clinical standard molecular diagnostic platform has historically been DNAm microarray. But in recent years, one emerging technology - nanopore sequencing - has grossly decreased the turnaround time and complexity of DNAm-based subtyping, all while providing a more comprehensive picture of the tumor. Recent studies have shown promising results for implementation of nanopore DNAm sequencing in intraoperative settings, where molecular diagnoses within a surgical timeframe (60-90 minutes) are able to inform the degree of tumor resection. However, these studies rely on distributing precious material between histopathology and molecular pathology, together limiting availability for later analyses. To overcome this limitation, we sought to demonstrate the molecular diagnostic utility of tumor aspirate, a waste material generated from ultrasonic aspiration during surgical resection. We performed whole genome nanopore DNAm sequencing on 8 matched pairs of standard patient tissue and tumor aspirate (16 total samples) and utilized the previously-developed machine learning classifier Sturgeon to compare tumor (sub)types between samples. For each pair, we demonstrated moderate to high depth genomic coverage, generated digital copy number variation plots for identifying large-scale alterations, classified the sample's tumor (sub)type, and validated key subtype-defining features. Copy number variation profiles, DNAm-based tumor (sub)type classifications, and subtype-defining features were all consistent between matched samples, and detailed analyses of 5mC and 5hmC DNAm modifications revealed strong correlations amongst tumor subtypes regardless of sample origin, together demonstrating the utility of tumor aspirate for molecular analyses. Additionally, as current machine learning classifiers for CNS tumors have been trained on DNAm microarray data from both adult and pediatric patients and are well-known to underperform in pediatric populations, the matched pairs were also used to identify shortcomings in existing classification, including confident misclassification of rare tumor entities, inappropriate labels for some common pediatric tumor (sub)types, and insufficient computational adjustment for low tumor purity samples. While adjustments to DNAm preprocessing led to correct, confident classification of all 16 samples, the current project objectives involve utilizing lower depth data, such as that collected during intraoperative nanopore sequencing, to achieve similar accuracy and confidence, particularly in pediatric contexts. Citation Format: Allison A. Murray, Breanna E. Mann, Andrew B. Satterlee, David E. Kram, Jeremy R. Wang. Whole genome nanopore DNA methylation sequencing for rapid molecular classification of CNS tumors [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Brain Cancer; 2026 Mar 23-25; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2026;86(6_Suppl):Abstract nr B004.