Towards accessible molecular diagnostics for central nervous system tumours

The 2021 WHO classification represents a significant shift in Central Nervous System (CNS) tumour diagnostics, emphasising the integration of molecular alterations alongside traditional histopathology. Among the advancements in molecular diagnostics, methylation-based classification using the Heidelberg Molecular Neuropathology (MNP) classifier (molecularneuropathology.org) has become an essential diagnostic tool. Conventional molecular testing often involves multiple assays such as DNA/RNA sequencing, methylation arrays, immunohistochemistry among others, which are resource-intensive and limited to high-throughput settings due to their complexity, costs, and lengthy turnaround times. In this work, I introduce two tools aimed at improving the accessibility and affordability of CNS tumour molecular diagnostics: Rapid-CNS2 and MNP-Flex. Rapid-CNS2 is a nanopore sequencing workflow that employs adaptive sampling to efficiently detect mutations, copy number alterations, gene fusions, target gene methylation, and perform methylation classification, all in a single test. This system is flexible, allowing immediate testing on individual samples and customisable targets via a simple text file. I formulated and subsequently validated the pipeline using 252 samples, including archival and diagnostic frozen sections. I developed ad-hoc models for methylation classification and MGMT promoter methylation detection. I employed publicly available state-of-the-art tools for pre-processing, variant calling and annotation, and devised computational acceleration strategies. Additionally, I demonstrate the potential of the pipeline to report results in an intraoperative time-frame with 18 samples from two independent centres. Thus, Rapid-CNS2 offers real-time methylation classification and DNA copy-number reporting within a 30-minute intraoperative window, followed by comprehensive molecular profiling within 24h, covering the entire spectrum of molecular alterations relevant for diagnosis and targeted therapies for CNS tumour subtypes- drastically reducing the weeks-long turnaround required by conventional methods. To further enhance accessibility of the MNP classifier, I developed MNP-Flex, a platform-independent version of the MNP classifier, covering 184 CNS tumour classes. I validated MNP-Flex on a global cohort of over 78,000 samples, including both frozen and formalin-fixed paraffin-embedded (FFPE) samples processed using five different methylation profiling technologies. With clinically relevant thresholds, MNP-Flex achieved accuracies of 99.6% for methylation families and 99.2% for methylation classes. Together, Rapid-CNS2 and MNP-Flex offer a comprehensive workflow for CNS tumour diagnostics. Rapid-CNS2 provides real-time, intraoperative reporting of broad methylation classification and copy number variations to guide surgical strategy, while the complete molecular profile and fine-grained methylation classification with MNP-Flex is available the next day, informing clinical care and therapeutic decisions. The workflow is cost-effective, uses compact equipment, and employs straightforward laboratory and bioinformatics tools. Rapid-CNS2 is available on GitHub, and MNP-Flex can be accessed via a research-use web service at https://mnp-flex.org. This integrated approach aims to streamline CNS tumour molecular diagnostics, broadening global access to precise, molecularly-informed classification and ultimately improving patient outcomes.