Robust molecular subgrouping and reference-free aneuploidy detection in medulloblastoma using low-depth whole genome bisulfite sequencing.

Medulloblastoma comprises four principal molecular disease groups and their component subgroups, each with distinct molecular and clinical features. Group assignment is currently achieved diagnostically using Illumina DNA methylation microarray. Whole-genome sequencing (WGS) capacity is rapidly expanding in the clinical setting and the development of platform-independent, sequence-based assays of molecular group offers significant potential. Specifically, whole-genome bisulfite sequencing (WGBS) enables assessment of genome-wide methylation status at single-base resolution, however its routine application has been limited by high DNA input requirements, cost, and a lack of pipelines tailored to more rapidly-acquired and cost-effective low-depth (< 10x) sequencing data. We utilised WGBS data for 69 medulloblastomas, comprising 35 in-house low-depth (~ 10x) and 34 publicly available high-depth (~ 30x) samples, alongside cerebellar controls (n = 8), all with matched DNA methylation microarray data. We assessed quality (QC) and imputation approaches using low-pass WGBS data, assessed inter-platform correlation and identified molecular groups and subgroups by directly integrating matched/associated loci from WGBS sample data with the MNP classifier probeset. We further assessed and optimised reference-free aneuploidy detection using low-pass WGBS and assessed concordance with microarray-derived calls. We developed and optimised pipelines for processing, QC, and analysis of low-pass WGBS data, suitable for routine molecular subgrouping and reference-free aneuploidy assessment. We demonstrate that low-pass WGBS data can (i) be integrated into existing array-trained models with high assignment probabilities for both principal molecular groups (97% concordance) and molecular subgroups (94.2% concordance), and (ii) detect clinically relevant focal copy number changes, including SNCAIP, with greater sensitivity than microarray approaches. Low-pass WGBS performs equivalently to array-based methods at comparable cost. Finally, its ascertainment of the full methylome enables elucidation of additional biological complexity and inter-tumoural heterogeneity that has hitherto been inaccessible. These findings provide proof-of-concept for clinical adoption of low-pass WGBS, applied using standard WGS technology.