Abstract / Summary
Up to one-third of Wilms tumours (WTs) arise in the context of genetic or epigenetic predisposition, but identifying predisposed patients remains challenging because germline variants can be missed due to variable clinical presentation, mosaicism, and cryptic variants. We tested whether tumour-derived molecular patterns can reveal underlying predisposition and thereby complement conventional diagnostic approaches. DNA methylation profiling of 138 WTs identified five molecular clusters that captured biological and predisposition-associated heterogeneity, including enrichment for Beckwith-Wiedemann Spectrum tumours, and a distinct WT1-driven subgroup. Cluster-guided reanalysis uncovered previously unrecognised WT1 pathogenic variants, including a newly identified germline variant that led to surveillance enrolment. A random forest classifier predicted WT1-driven tumours with ~90% accuracy across independent validation cohorts. Tumour methylation profiling identifies biologically and clinically relevant WT subgroups and reliably detects a WT1-driven tumour methylation signature. This approach can guide targeted variant reanalysis, and uncover clinically relevant germline variants missed by standard diagnostic workflows.