Network Biology: Exploring Methylation Features in Cancer through Machine Learning

Despite the growing understanding of DNA methylation’s role in cancer biology, many existing studies focus on analyzing individual methylation sites or predefined gene sets, leaving the interactions between co-methylated regions largely unexplored. In this study, we apply a network-based approach to explore methylation features across six cancer types using the OhmNet model. By constructing co-methylation networks and utilizing multi-layer network embedding, we identify significant co-methylation patterns associated with key genes implicated in tumorigenesis. Through pathway enrichment analysis, we discovered key pathways related to cell adhesion and axonogenesis, suggesting a novel link between DNA methylation and nerve-cancer crosstalk. Our work not only reveals unique insights into the methylation landscape of cancers but also introduces a scalable, label-free network-based approach for studying complex epigenetic regulation.