The multilayer community structure of medulloblastoma

Abstract

Biomedical multilayer networks offer a wide range of possibilities for the interpretation of the molecular basis of diseases; a particularly challenging task in the case of rare diseases, where the number of cases is small in comparison with the size of the associated multi-omics datasets. In this work, we develop a dimensionality reduction methodology to identify the minimal set of genes that characterize disease subgroups based on their persistent association in the multilayer network at different levels of resolution. We apply this approach to the study of a cohort of patients affected by medulloblastoma, a childhood brain tumor, using proteogenomic data. Our approach is able to recapitulate known medulloblastoma subtypes (accuracy > 94%) and offers a clear characterization of the associated gene functions, with the downstream implications for diagnosis and therapeutic interventions. We verified the general applicability of our method by applying it to an independent dataset, achieving very high performances (accuracy > 98%). Overall, this approach opens the door to a new generation of multilayer-based methods able to overcome the specific dimensionality limitations of the rare disease datasets.