Multi-omics data integration for pediatric acute leukemia

Sammanfattning

B-cell precursor acute lymphoblastic leukemia (BCP-ALL) is the most common pediatric cancer, characterized by diverse genetic alterations, reflecting on patient outcomes. Despite its severity, BCP-ALL cure rates often exceed 90%. However, patients with relapsed ALL have poorprognosis. In addition, patients often suffer from adverse effects or die, if they are over-treated. In this project, multi-omics integration was applied to construct a functional map of BCP-ALL, incorporating data from 1,231 Nordic pediatric patients to elucidate the molecular factorsunderlying differential patient response. DNA methylation (DNAm), gene expression (GEX), single nucleotide variations (SNV), and ex vivo drug response (EVDR) -data were integratedusing Multi-Omics Factor Analysis (MOFA). MOFA was employed to identify major axes ofvariation both within and across these data layers, capturing key patterns linked to diseasebiology. The integration identified 10 cross-modal elements (CMEs), capturing up to 50% of variation inthe omics datasets. Several of the CMEs exhibited a significant correlation to specific molecularsubtypes of BCP-ALL, reflecting key biological characteristics of the disease. Gene SetEnrichment Analysis (GSEA) revealed associations between the CMEs and biological pathwaysrelevant to carcinogenesis, cell cycle, cell metabolism, as well as processes specifically linked to ALL development, such as immune regulation and lymphocyte function. Furthermore, survivalanalysis, combined with a multivariate Cox regression, demonstrated that CMEs could serve asputative prognostic markers of clinical outcomes, particularly event-free survival (EFS). Overall, data integration detected intricate disease patterns with potential clinical revelations, paving theway to a more holistic exploration of the landscape of BCP-ALL. Finally, an interactive resource for Visualizing ALL through Multi-Omics integration (vALLmo) was developed to enable user-friendly visual exploration of the CMEs and associated data.