Mapping Molecular Transitions in Barrett's-Associated Oesophageal Adenocarcinoma via Multi-Omics Integration and...

Background/Objectives: Interpretation of high-dimensional molecular profiling data requires computational strategies that go beyond single-layer analyses to resolve coordinated biological programs. Although multi-omics integration has advanced latent structure discovery, translating these cross-omics signals into interpretable, functionally meaningful molecular states remains a central challenge. Methods: Here, we present an interpretable, pathway-centric multi-omics integration framework that combines expression, DNA methylation, and copy-number alteration data to capture nonlinear cross-omics interactions and enable biologically grounded representations of molecular states. We apply this framework towards oesophageal adenocarcinoma, a malignancy that remains relatively underexplored in integrative multi-omics and systems-level analyses, despite its rising incidence and poor prognosis. Results: Using samples spanning Barrett's oesophagus and oesophageal adenocarcinoma from the Oesophageal Cancer Clinical and Molecular Stratification consortium, we demonstrate the framework's ability to resolve coordinated oncogenic, metabolic, immune, and cell-cycle programs that evolve across disease states. Conclusions: Together, this work establishes a scalable and interpretable computational strategy for pathway-based multi-omics integration, along with providing a generalisable approach for molecular state assignment in complex biological systems.