Multi-Omic Bicluster Association Analysis (MOBAA)—a tool for identifying population subgroups with distinct multi-omics molecular profiles

Motivation: The increasing availability of multi-omic datasets from the same individuals presents the opportunity to uncover distinct molecular profiles across subgroups within a study population. These profiles may be linked to specific biological traits, such as disease status, and have distinct disease trajectories. Importantly, they could reveal robust, multi-layered molecular signatures with potential applications in early diagnosis, prognosis, and treatment advancing precision medicine. Although several integrative multi-omic methods have been developed in recent years, most are tailored to population-level analyses and are not well-suited for identifying signals specific to subpopulations. Results: We developed MOBAA (Multi-Omic Bicluster Association Analysis), a novel data-driven integrative machine-learning framework for identifying subgroups within a study population that exhibit distinct multi-omic molecular profiles. MOBAA is scalable and capable of handling multiple omics simultaneously without relying on parametric distributional assumptions. It combines biclustering algorithms with hierarchical clustering-based module identification and uses permutation-derived empirical P-values. This approach provides a comprehensive and intuitive view of underlying biological variation and population heterogeneity facilitating discovery of complex, multi-layered molecular signatures. Availability and implementation: The code is available as MOBAA R package. All source code as well as comprehensive documentation and examples are provided at https://github.com/pmishra912/MOBAA.