Interpretable spatial multi-omics data integration and dimensionality reduction with SpaMV

Spatial multi-omics technologies revolutionized our understanding of biological systems by providing spatially resolved molecular profiles from multiple perspectives. Existing spatial multi-omics integration methods often assume that data from different omics share a common underlying distribution and aim to project them into a single unified latent space. This assumption, however, obscures the unique insights offered by each omics, thereby limiting the full potential of multi-omics analyses. To address this limitation, we develop the Spatial Multi-View (SpaMV) representation learning algorithm, which explicitly captures both shared information across omics and the distinct, omics-specific information, enabling a more comprehensive and interpretable representation of spatial multi-omics data. Through extensive evaluation on both simulated and real-world datasets, SpaMV demonstrates superior spatial domain clustering performance and offers topic modeling with more interpretable dimensionality reduction for downstream analysis. Moreover, our method more effectively discovers interpretable omics-specific biomarkers than existing approaches, highlighting its strength in disentangling multi-omics signals. Spatial multi-omics tools can reveal complementary molecular information across tissues, but integrating shared and omics-specific signals remains challenging. Here, the authors develop SpaMV, an interpretable representation learning framework that disentangles shared and omics-specific information to improve spatial multi-omics integration, spatial domain analysis, and biologically meaningful topic discovery.