Multiscale multimodal integration of spatial omics through a deep generative model with CORAL.

CORAL, a probabilistic deep generative model, integrates spatial transcriptomic, proteomic, metabolomic, and epigenomic data across scales to generate unified representations with broad molecular coverage and high spatial resolution.

• Why it matters: Spatial omics integration is hindered by mismatched spatial resolution, cross-modal alignment, and incomplete molecular coverage across modalities and platforms.
• What they did: CORAL uses graph attention to jointly model high-resolution datasets with limited molecular coverage alongside low-resolution genome-wide datasets, then evaluates domain detection, high-resolution molecular prediction, spatially variable genes, and cell-cell interaction inference on synthetic and experimental data.
• The result: In hepatocellular carcinoma patient spatial transcriptomic and proteomic data before and after immunotherapy, CORAL identified organized immune-cell subsets and signaling pathways linked to therapeutic resistance.

AI summary of the abstract · may contain errors · not medical advice · read the paper.