The CORAL package, a deep generative model for multiscale multimodal integration of spatial omics, is available at https://github.com/zou-group/CORAL and archived on Zenodo (https://doi.org/10.5281/zenodo.22263988). All custom code generated for the study, including data preprocessing, model training, benchmarking and ablation analyses, simulation code and notebooks that generate every main and Extended Data figure panel, is available at https://github.com/siyuh/CORAL_reproducibility and https://doi.org/10.5281/zenodo.22263988. Analyses were performed in Python (v3.12), with CORAL model training, downstream analyses and figure generation using PyTorch (v2.9.1), PyTorch Geometric (v2.7.0), Scanpy (v1.11.5), AnnData (v0.12.6), NumPy (v2.3.5), pandas (v2.3.3), SciPy (v1.16.3), scikit-learn (v1.7.2), leidenalg (v0.11.0), Python-igraph (v1.0.0), umap-learn (v0.5.9), NetworkX (v3.6), matplotlib (v3.10.7), seaborn (v0.13.2), esda (v2.10.0)