scviva-legacy (formerly niche-VI)

Archived. This is the original research implementation of scVIVA, kept to reproduce the results of the scVIVA paper. scVIVA is now maintained in scVIVA-Tools (GitHub), installed with pip install scviva-tools and used as scviva.SCVIVA. Use that for new work. This repository receives no further development.

The code that produces the paper's figures is in scviva_paper. It imports this package as nichevi.

Which version to install

All paper results were produced with commit 34a85af ("iLISI computation"). Install that exact commit:

pip install git+https://github.com/YosefLab/scviva-legacy.git@34a85af59a441d69ed7aeb90d4dda98f3dcbf047

The repository was renamed from niche-VI. GitHub redirects the old URL, so git+https://github.com/YosefLab/niche-VI.git@... still works. The Python import name is still nichevi, and the distribution name is still niche-VI. Neither was renamed, because the trained checkpoints are loaded through this package and its classes.

The paper environment also pinned scvi-tools to development commit 3e275ff (installed from git) and scib-metrics to the 0.5.7 release. The full environment is envs/scvi.yml in the reproducibility repo.

Python 3.12 was used. The package declares >=3.9.

Citation

@article{levy2025scviva,
  title={scVIVA: a probabilistic framework for representation of cells and their environments in spatial transcriptomics},
  author={Levy, Nathan and Ingelfinger, Florian and Bakulin, Artemii and Cinnirella, Giacomo and Boyeau, Pierre and Nadler, Boaz and Ergen, Can and Yosef, Nir},
  journal={bioRxiv},
  pages={2025--06},
  year={2025},
  publisher={Cold Spring Harbor Laboratory}
}