novae 1.1.1

💫 Graph-based foundation model for spatial transcriptomics data

Novae is a deep learning model for spatial domain assignments of spatial transcriptomics data (at both single-cell or spot resolution). It works across multiple gene panels, tissues, and technologies. Novae offers several additional features, including: (i) native batch-effect correction, (ii) analysis of spatially variable genes and pathways, and (iii) architecture analysis of tissue slides.

[!NOTE] Novae was developed by the authors of sopa and is part of the scverse ecosystem. Read our article here.

[!TIP] Recent updates:

1. We added a new improved model, check the corresponding tutorial here.
2. We also added a tutorial for automated labeling.

Documentation

Check Novae's documentation to get started. It contains installation explanations, API details, and tutorials.

Overview

(a) Novae was trained on a large dataset, and is shared on Hugging Face Hub. (b) Illustration of the main tasks and properties of Novae. (c) Illustration of the method behind Novae (self-supervision on graphs, adapted from SwAV).

Installation

novae can be installed from PyPI on all OS, for any Python version >=3.11.

pip install novae

[!NOTE] See this installation section for more details about extras and other installations modes.

Usage

Here is a minimal usage example. For more details, refer to the documentation.

import novae

# compute cell neighbors
novae.spatial_neighbors(adata)

# load a pre-trained model
model = novae.Novae.from_pretrained("prism-oncology/novae-human-0")

# compute spatial domains
model.compute_representations(adata, zero_shot=True)
model.assign_domains(adata)

Cite us

Our article is published in Nature Methods. You can cite Novae as below:

Blampey, Q., Benkirane, H., Bercovici, N. et al. Novae: a graph-based foundation model for spatial transcriptomics data.
Nat Methods (2025). https://doi.org/10.1038/s41592-025-02899-6