---
license: mit
pipeline_tag: mask-generation
tags:
- model_hub_mixin
- pytorch_model_hub_mixin
---

<!-- This model has been pushed to the Hub using the [PytorchModelHubMixin](https://huggingface.co/docs/huggingface_hub/package_reference/mixins#huggingface_hub.PyTorchModelHubMixin) integration:
- Library: ynyou/CIFM
- Docs: [More Information Needed] -->

# CIFM: Cellular Interaction Foundation Model

## Overview
This is the PyTorch implementation of the CIFM model -- an AI foundation model that can simulate the activities within living tissues (AI virtual tissue).
The current version of CIFM has 100M parameters and is trained on around 23M cells of spatial genomics. The signature functions of CIFM are:
- **Embedding** of celllular microenvironments via ```embeddings = model.embed(adata)``` (the 1st Figure below panel D top);
- **Inference/simulation** of cellular gene expressions within a certain microenvironment via ```expressions = model.predict_cells_at_locations(adata, target_locs)``` (the 1st Figure below panel D bottom, and the 2nd Figure below).

The detailed usage of the model can be found in the [tutorial](https://huggingface.co/ynyou/CIFM/blob/main/test.ipynb).
Before running the tutorial, please set up an environment following the [environment instruction](https://huggingface.co/ynyou/CIFM#environment).

More information about the model can be found in the [preprint](https://www.biorxiv.org/content/10.1101/2025.01.25.634867v1).

![](./figures/cifm.png)
![](./figures/autoregressive.gif)

## Environment
I use conda to manage the environment ```conda create -n $MYENV python=3.11```, but it is not the only way to do that.
After entering the environment, you can clone the model via:
```
# Make sure you have git-lfs installed (https://git-lfs.com), e.g. as:
git lfs install
# or if you are under conda environment, you can also install git-lfs via conda as:
conda install anaconda::git-lfs

# clone the model and enter the directory
git clone https://huggingface.co/ynyou/CIFM
cd ./CIFM/
git lfs install ; git lfs pull
```

The environment can be set up via:
```
# my python version is 3.11, which is manage when creating conda environment: conda create -n $MYENV python=3.11
pip install torch==2.1.0 --index-url https://download.pytorch.org/whl/cu118
pip install torch-scatter torch-sparse torch-cluster torch-geometric -f https://data.pyg.org/whl/torch-2.1.0+cu118.html
pip install scanpy e3nn lightning==2.1.0 transformers numpy==1.26.4
```
where the compatibility between ```torch``` and ```torch-geometric``` is not always guaranteed (lines 2-3), since these two guys are very picky on the platform and the version of other packages. You may need trial-and-error to find the right versions of ```torch``` and ```torch-geometric``` that works for you, e.g., in some machine I once installed via:
```
pip install torch==2.0.1 --index-url https://download.pytorch.org/whl/cu117
pip install torch-scatter torch-sparse torch-cluster torch-geometric -f https://data.pyg.org/whl/torch-2.0.1+cu117.html
```

## Citation
If you use this code for you research, please cite our paper.
```
@article{you2025building,
  title={Building Foundation Models to Characterize Cellular Interactions via Geometric Self-Supervised Learning on Spatial Genomics},
  author={You, Yuning and Wang, Zitong and Fleisher, Kevin and Liu, Rex and Thomson, Matt},
  journal={bioRxiv},
  pages={2025--01},
  year={2025},
  publisher={Cold Spring Harbor Laboratory}
}
```