Installation

Software Dependencies

The following core dependencies are required:

scanpy
numpy
pandas
anndata
matplotlib
scipy
scikit-learn
psutil
cloudpickle
tqdm
leidenalg
networkx
docrep
jax
ott-jax[neural]
torch
torch-geometric
wrapt
rich
plotly
pyyaml
torch-geometric
torch-cluster
torch-scatter
torch-sparse

Installation Guide

Using Conda and Pip

We assume that you already have Python installed (preferably within a Conda virtual environment).

If you have CUDA-compatible devices, we recommend installing PyTorch and JAX with CUDA support beforehand.

To install PyTorch with CUDA support, please follow the official guide: https://pytorch.org/get-started/locally
To install JAX (and JAXLIB) with CUDA support, refer to: https://docs.jax.dev/en/latest/installation.html

Additionally, AlignDG requires several PyTorch Geometric (PyG) dependencies. It is strongly recommended to install these packages manually to avoid potential build issues (e.g., hanging at the setup step when installing torch_cluster via pip — a known issue).

Install the compatible PyG packages with the following command:

pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-${TORCH}+${CUDA}.html

Replace ${TORCH} and ${CUDA} with your specific PyTorch and CUDA versions, respectively.

For example, if you are using PyTorch 2.8.0 with CUDA 12.9, run:

pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.8.0+cu129.html

Once all dependencies are installed, you can install AlignDG via pip:

pip install aligndg

To verify the installation:

import aligndg
print(aligndg.__version__)

—

From Source (GitHub)

You can also install AlignDG from source.

Clone the repository from GitHub:

git clone https://github.com/xkmaxidian/AlignDG.git

Navigate to the source directory:
```
cd your_path/AlignDG_package
```

Build and install the package:

python setup.py build
python setup.py install

—

Notes - We strongly recommend using Python ≥ 3.10 and pip ≥ 22.0. - To avoid dependency conflicts, always install AlignDG in a clean virtual environment (e.g., via Conda or venv). - For GPU users, ensure that your CUDA toolkit and drivers are correctly configured before installing PyTorch/JAX.