Xenium single cell miso

This repo was forked from the original miso repository and modified to accomodate large xenium datasets with batching and cagra knn to calculate adjacency matrices, with several other changes. This version does not use the HIPT model for H&E images and instead takes as input pre-calculated foundation model embeddings that have already been associated to cells (see here for an example implementation).

Setup:

mamba create -n miso_rapids python=3.11.11
mamba activate miso_rapids
mamba install -c rapidsai -c conda-forge -c nvidia cuvs cuda-version=12.8
mamba install -c conda-forge ipykernel ipywidgets
git clone https://github.com/Tyler-Lam/miso.git
cd miso
python -m pip install .

Software Requirements

python==3.11.11
einops==0.8.1
numpy==2.2.6
opencv-python==4.10.0
Pillow==11.2.1
scanpy==1.11.2
scikit-image==0.25.2
scikit-learn==1.7.0
scipy==1.15.2
setuptools==75.8.2
torch==2.4.0
torchvision==0.19.0
tqdm==4.67.1