arcinstitute/ST-Parse

State Transition model checkpoints are meant to be used with the Arc State package.

State is distributed via uv. To train models on your own AnnDatas, please see the repository README.

Running inference from pypi

uv tool install arc-state

To generate predictions:

state tx infer --model_dir <ST-Parse_PATH> --pert_col cytokine --adata <INPUT_ADATA>.h5ad --output <OUTPUT_DIR>

This will group the cells in your input AnnData by the specified pert_col and run inference with a pretrained ST checkpoint.

Running from source

# Clone repo
git clone github.com:arcinstitute/state
cd state

# Initialize venv
uv venv

# Install
uv tool install -e .

To generate predictions:

state tx infer --model_dir <ST-Parse_PATH> --pert_col cytokine --adata <INPUT_ADATA>.h5ad --output <OUTPUT_DIR>

For model licenses please see MODEL_ACCEPTABLE_USE_POLICY.md, MODEL_LICENSE.md, and LICENSE.md.