Metadata-Version: 2.4
Name: arielspatial
Version: 0.1.0a14
Summary: Spatial multiomics integration.
Author: Tien-Thanh Bui
License-Expression: MIT
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: anndata
Requires-Dist: anndata2ri
Requires-Dist: mudata
Requires-Dist: muon
Requires-Dist: numpy
Requires-Dist: pandas
Requires-Dist: rpy2
Requires-Dist: scanpy
Requires-Dist: scipy
Requires-Dist: torch
Requires-Dist: tqdm
Provides-Extra: gpu
Requires-Dist: cupy-cuda13x; extra == "gpu"
Requires-Dist: rapids-singlecell-cu13; extra == "gpu"
Dynamic: license-file

# ARIELSpatial

ARIEL is a library for spatial multiomic integration with modular encoders,
graph backbones, multimodal aggregators, probabilistic decoders, loss
composition, clustering, profiling, and reproducible seeding.

## What this release contains

Version 0.1.0 ships the data-preparation and seeding layers:

- `ARIELSpatial.preprocessing`: RNA preprocessing with Seurat SCTransform and
  PCA, ADT CLR normalization and PCA, in-tissue spot filtering with spatial
  coordinates, and barcode and sparse-matrix utilities.
- `ARIELSpatial.seed`: process-wide seeding and namespaced Python, NumPy, and
  PyTorch random generators, including DataLoader worker seeding.

## Installation

Releases are currently alpha pre-releases, so pip needs `--pre`:

```bash
pip install --pre arielspatial
```

The package is imported as `ARIELSpatial`:

```python
import ARIELSpatial
```

### Requirements

- Python 3.10 or newer on Linux.
- An R installation with the Seurat, SeuratObject, sctransform, Matrix,
  dplyr, and SingleCellExperiment packages. RNA preprocessing runs Seurat
  SCTransform through `rpy2` and `anndata2ri`.
- Optional GPU acceleration: an NVIDIA GPU with a CUDA 13 runtime and the
  `gpu` extra, which installs `cupy-cuda13x` and `rapids-singlecell-cu13`:

  ```bash
  pip install --pre "arielspatial[gpu]"
  ```

## License

MIT. See [LICENSE](LICENSE).
