Installation#

Compatibility#

  • Version: 1.0.0 (Release Date: September 2026)

  • Tested Python versions: 3.12–3.14

  • Tested platforms: Linux, Windows, and macOS on Apple Silicon

Linux and Windows#

Install from PyPI:

python -m pip install bcmp

macOS (Apple Silicon)#

Use the Conda-forge environment. Download bcmp.yml and run:

conda env create -f bcmp.yml
conda activate bcmp
python -m pip install --no-deps bcmp

Verify the installation#

import bcmp
from bcmp.datasets import diabetic_kidney_lite

adata = diabetic_kidney_lite()
print(adata.shape)

Expected output:

(1500, 27980)

Install from source#

For the development version:

git clone https://github.com/Guan-Pujun/bcmp-python.git
cd bcmp-python
python -m pip install .

scIB metric calculation#

Batch-effect-removal metrics for candidate integration embeddings are calculated with scib-metrics. Follow the official scib-metrics installation guide; for the PyPI release, run:

python -m pip install scib-metrics

Dependencies#

Installing bcmp resolves and installs the following core packages:

  • anndata

  • numpy

  • pandas

  • scipy

  • numba

  • annoy

  • igraph and leidenalg

  • scikit-learn and scikit-misc

  • umap-learn

For typical use, we recommend:

  • scib-metrics