Metadata-Version: 2.4
Name: scDecant
Version: 0.1.3
Summary: DECANT prioritizes biological heterogeneity over batch effects during feature selection for single-cell multi-omics data.
Author: Ziyi Zou, Heyang Hua, Siyu Li, Yuhang Jia, Ning Sun, Shengquan Chen
License-Expression: MIT
Project-URL: Repository, https://github.com/BioX-NKU/DECANT
Keywords: single-cell,feature selection,batch effect,data integration,single-cell omics,scRNA-seq,scDNAm,scATAC-seq
Classifier: Programming Language :: Python :: 3
Classifier: Operating System :: OS Independent
Classifier: Intended Audience :: Science/Research
Classifier: Topic :: Scientific/Engineering :: Bio-Informatics
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: numpy>=1.23
Requires-Dist: pandas>=1.5
Requires-Dist: scipy>=1.9
Requires-Dist: scikit-learn>=1.1
Requires-Dist: scanpy>=1.9
Requires-Dist: anndata>=0.8
Requires-Dist: h5py>=3.0
Requires-Dist: igraph>=0.10
Requires-Dist: leidenalg>=0.9
Requires-Dist: scib>=1.1
Dynamic: license-file

# scDecant

Batch effects increasingly confound biological structure in single-cell atlases. Such effects are usually corrected during downstream integration, but are rarely weighed against biological signal during feature selection. DECANT starts batch control at feature selection by weighing biological discrimination against batch association for each gene, genomic region or accessibility peak, thereby providing downstream methods with a less batch-dominated and biologically informative feature set. Across nine datasets spanning scRNA-seq, scDNAm and scATAC-seq, each analysed with four modality-matched downstream workflows, DECANT outperformed the corresponding baselines in the vast majority of comparisons, improving the balance between biological conservation and batch mixing across all three modalities. Beyond these benchmarks, the selected features enabled regulatory pathway analysis at cell-type resolution, uncovered additional candidate differentially methylated regions and showed sharper brain-tissue specificity in cortical methylation data.

## Installation

```bash
pip install scDecant
```

Python 3.10 or later is required.

## Quick start

```python
import scanpy as sc
from decant import run_decant

adata = sc.read_h5ad("path/to/pancreas.h5ad")
result = run_decant(adata, omics="rna", batch_key="tech", n_features=2000)
selected = adata[:, result["selected_indices"]]
```

Set `omics` to `"rna"`, `"meth"`, or `"atac"`. The selected feature set can be
passed to a downstream dimensionality-reduction, integration, or clustering
method.

## Source code and tutorials

Source code, per-omics tutorial notebooks, and documentation are available at
https://github.com/BioX-NKU/DECANT.
