WAD Documentation

License

WAD is a linear programming method for L1-minimal reconstruction loss in wavelet-ATAC data deconvolution.

WAD_pipeline

Usage

Input

  • single-cell ATAC reference with defined cell types

  • bulk tissue samples for deconvolution

Output

  • deconvolution_results.tsv

Example Execution

WAD \
--scATAC sample1_celltypeA.bw --cell_type celltypeA \
--scATAC sample1_celltypeB.bw --cell_type celltypeB \
--scATAC sample1_celltypeC.bw --cell_type celltypeC \
--scATAC sample2_celltypeA.bw --cell_type celltypeA \
--scATAC sample2_celltypeB.bw --cell_type celltypeB \
--scATAC sample2_celltypeC.bw --cell_type celltypeC \
--bulk_tissue bulk_sample1.bw \
--bulk_tissue bulk_sample2.bw \
--chrom_sizes hg38.chrom.sizes \
--output_dir deconvolution_results

Example Output

BulkSample      CellTypeA       CellTypeB       CellTypeC
bulk_sample1       40%             50%             10%
bulk_sample2       35%             50%             15%

Installation

PyPI / Pip

WAD can be installed from PyPI using pip:

pip install WAD

We recommend installing using a conda environment:

conda create -n WAD_env
conda activate WAD_env
conda install pip
pip install WAD

Requirements

The list of package version requirements is available in setup.py.

- python == 3.12.7
- numpy >2.0, <2.3
- pandas == 2.3.1
- pyBigWig == 0.3.24
- PyWavelets == 1.7.0
- click == 8.1.7
- scipy == 1.13.1
- cvxpy == 1.6.6
- ortools >=9.10, <9.12
- pyarrow == 20.0.0

License

WAD is released under an MIT license.

API Reference