scarp 1.0.0

Incorporating network diffusion and peak location information for better single-cell ATAC-seq data analysis

---

SCARP method for scATAC-seq data analysis

Incorporating network diffusion and peak location information for better single-cell ATAC-seq data analysis

Authors

• [e-mail]
• [e-mail]

Pipeline

Installation

We recommend to create a new environment with Python 3.10:

conda create -n py310 python=3.10
conda activate py310

The scarp package can be installed via pip:

pip install scarp

Dependencies

scanpy (>=1.9.5)
numpy (>=1.25.2)
scipy (>=1.11.3)
pandas (>=2.1.1)

Usage

Please checkout the tutorials at here.

1. Preparing your scATAC-seq data in h5ad format

You can downloaded a example data from here.

import scanpy as sc

data_name = 'Leukemia'
data = sc.read_h5ad('./Example_data/Leukemia.h5ad')

2. Running SCARP easily in one step

from scarp import model

Cells_df = model.SCARP(adata=data,
                       data_name=data_name,
                       plot_SD=True,
                       verbose=True
                       )

parameter descriptions:

parameter name	description	type	default
adata	input scATAC-seq data	h5ad	None
data_name	name of this dataset	str	None
m	parameter to control NR diffusion intensity	float	1.5
gamma	parameter to control the threshold for merging adjacent chromosomes	int	3000
beta	parameter to control the extent to which prior edge weight decays	int	5000
return_shape	shape of the returned matrix	str	'CN'
peak_loc	use peak location prior information or not	bool	True
parallel	parallel computing or not. 0 means automatically determined	int	0
plot_SD	plot the SDs of PCs or not	bool	True
fig_size	figure size of the SD plot	tuple	(4,3)
save_file	if plot_std is True, the file path you want to save	str	None
verbose	print the process or not	bool	True

3. Or you can run SCARP step by step

(1) Obtaining the NR diffused matrix

t, diffusion_mat = model.SCARP_diffusion_mat(adata=data)

(2) Computing the retained dimension

k = model.SCARP_SD_plot(data=diffusion_mat,
                        peaks_num=Peaks_num,
                        title=data_name,
                        plot_SD=True)

(3) Dimensional reduction

Cell_embedding = model.SCARP_cell_embedding(diffusion_mat=diffusion_mat,
                                            kept_comp=k)

Reproduce results

For reproducibility, we provide all the necessary scripts and data here.