Single-Cell ATAC-seq analysis via Latent feature Extraction
2021.04 A new online integration tool SCALEX on scRNA-seq and scATAC-seq is available!
2021.03.23 Introduce the highly_variable_genes from scanpy to filter peaks
2021.01.14 Update to compatible with h5ad file and scanpy
SCALE neural network is implemented in Pytorch framework.
Running SCALE on CUDA is recommended if available.
install from PyPI
pip install scale
install latest develop version from GitHub
pip install git+https://github.com/jsxlei/SCALE.git
or download and install
git clone git://github.com/jsxlei/SCALE.git cd SCALE python setup.py install
Installation only requires a few minutes.
- h5ad file
- count matrix file:
- row is peak and column is barcode, in txt / tsv (sep="\t") or csv (sep=",") format
- mtx folder contains three files:
- count file: count in mtx format, filename contains key word "count" / "matrix"
- peak file: 1-column of peaks chr_start_end, filename contains key word "peak"
- barcode file: 1-column of barcodes, filename contains key word "barcode"
- h5mu file, e.g. filename.h5mu/atac
SCALE.py -d [input]
Output will be saved in the output folder including:
- model.pt: saved model to reproduce results cooperated with option --pretrain
- adata.h5ad: saved data including Leiden cluster assignment, latent feature matrix and UMAP results.
- umap.pdf: visualization of 2d UMAP embeddings of each cell
Get binary imputed data in adata.h5ad file using scanpy adata.obsm['binary'] with option --binary (recommended for saving storage)
SCALE.py -d [input] --binary
or get numerical imputed data in adata.h5ad file using scanpy adata.obsm['imputed'] with option --impute
SCALE.py -d [input] --impute
- save results in a specific folder: [-o] or [--outdir]
- embed feature by tSNE or UMAP: [--embed] tSNE/UMAP
- filter low quality cells by valid peaks number, default 100: [--min_peaks]
- filter low quality peaks by valid cells number, default 3: [--min_cells]
- filter peaks by selecting highly variable features, default 100,000: [--n_feature], disable by [--n_feature] -1.
- modify the initial learning rate, default is 0.002: [--lr]
- change iterations by watching the convergence of loss, default is 30000: [-i] or [--max_iter]
- change random seed for parameter initialization, default is 18: [--seed]
- binarize the imputation values: [--binary]
Look for more usage of SCALE
Use functions in SCALE packages.
import scale from scale import * from scale.plot import * from scale.utils import *
Tutorial Forebrain Run SCALE on dense matrix Forebrain dataset (k=8, 2088 cells)
Lei Xiong, Kui Xu, Kang Tian, Yanqiu Shao, Lei Tang, Ge Gao, Michael Zhang, Tao Jiang & Qiangfeng Cliff Zhang. SCALE method for single-cell ATAC-seq analysis via latent feature extraction. Nature Communications, (2019).