The R-based implementation of RA3

The R-based implementation of "A reference-guided approach for epigenetic characterization of single cells".
The source code for the reproduction of results in the manuscript can be found here.

Installation guide

Install the released version of RA3 package using devtools from Github:

install.packages("devtools")
devtools::install_github("cuhklinlab/RA3")

Package Seurat is needed for current version of RA3. Install Seurat using the following chunk, or check here.

# Enter commands in R (or R studio, if installed)
install.packages('Seurat')

Troubleshooting dependencies

At this point, there might be some missing dependencies from CRAN or Bioconductor.

Installing following packages from Bioconductor requires R version > 4.0.0 and Bioconductor version 3.11. Use following chunk:

if (!requireNamespace("BiocManager", quietly = TRUE))
    install.packages("BiocManager")
BiocManager::install(c('chromVAR', 'motifmatchr', 'SummarizedExperiment', 'BSgenome.Hsapiens.UCSC.hg19', 'slingshot', 'JASPAR2016'))

It will take a few minutes to install the RA3 package, mainly for preparing the embedded demo-data.

Functions

This package includes following main functions:

runRA3 runs RA3 for the analysis of single-cell chromatin accessibility sequencing data.
Dataprep normalizes the input count matrix by TF-IDF.
RA3_EM performs an EM algorithm to estimate parameters of the RA3 model.
RA3_clustering deploys louvain clustering on the output of RA3.
RA3_TrajInf does trajectory inference based on the output of RA3.
RA3_motif runs motif enrichment based on the output of RA3.
RA3_pcatsne deploys PCA and t-SNE for dimension reduction on normalized scCAS data.
RA3_RefProj deploys a reference projection approach on the TF-IDF normalized scCAS data, then uses t-SNE for further dimension reduction.

Documentation

Please check the vigenette for a tutorial. Two examples are contained for a quick start of RA3.

System requirements

Software requirements

The package has been tested on the following operating systems:

Linux: CentOS Linux release 7.7.1908
macOS: Catalina (10.15.4)
Windows: Windows 10

Hardware requirements

The package has been tested on both Normal Personal Computer and High-Performance Computing Cluster.

License

This package is built under license GNU GENERAL PUBLIC LICENSE (GPL).