scPipe

scPipe is an R package that allows barcode demultiplexing, transcript mapping and quality control of raw sequencing data generated by

• multiple 3 prime end sequencing protocols for scRNA-Seq including CEL-seq, MARS-seq, Chromium 10x and Drop-seq and,
• various scATAC-Seq platforms including sci_ATAC, sc-ATAC, 10X, etc.

RNA-Seq module of scPipe produces a count matrix that is essential for downstream analysis along with a user-friendly HTML report that summarises data quality. These results can be used as input for downstream analyses including normalization, visualization and statistical testing.

ATAC-Seq module of scPipe contains capabilities of pre-processing scATAC-Seq data through scPipe. This functionality allows barcode demultiplexing, peak calling and quality control of raw sequencing data generated by multiple single-cell ATAC-Seq sequencing protocols including 10X, scATAC-Seq, dscATAC-Seq, dsciATAC-Seq, sciATAC-Seq, plate-based ATAC-Seq and scHTS-Seq. ATAC-Seq module also produces a feature-barcode count matrix that is essential for downstream analysis along with a user-friendly HTML report that summarises data quality.

The scATAC-Seq preprocessing module of the package is under active development. Feel free to ask any questions or submit a pull request.

• [01/04/2021] scPipe now uses samtools to remove duplicates from scATAC-Seq data
• [01/04/2021] scPipe now uses MACS3 for scATAC-Seq peak calling
• [01/04/2021] scPipe now uses sinto for scATAC-seq fragment file generation
• [15/02/2021] scPipe scATAC-Seq module also now uses the SingleCellExperiment class.
• [21/01/2021] now uses macsr for peak calling (yet via R developmental version, R.0.1)
• [03/01/2021] add the cell calling function to scATAC-Seq module for ScPipe (using the package DropletUtils)
• [12/12/2020] complete 1st version of scATAC-Seq module for ScPipe
• [13/05/2020] initiate scATAC-Seq module for scPipe
• [21/09/2017] scPipe now uses the SingleCellExperiment class.

Installation

From Bioconductor

if (!requireNamespace("BiocManager", quietly=TRUE))
    install.packages("BiocManager")
BiocManager::install("scPipe")

From GitHub (Developmental version)

install.packages("devtools")
devtools::install_github("LuyiTian/scPipe")

Getting started

The general workflow of scPipe is illustrated in the following figure:

Concept for scRNA-Seq preprocessing

• The sc_trim_barcode function will reformat each read and put the cell barcode and UMI sequence into the fastq read names: @ACGATCGA_TAGAGC#SIMULATE_SEQ::002::000::0000::0 AAGACGTCTAAGGGCGGTGTACACCCTTTTGAGCAATGATTGCACAACCTGCGATCACCTTATACAGAATTAT+AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA

• After alignment, the sc_exon_mapping function will put the cell barcode and UMI into the bam file with different tags, together with gene information: AAAGTCAA_AACTCA#SIMULATE_SEQ::007::000::0013::10 0 ERCC-00171 142 40 73M * 0 0 GCCTCGGGAATAAGCTGACGGTGACAAGGTTTCCCCCTAATCGAGACGCTGCAATAACACAGGGGCATACAGT AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA HI:i:1 NH:i:1 NM:i:0 GE:Z:ERCC-00171 YC:Z:AAAGTCAA YM:Z:AACTCA YE:i:-364. In this example the cell barcode is AAAGTCAA with tag YC, the UMI is AACTCA with tag YM and the gene that this read maps to is ERCC-00171 with tag GE. This read is located 364 bp upstream of the transcription end site (TES), which is stored in the YE tag.

• The sc_demultiplex function will look for the cell barcode in BAM file (by default in the YC tag) and compare it against the known cell barcode annotation file, which is a csv file consisting of two columns. The first column is the cell name and second column is the cell barcode. For Chromium 10x and Drop-seq data we can run sc_detect_bc to find the barcodes and generate the cell barcode annotation file before running sc_demultiplex. An example barcode annotation file is available in the package from system.file("extdata", "barcode_anno.csv", package = "scPipe"). The output of sc_demultiplex will be multiple csv files corresponding to each cell. Each file has three columns, the first of which contains the gene id, the second column contains the UMI sequence and third column gives the relative location of the read to the TES. These files are used for sc_gene_counting.

For further examples see the vignette.

Concept for scATAC-Seq preprocessing

• The sc_atac_trim_barcode function will reformat each read and put the cell barcode and UMI sequence into the fastq read names: @ACGATCGA_TAGAGC#SIMULATE_SEQ::002::000::0000::0 AAGACGTCTAAGGGCGGTGTACACCCTTTTGAGCAATGATTGCACAACCTGCGATCACCTTATACAGAATTAT+AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA

• The sc_atac_aligning function will align the reformatted fastq files and create bam files.

• After alignment, the sc_atac_bam_tagging function will put the cell barcode (and UMI, if available) into the bam file with different tags: AAAGTCAA_AACTCA#SIMULATE_SEQ::007::000::0013::10 0 ERCC-00171 142 40 73M * 0 0 GCCTCGGGAATAAGCTGACGGTGACAAGGTTTCCCCCTAATCGAGACGCTGCAATAACACAGGGGCATACAGT AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA HI:i:1 NH:i:1 NM:i:0 GE:Z:ERCC-00171 YC:Z:AAAGTCAA YM:Z:AACTCA YE:i:-364. In this example the cell barcode is AAAGTCAA with tag YC, the UMI is AACTCA with tag YM and the gene that this read maps to is ERCC-00171 with tag GE. This read is located 364 bp upstream of the transcription end site (TES), which is stored in the YE tag.

• The sc_atac_peak_calling function can be used to call peaks using macsr via Linux/Mac environment(). However, macsr is yet not compatible with Windows.

• The sc_atac_feature_counting function will generate a feature-count matrix for the alignment and an input feature file (a genome, a bed file format of features, for example one generated through sc_atac_peak_calling or MACS2/3). If the feature file is a genome.fasta file a genome_bin approach is used to create the features. It would also generate quality statistics that will get stored in the scPipe_atac_stats folder within the working directory. Cell calling is a function implemented in this function to identify the "true" cells.

• The function sc_atac_create_sce generates the Single Cell Experiment object from the feature-count matrix and the quality scores acquired throughout the pipeline. It also allows the user to generate a HTML report which can alternatively be created by the sc_atac_create_report function.

• The function sc_atac_create_report can be run within the sc_atac_create_sce or independently to create a report based on the quality statistics available through the processed pipeline.

A minimal example for scATAC-Seq module of scPipe is available here. For further examples see the relevant vignette.

Acknowledgments

This package is inspired by the scater, scran and scATAC=pro packages. The idea to put cell barcode and UMI sequences into the BAM file is from Drop-seq tools. Also some features of the scPipe-ATAC module were inspired by the scATAC-pro and SnapTools packages. We thank Dr Aaron Lun for suggestions on package development.