In drisso/scRNAseq: Collection of Public Single-Cell RNA-Seq Datasets

library(BiocStyle)
knitr::opts_chunk$set(error=FALSE, message=FALSE, warning=FALSE)

Specifying the samples

We obtain a single-cell RNA sequencing dataset of the mouse mammary gland from @bach2017differentiation. we define all of the samples to be pulled down.

accessions <- c(
    "GSM2834498", "GSM2834499", 
    "GSM2834500", "GSM2834501", 
    "GSM2834502", "GSM2834503", 
    "GSM2834504", "GSM2834505")

samples <- c(
    "NP_1", "NP_2",
    "G_1", "G_2",
    "L_1", "L_2",
    "PI_1", "PI_2")

conditions <- c(NP="Nulliparous",
    G="Gestation",
    L="Lactation",
    PI="Post-involution")[sub("_.*", "", samples)]

data.frame(accessions, samples, conditions)

Downloading the count data

We then download and cache the assorted count matrices using the r Biocpkg("BiocFileCache") package.

library(BiocFileCache)
bfc <- BiocFileCache("raw_data", ask = FALSE)

library(Matrix)
library(S4Vectors)
template.path <- "ftp://ftp.ncbi.nlm.nih.gov/geo/samples/%snnn/%s/suppl"
collected.counts <- collected.rowdata <- collected.coldata <- list()
for (i in seq_along(accessions)) {
    curacc <- accessions[i]
    subacc <- substr(curacc, 1, 7)
    base.path <- sprintf(template.path, subacc, curacc)
    header <- paste0(curacc, "%5F", sub("_", "%5F", samples[i]), "%5F")

    barcode.fname <- bfcrpath(bfc, file.path(base.path,
        paste0(header, "barcodes%2Etsv%2Egz")))
    gene.fname <- bfcrpath(bfc, file.path(base.path,
        paste0(header, "genes%2Etsv%2Egz")))
    counts.fname <- bfcrpath(bfc, file.path(base.path,
        paste0(header, "matrix%2Emtx%2Egz")))

    collected.counts[[i]] <- as(readMM(counts.fname), "dgCMatrix")
    gene.info <- read.table(gene.fname, stringsAsFactors=FALSE)
    colnames(gene.info) <- c("Ensembl", "Symbol")
    collected.rowdata[[i]] <- DataFrame(gene.info)
    collected.coldata[[i]] <- DataFrame(
        Barcode=readLines(barcode.fname),
        Sample=samples[i],
        Condition=conditions[i])
}

We verify that all of the row data matches up, and that all the dimensions are consistent.

stopifnot(length(unique(collected.rowdata))==1L)

X <- vapply(collected.coldata, nrow, 0L)
Y <- vapply(collected.counts, ncol, 0L)
stopifnot(identical(X, Y))
X

Saving for upload

We save these to file for upload to r Biocpkg("ExperimentHub").

path <- file.path("scRNAseq", "bach-mammary", "2.0.0")
dir.create(path, showWarnings=FALSE, recursive=TRUE)
saveRDS(collected.rowdata[[1]], file=file.path(path, "rowdata.rds"))
for (i in seq_along(samples)) {
    saveRDS(collected.counts[[i]], file=file.path(path, sprintf("counts-%s.rds", samples[i])))
    saveRDS(collected.coldata[[i]], file=file.path(path, sprintf("coldata-%s.rds", samples[i])))
}