Analyze snATAC-seq data of basal cell carcinoma sample SU008_Tumor_Pre in GEO (GSE129785).
library(scPloidy)
library(dplyr)
#>
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#>
#> filter, lag
#> The following objects are masked from 'package:base':
#>
#> intersect, setdiff, setequal, union
library(tidyr)
library(gplots)
#>
#> Attaching package: 'gplots'
#> The following object is masked from 'package:stats':
#>
#> lowess
You can skip the preprocessing and start from section CNV.
Download GSE129785_scATAC-TME-All.cell_barcodes.txt.gz from below and gunzip https://www.ncbi.nlm.nih.gov/geo/download/?acc=GSE129785&format=file&file=GSE129785%5FscATAC%2DTME%2DAll%2Ecell%5Fbarcodes%2Etxt%2Egz
Download GSM3722064_SU008_Tumor_Pre_fragments.tsv.gz from https://www.ncbi.nlm.nih.gov/geo/download/?acc=GSM3722064&format=file&file=GSM3722064%5FSU008%5FTumor%5FPre%5Ffragments%2Etsv%2Egz
The input window file window.hg37.20MB.bed
and resultant
peak file multi_tissue_peaks.hg37.20MB.bed
can be
downloaded from https://doi.org/10.6084/m9.figshare.23574066
To reproduce by yourself, download chromatin accessibility DHS_Index_and_Vocabulary_hg19_WM20190703.txt.gz from https://doi.org/10.5281/zenodo.3838751
Generate peaks for 20MB windows using peak_sum.R by yardimcilab in https://github.com/yardimcilab/RIDDLER/blob/main/util/peak_sum.R
See vignette of R package scPloidy. Load setting for hg19 genome.
simpleRepeat = readr::read_tsv(
"~/human/publichuman/hg37_ucsc/simpleRepeat.chrom_chromStart_chromEnd.txt.gz",
col_names = c("chrom", "chromStart", "chromEnd"))
rmsk = readr::read_tsv(
"~/human/publichuman/hg37_ucsc/rmsk.Simple_repeat.genoName_genoStart_genoEnd.txt.gz",
col_names = c("chrom", "chromStart", "chromEnd"))
simpleRepeat = rbind(simpleRepeat, rmsk)
rm(rmsk)
# convert from 0-based position to 1-based
simpleRepeat[, 2] = simpleRepeat[, 2] + 1
simpleRepeat = GenomicRanges::makeGRangesFromDataFrame(
as.data.frame(simpleRepeat),
seqnames.field = "chrom",
start.field = "chromStart",
end.field = "chromEnd")
# remove duplicates
simpleRepeat = GenomicRanges::union(simpleRepeat, GenomicRanges::GRanges())
window = read.table("window.hg37.20MB.bed", header = FALSE)
colnames(window) = c("chr", "start", "end", "window")
at = GenomicRanges::makeGRangesFromDataFrame(window[, 1:3])
barcodesuffix = paste0(".", window$window)
Compute and save fragmentoverlap.
sample = "GSM3722064"
tissue = "SU008_Tumor_Pre"
bc = sc$Barcodes[sc$Group == tissue]
SU008_Tumor_Pre_fragmentoverlap =
fragmentoverlapcount(
paste0("SRX5679934/", sample, "_", tissue, "_fragments.tsv.gz"),
at,
excluderegions = simpleRepeat,
targetbarcodes = bc,
Tn5offset = c(1, 0),
barcodesuffix = barcodesuffix
)
You can skip above and load preprocessed data attached to the package. The data file GSE129785_SU008_Tumor_Pre.RData is also available from https://doi.org/10.6084/m9.figshare.23574066
Infer CNVs.
levels = c(2, 4)
result = cnv(SU008_Tumor_Pre_fragmentoverlap,
SU008_Tumor_Pre_windowcovariates,
levels = levels,
deltaBICthreshold = -600)
#> [1] "Computing span = 2"
#> [1] "Computing span = 3"
#> [1] "Computing span = 4"
#> [1] "Computing span = 5"
#> [1] "Computing span = 6"
#> [1] "Computing span = 7"
#> [1] "Computing span = 8"
#> [1] "Computing span = 9"
#> [1] "Computing span = 10"
#> [1] "Computing span = 2"
#> [1] "Computing span = 3"
#> [1] "Computing span = 4"
#> [1] "Computing span = 5"
#> [1] "Computing span = 6"
#> [1] "Computing span = 7"
#> [1] "Computing span = 8"
#> [1] "Computing span = 9"
#> [1] "Computing span = 10"
Attach the result to fragmentoverlap
.
windowcovariates = SU008_Tumor_Pre_windowcovariates
windowcovariates$w =
as.numeric(sub("window_", "", windowcovariates$window))
fragmentoverlap = SU008_Tumor_Pre_fragmentoverlap
fragmentoverlap$cell =
sub(".window.*", "", fragmentoverlap$barcode)
fragmentoverlap$window =
sub(".*window", "window", fragmentoverlap$barcode)
fragmentoverlap$w =
as.numeric(sub("window_", "", fragmentoverlap$window))
x = match(fragmentoverlap$barcode,
result$cellwindowCN$barcode)
fragmentoverlap$CN = result$cellwindowCN$CN[x]
fragmentoverlap$ploidy.moment.cell = result$cellwindowCN$ploidy.moment.cell[x]
fragmentoverlap = fragmentoverlap[!is.na(fragmentoverlap$CN), ]
# For better hierarchical clustering
fragmentoverlap$pwindownormalizedcleanedceiled =
pmin(fragmentoverlap$CN, min(levels) * 2)
Make dataframe for plotting.
dataplot =
fragmentoverlap %>%
dplyr::select("w", "cell", "pwindownormalizedcleanedceiled") %>%
tidyr::pivot_wider(names_from = "w", values_from = "pwindownormalizedcleanedceiled")
dataplot = as.data.frame(dataplot)
rownames(dataplot) = dataplot$cell
dataplot = dataplot[, colnames(dataplot) != "cell"]
dataplot = as.matrix(dataplot)
n = max(as.numeric(colnames(dataplot)))
dataplot = dataplot[, match(as.character(1:n), colnames(dataplot))]
colnames(dataplot) = as.character(1:n)
Plot.
breaks = c(0, min(levels) - 1, min(levels) + 1, min(levels) * 2)
x = windowcovariates
x$chr[duplicated(windowcovariates$chr)] = NA
x = x$chr[match(colnames(dataplot), x$w)]
RowSideColors =
unlist(
lapply(
fragmentoverlap$ploidy.moment.cell[
match(rownames(dataplot), fragmentoverlap$cell)],
function (x) { which(sort(levels) == x)}))
RowSideColors = topo.colors(length(levels))[RowSideColors]
gplots::heatmap.2(
dataplot,
Colv = FALSE,
dendrogram = "none",
breaks = breaks,
col = c("blue", "gray80", "red"),
trace = "none", labRow = FALSE, na.color = "white",
labCol = x,
RowSideColors= RowSideColors)