LoadPancreasData

In cytomapper: Visualization of highly multiplexed imaging data in R

knitr::opts_chunk$set(echo = TRUE)
knitr::opts_knit$set(root.dir = file.path("..", "extdata"))

Script to download single-cell data from the pancreas IMC dataset and format the data as a SingleCellExperiment object

• Publication: Damond et al. A Map of Human Type 1 Diabetes Progression by Imaging Mass Cytometry. Cell Metab. 2019 Mar 5;29(3):755-768
• Dataset: accessible from Mendeley Data

library(S4Vectors)
library(SingleCellExperiment)
library(cytomapper)

Read-in single-cell data

Here, a subset of single-cell data, corresponding to 100 images from the full dataset is downloaded.

# Download the zipped folder image and unzip it
url.cells <- ("https://data.mendeley.com/public-files/datasets/cydmwsfztj/files/f1e3b8dc-56be-4172-bbc4-3a6f9de97563/file_downloaded")
download.file(url.cells, destfile = "CellSubset.zip")
unzip("CellSubset.zip")
file.remove("CellSubset.zip")

# Read-in the data
cells <- read.csv("CellSubset.csv", stringsAsFactors = FALSE)

# Order the dataset by ImageNumber and ObjectNumber
cells <- cells[order(cells$ImageNumber, cells$ObjectNumber), ]

Read-in Image metadata

# Download the zipped folder image and unzip it
url.image <- ("https://data.mendeley.com/public-files/datasets/cydmwsfztj/files/0b236273-d21b-4566-84a2-f1c56324a900/file_downloaded")
download.file(url.image, destfile = "Image.zip")
unzip("Image.zip")
file.remove("Image.zip")

# Read-in the data
image <- read.csv("All_Image.csv", stringsAsFactors = FALSE)

Read-in cell type information

# Download the zipped folder image and unzip it
url.celltypes <- ("https://data.mendeley.com/public-files/datasets/cydmwsfztj/files/59e8da72-5bfe-4289-b95b-28348a6e1222/file_downloaded")
download.file(url.celltypes, destfile = "CellTypes.zip")
unzip("CellTypes.zip")
file.remove("CellTypes.zip")

# Read-in the data
celltypes <- read.csv("CellTypes.csv", stringsAsFactors = FALSE)

Read-in donor information

# Download the zipped folder image and unzip it
url.donors <- ("https://data.mendeley.com/public-files/datasets/cydmwsfztj/files/9074990e-1b93-4c79-8c49-1db01a66398b/file_downloaded")
download.file(url.donors, destfile = "Donors.zip")
unzip("Donors.zip")
file.remove("Donors.zip")

# Read-in the data
donors <- read.csv("Donors.csv", stringsAsFactors = FALSE)

Load relevant cell-specific metadata

cell.metadata <- DataFrame(ImageNumber = cells$ImageNumber,
                           CellNumber = cells$ObjectNumber,
                           Pos_X = cells$Location_Center_X,
                           Pos_Y = cells$Location_Center_Y,
                           ParentIslet = cells$Parent_Islets,
                           ClosestIslet = cells$Parent_ExpandedIslets,
                           Area = cells$AreaShape_Area,
                           NbNeighbours = cells$Neighbors_NumberOfNeighbors_3)

Load relevant image-specific metadata

image.metadata <- DataFrame(ImageNumber = image$ImageNumber,
                            ImageFullName = image$FileName_CleanStack,
                            slide = image$Metadata_Slide,
                            width = image$Width_CleanStack,
                            height = image$Height_CleanStack)

Merge cell and image metadata

cell.metadata <- merge(cell.metadata, image.metadata, by="ImageNumber")

Add image names

This information is used by cytomapper to match single-cell data with images and masks

cell.metadata$ImageName <- sub("_a0_full_clean.tiff", "", cell.metadata$ImageFullName)

Import cell types

# Add cell ids to cell metadata (format: "ImageName_CellNumber")
cell.metadata$id <- paste(cell.metadata$ImageName, cell.metadata$CellNumber, sep="_")

# Merge cell metadata and cell type information
cell.metadata <- merge(cell.metadata,
                       celltypes[, c("id", "CellCat", "CellType")],
                       by="id")

Import donor metadata

cell.metadata <- merge(cell.metadata, donors, by="slide")

Order the cell metadata dataset and add rownames

# Rows are ordered by ImageNumber and CellNumber
cell.metadata <- cell.metadata[order(cell.metadata$ImageNumber, cell.metadata$CellNumber), ]

# Cell ids are used as row names
rownames(cell.metadata) <- cell.metadata$id

Load panel data

The panel contains antibody-related metadata. The channel-mass file is used to match panel information and image stack slices.

# Import panel
url.panel <- ("https://data.mendeley.com/public-files/datasets/cydmwsfztj/files/2f9fecfc-b98f-4937-bc38-ae1b959bd74d/file_downloaded")
download.file(url.panel, destfile = "panel.csv")
panel <- read.csv("panel.csv")

# Import channel-mass file
url.channelmass <- ("https://data.mendeley.com/public-files/datasets/cydmwsfztj/files/704312eb-377c-42e2-8227-44bb9aca0fb3/file_downloaded")
download.file(url.channelmass, destfile = "ChannelMass.csv")
channel.mass <- read.csv("ChannelMass.csv", header = FALSE)

Select relevant channels and match them with image stack slices

# Match panel and stack slice information
panel <- panel[panel$full == 1,]
panel <- panel[match(channel.mass[,1], panel$MetalTag),]

# Add short protein names as panel rownames
rownames(panel) <- panel$shortname

Load single cell measurements

Here, we import the mean intensity per cell

cur_counts <- cells[, grepl("Intensity_MeanIntensity_CleanStack", colnames(cells))]

Reorder the counts channels (based on channel number)

channelNumber <- as.numeric(sub("^.*_c", "", colnames(cur_counts)))
cur_counts <- cur_counts[, order(channelNumber, decreasing = FALSE)]

Create the SingleCellExperiment (SCE) object

sce <- SingleCellExperiment(assays = list(counts = t(as.matrix(cur_counts))))

Add transformed counts as a new assay

exprs = asinh-transformed counts

assay(sce, "exprs") <- asinh(counts(sce)/1)

Set dimnames

rownames(sce) <- rownames(panel)
colnames(sce) <- rownames(cell.metadata)

Store metadata in the SCE object

colData(sce) <- cell.metadata
rowData(sce) <- panel
sce

Save SCE

saveRDS(sce, "pancreas_sce.rds")

Delete unneeded CSV files from the extdata directory

file.remove("All_Image.csv", "CellSubset.csv", "CellTypes.csv", "Donors.csv", "panel.csv", "ChannelMass.csv")

cytomapper documentation built on Jan. 30, 2021, 2:01 a.m.