Skip to contents

How to create a SpatialFeatureExperiment object

Kayla Jackson

2023-11-29

Source: vignettes/create_sfe_v2.Rmd
create_sfe_v2.Rmd

Downloading the data

The data used are from a recent publication, High Resolution Slide-seqV2 Spatial Transcriptomics Enables Discovery of Disease-Specific Cell Neighborhoods and Pathways and are available for download from GEO (Accession Number: GSE190094.

We will demonstrate use of ffq to access FTP links for downloading the relevant data. We will only download the data for a single WT sample. The commented line shows how to install ffq from the R terminal.

# system("pip install ffq")
system("ffq -l1 GSM5713341")

The output of the command is metadata for GSM5713341. We can use curl or wget to download files with FTP links one-by-one.

Files beginning with ftp:// can be read directly with the R package vroom. Files do not have be uncompressed before reading. These files will be automatically downloaded and uncompressed. We will use this method here, but the commented lines show how to download the files using curl.

# system("curl -O ftp://ftp.ncbi.nlm.nih.gov/geo/samples/GSM5713nnn/GSM5713341/suppl/GSM5713341_Puck_191112_04_MappedDGEForR.csv.gz")

# system("curl -O ftp://ftp.ncbi.nlm.nih.gov/geo/samples/GSM5713nnn/GSM5713341/suppl/GSM5713341_Puck_191112_04_BeadLocationsForR.csv.gz")

# list.files(pattern = "*.gz")

Reading in the data 

mtx <- vroom("ftp://ftp.ncbi.nlm.nih.gov/geo/samples/GSM5713nnn/GSM5713341/suppl/GSM5713341_Puck_191112_04_MappedDGEForR.csv.gz")

centroids <- vroom("ftp://ftp.ncbi.nlm.nih.gov/geo/samples/GSM5713nnn/GSM5713341/suppl/GSM5713341_Puck_191112_04_BeadLocationsForR.csv.gz")

Construct a SFE object

The count matrix and bead locations are provided by the authors. We will pass these to the constructor for the SpatialFeatureExperiment object. The files are read in as data frames. We will convert the gene count matrix to a matrix and then a sparse dgCMatrix.

# Note: if using Google Colab, this step might run out of RAM
# If this happens, please upgrade to Colab Pro

rn <- mtx$Row
mtx <- as.matrix(mtx[,-1])

rownames(mtx) <- rn
mtx <- as(mtx, "dgCMatrix")

Here, spot locations are provided as a CSV file. There are two columns of particular interest, namely xcoord and ycoord. The barcode column corresponds to the barcodes in the count matrix. Before calling the SpatialFeatureExperiment constructor, the spatial coordinates must be converted to a sf data frame using df2sf(). The coordinates are centroid positions, so we will indicate that geometryType="POINT".

colnames(centroids)[1] <- "ID"

centroids <- df2sf(
  centroids, geometryType = "POINT",
  spatialCoordsNames=c("xcoord","ycoord"))

Now we have the ingredients to create a SFE object. The values to the assays and colGeometries arguments must be passed as a list as shown below.

sfe <- SpatialFeatureExperiment(
  assays = list(counts = mtx),
  colGeometries = list(centroids = centroids)
)

sfe

Session info