Interoperability: How to convert Spectre data.table to object compatible with CATALYST package?

[denvercal1234GitHub]

Hi there,

I lov Spectre package especially with the well-crafted tutorials and explanation. But, I also hope to use some features provided by the CATALYST package.

CATALYST package organizes their data as a SingleCellExperiment class (https://bioconductor.org/packages/release/bioc/vignettes/CATALYST/inst/doc/differential.html).

Do you have a recommended way to do this?

Thank you for your help!

[tomashhurst]

@denvercal1234GitHub it's pretty straightforward to construct a SingleCellExperiment from a data.table, and something I have been meaning to create a quick tutorial for. @ghar1821 would you have time to write up a quick example?

[ghar1821]

Yep I'll have time to write up a quick tutorial. I'll post a link here when I've written it up!

[denvercal1234GitHub]

Thank you so much, @ghar1821 !

[denvercal1234GitHub]

hI @ghar1821 - Would you mind letting me know whether you had gotten a chance to write the tutorial for converting Spectre data.table to SCE object? Thank you so much!

[ghar1821]

@denvercal1234GitHub
I've written up a full tutorial here: https://immunedynamics.io/spectre/tutorials/datatable_interoperability/sce_support.html

Let me know if you need help.

[ghar1821]

@denvercal1234GitHub

Sorry for the huge delay! I haven't written a full tutorial, but I'll attach a brief one here which should help:

With sce, cell-specific (e.g., what sample or group each cell belongs to) or marker-specific (e.g., what are the channels) must be separate from the actual data.
See the following where I mock up my data and store it as dt, and the cell- and marker-specific information as cell_info and marker_info.

dt <- data.table(
    CD4=sample(1:100, 20, replace = TRUE),
    CD8=sample(1:100, 20, replace = TRUE),
    CD3=sample(1:100, 20, replace = TRUE))

marker_info <- data.table(
    name=c("CD4", "CD8", "CD3"),
    type=rep("Cell type", 3)
)

cell_info <- data.table(
    cell_id=paste0("Cell_", c(1:20)),
    sample="01",
    group="healthy"
)

Next, to load them up into sce, we start with just the data, like the following:

library(SingleCellExperiment)
sce <- SingleCellExperiment(assays=list(counts=t(dt)))

Note a couple of things:

1. We store the data in an assay call count in sce.
2. I transformed the data so the cells are columns, and markers are row. This is the standard for sce and scRNAseq data.

Then we load up the marker- and cell- specific information as rowData and colData respectively, while also giving the row and column in the sce object a unique id. For markers (rows), it can just be the name, for cells (columns), you will probably have to make it up using something like paste0("CellId"_c(1:nrow(dt))) if you don't have anything that uniquely identifies each cell.

# Load up the marker information.
# Unique name for each row seem to already be loaded from previous step. 
rowData(sce) <- DataFrame(marker_info, row.names = marker_info$name)

# Give each cell a unique ID, already embedded in my original data table
colnames(sce) <- cell_info$cell_id
# Then load up the rest of the information
colData(sce) <- DataFrame(cell_info, row.names = cell_info$cell_id)

Check your object:

sce

and it should look like this:

class: SingleCellExperiment 
dim: 3 20 
metadata(0):
assays(1): counts
rownames(3): CD4 CD8 CD3
rowData names(2): name type
colnames(20): Cell_1 Cell_2 ... Cell_19 Cell_20
colData names(3): cell_id sample group
reducedDimNames(0):
mainExpName: NULL
altExpNames(0):

Hope this help! Let me know if you get stuck or need more info. I'll write up a proper tutorial in the coming days.