Xenium data loading failed--out of memory

[lh12565]

Hi, I have a Xenium data which has 36 samples. When loading the Xenium data into the Giotto object, it consumes excessive memory (>1 T) and fails to complete the reading process.

instrs = createGiottoInstructions(
    save_dir = results_folder,
    save_plot = TRUE,
    show_plot = FALSE,
    return_plot = FALSE,
    python_path = python_path
)
feat_types <- c(
    "rna",
    "UnassignedCodeword", 
    "NegControlCodeword",  
    "NegControlProbe"
)
split_keywords = list(
    c("BLANK"),
    c("NegControlCodeword"),
    c("NegControlProbe", "antisense")
)

xenium_gobj <- createGiottoXeniumObject(
    xenium_dir = file.path(out_path),
    qv_threshold = 20,
    feat_type = feat_types,
    split_keyword = split_keywords,
    instructions = instrs
)

I found the “transcript” data is too large. When I load it like below, the subsequent code throws errors, and I'm unsure how to resolve this issue.

xenium_gobj <- createGiottoXeniumObject(
    xenium_dir = file.path(data_path),
    qv_threshold = 20, # qv of 20 is the default and also what 10x uses
    feat_type = feat_types,
    split_keyword = split_keywords,
    load_transcripts=F,
    load_expression=T,
    load_cellmeta=T,
    instructions = instrs
)

...

xenium_gobj <- [calculateOverlapRaster](https://drieslab.github.io/GiottoClass/reference/calculateOverlapRaster.html)(xenium_gobj,
    spatial_info = 'cell',
    feat_info = 'rna'
)


Error: Giotto object contains no feature point information

Can Giotto load just one single sample from Xenium data？ Thanks!

[matt-sd-watson]

Hello @lh12565 As an alternative, you can visualise multiple Xenium samples from spatialdata using our in-browser application rakaia: https://github.com/camlab-bioml/rakaia. When reading from multiple zarr spatialdata directories, we perform lazy loading so minimal memory is required to import and render multiple Xenium samples.