Cannot convert the sce to seurat object

[Ahmedalaraby20]

Hey guys,
Congratulation on such amazing work,
I am just having some issues converting the sce to Seurat object

allseurat <- as.Seurat(All)

Error in UseMethod(generic = "as.sparse", object = x) : 
  no applicable method for 'as.sparse' applied to an object of class "c('DelayedMatrix', 'DelayedArray', 'DelayedUnaryIsoOp', 'DelayedUnaryOp', 'DelayedOp', 'Array', 'RectangularData')"

Would appreciate if you could help with that :D
Thanks a lot

[federicomarini]

Hi Ahmed,
From the error message you are posting, this is because the casting function to Seurat is not able to handle the conversion of the assays slots, which are (cleverly as it frees up memory while you have loaded the objects) stored with the DelayedArray backend.
If your machine supports that, I think you could enforce the conversion to a "classical" in-memory matrix, just be ready to use up a substantial amount of RAM.
Feel free to follow up on this on the Bioc slack (https://bioc-community.herokuapp.com/), there might be cleverer ways to do this - can just be that I am not fully aware of the palette of methods to handle this.
Federico

[massonix]

Hi Ahmed,

This is indeed an issue on Seurat's side, as described in this issue. Following their advice, you could run this before conversion:

counts(sce) <- as(counts(sce, withDimnames = FALSE), "dgCMatrix")

Ramon

[Ahmedalaraby20]

Thanks alot guys,
I switched to Scanpy :D