how to accelarate the scMerge2 process

[xflicsu]

I use scMerge2 to integrate about 150K cells.
Now, it costed about 5 hours and still run the "Running RUV" step with 120 CPU.
I wonder how to accelarate the process like your paper mentioned?
Thanks!

##################
scMerge2_res <- scMerge2(exprsMat = logcounts(sce),
batch = sce$orig.ident,condition=sce$type,chosen.hvg=hgvs,return_matrix = FALSE,
verbose = TRUE,use_bpparam = BiocParallel::SerialParam()
)
[1] "Cluster within batch"
[1] "Normalising data"
[1] "Constructing pseudo-bulk"
Dimension of pseudo-bulk expression: [1] 33341 16083
[1] "Identifying MNC using pseudo-bulk:"
[1] "condition_mode"
[1] "Running RUV"

[YingxinLin]

Hi thank you for your interest in scMerge2.

I am wondering how many batches and conditions do you have for your dataset? (assuming condition of the sample is included as sce$type here). If you want to run scMerge2 in parallel, you can set use_bpparam = BiocParallel::MulticoreParam(workers = ncores).

Best wishes,
Yingxin

[xflicsu]

Hi thank you for your interest in scMerge2.

I am wondering how many batches and conditions do you have for your dataset? (assuming condition of the sample is included as sce$type here). If you want to run scMerge2 in parallel, you can set use_bpparam = BiocParallel::MulticoreParam(workers = ncores).

Best wishes, Yingxin

Thanks for your quick response!

I have 150k cells with 40 samples (sce$orig.ident) and 3 conditions (sce$type).
The process work in parallel.

[DarioS]

Did you replace use_bpparam = SerialParam() by use_bpparam = MulticoreParam(workers = ncores) in the end?