vector size

[antoine4ucsd]

Hello I was able to install your great package. it works fine. but I have an error with the following: My goal is to do a power analyses for 20 samples with 500 and 1000 (and ideally 5000) cells per sample. I started with smthg like that

#read in count seurat object with spleed scRNAseq data
spleen.filt <- readRDS('spleen_filt_cells.rds', refhook = NULL)

#extract counts from Seurat object
counts <- as.matrix(GetAssayData(object = spleen.filt, slot = "counts"))

setupres <- Setup(ngenes = 10000, nsims = 10, 
                  p.DE = 0.05, pLFC = p.lfc,
                  n1 = c(10000,20000), n2 = c(10000,20000), 
                  Thinning = NULL, LibSize = 'equal',
                  estParamRes = estparam,
                  estSpikeRes = NULL,
                  DropGenes = TRUE,
                  setup.seed = 5299, verbose = TRUE) 

The provided count matrix has 3881 out of 3881 single cells and 21901 out of 33539 genes with at least 1 count.

# define log fold change
p.lfc <- function(x) sample(c(-1,1), size=x,replace=T)*rgamma(x, shape = 1, rate = 2)
# set up simulations
setupres <- Setup(ngenes = 10000, nsims = 5, # was initially set to 25
                  p.DE = 0.05, pLFC = p.lfc,
                  n1 = c(10000,20000), n2 = c(10000,20000), # number of cells for example, if 20 samples, this will evaluate 500 cells and 1000 cells
                  Thinning = NULL, LibSize = 'equal',
                  estParamRes = estparam,
                  estSpikeRes = NULL,
                  DropGenes = TRUE,
                  setup.seed = 5299, verbose = TRUE)

but I get an error due to the vector size... " cannot allocate a vector of size 1.3Go"

is there a way to make it work for such sample size?

thank you!

[bvieth]

Hello,

with your defined setup the matrices indeed will be very large. If you want to create 20 'pseudo' replicates of 500 vs 1000, you can just run the Setup() with nsims = 20 and n1 = 500 and n2 = 1000.

HTH Beate

[antoine4ucsd]

thank you!