Closed DarioS closed 6 months ago
I have a typical-sized data set and would not expect any numerical issues. But it happens. Latest release versions of R and packages.
library(SingleCellExperiment) library(scuttle) library(scran) SCE <- SingleCellExperiment(assays = list(counts = matrix(rpois(21711 * 111671, 10), ncol = 111671))) SCE <- logNormCounts(SCE) SCE$cellType <- rep(c("Fibroblasts", "Other", "Tumor cells"), c(40967, 7374, 63330)) scoreMarkers(SCE, SCE$cellType, lfc = 3) Warning messages: 1: In left.ncells * right.ncells : NAs produced by integer overflow 2: In left.ncells * right.ncells : NAs produced by integer overflow
The dimensions of the matrix and number of samples per class are identical to a biological data set.
Oops. Fixed by f180b0fe767731aaa53f501a8431356cb599684c in 1.31.2. Check if it works and I'll push it to release.
It was tried with biological data and the problem is gone.
I have a typical-sized data set and would not expect any numerical issues. But it happens. Latest release versions of R and packages.
The dimensions of the matrix and number of samples per class are identical to a biological data set.