AUCell scores change depending on how many genesets are used?

[andrewdchen]

Hello!

Noticed that the AUCell scores change depending on the number of genesets tested. I'm using the toy data below and the values are the different:
cells_AUC_1:

AUC for 2 gene sets (rows) and 500 cells (columns).

Top-left corner of the AUC matrix:
          cells
gene sets       Cell1 Cell2 Cell3 Cell4     Cell5
  geneSet1 0.04444444     0     0     0 0.0000000
  geneSet2 0.00000000     0     0     0 0.2571429

cells_AUC_2

AUC for 1 gene sets (rows) and 500 cells (columns).

Top-left corner of the AUC matrix:
          cells
gene sets       Cell1 Cell2 Cell3 Cell4      Cell5
  geneSet1 0.04444444     0   0.2     0 0.04444444

Is there a specific reason that this is the case or a specific parameter that avoid this? I tried setting normAUC=FALSE and the two scores are still different.

Code to reproduce:

library(AUCell)

set.seed(123)
exprMatrix <- matrix(data=sample(c(rep(0, 50000), sample(1:3, 50000, replace=TRUE))),
                     nrow=200, 
                     dimnames=list(paste("Gene", 1:200, sep=""), 
                                   paste("Cell", 1:500, sep="")))
exprMatrix <- as(exprMatrix, "dgCMatrix")

fewGenes <- sample(rownames(exprMatrix), 10)
otherGenes <- sample(rownames(exprMatrix), 5)

cells_AUC_1 <- AUCell_run(exprMatrix, list(geneSet1=fewGenes,
                 geneSet2=otherGenes))

cells_AUC_2<- AUCell_run(exprMatrix, list(geneSet1=fewGenes))