Discrepancy in results from v0.0.16 to 0.0.17

[DCarbonez]

First of all, thank you for releasing this package.

Results ran on these versions show small changes. Commands used:

# gene set
hsmg <-
  buildMSIGDB(species = "human",
              keytype = "SYMBOL",
              anntype = "HALLMARK")

# v0.0.16
res_HALLMARK_1 <-
  scgsva(obj = seurat_object, hsmg, assay = "RNA")

# v0.0.17
res_HALLMARK_2 <-
  scgsva(obj = seurat_object, hsmg, assay = "RNA")

# correlation plots
plot(x = res_HALLMARK_1@gsva$TNFA_SIGNALING_VIA_NFKB,y = res_HALLMARK_2@gsva$TNFA_SIGNALING_VIA_NFKB)

I checked & both results have the same results in the annot slot & were built upon the same seurat object.

Correlations between both results are shown int his plot

Is there an explanation for this?

Thank you

[guokai8]

Hi @DCarbonez ,
Thanks for testing the different versions. Here is a random seed used for the GSVA calculation in the GSVA package. I didn't change anything about the calculation for different versions. But I will check the discrepancy between different versions even just small changes. However, I am sure this won't affect the trend or the differences within the same version.

Best,
K