Closed jpreall closed 4 years ago
The memory and compute requirements of SoupX should not be particularly high (unless you are opting to use cell specific contamination estimates, which I do not recommend). Which step exactly is taking hours to run? Are you providing sparse matricies as inputs?
I have used SoupX successfully in the past on 10X v2 libraries, which have far fewer possible barcodes than the v3 chemistry. Calculating cell specific estimates takes a few minutes on my laptop on v2 libraries, but many hours on v3 chemistry and ultimately exhausts available vector memory. Is this a known issue with the current version? Adjusting counts also seems to hang.
I also experienced performance problems on a fresh install of R/Rstudio/SoupX on a new iMac.
Perhaps there is some dependency that is not configured properly on my machines for parallelization?