error when run computeCommunProb

[XinmiaoYan]

hi, nice tool, but I encountered a problem when running computeCommunProb, the issue is listed below:

cellchat <- computeCommunProb(cellchat)
Error: MultisessionFuture (future_sapply-1) failed to call grmall() on cluster RichSOCKnode #4 (PID 6869 on localhost ‘localhost’). The reason reported was ‘error writing to connection’. Post-mortem diagnostic: No process exists with this PID, i.e. the localhost worker is no longer alive. The total size of the 7 globals exported is 214.51 KiB. The three largest globals are ‘data.use’ (200.60 KiB of class ‘numeric’), ‘...future.FUN’ (13.63 KiB of class ‘function’) and ‘Rsubunits’ (224 bytes of class ‘list’)

thanks for your response

[sqjin]

@xinmiaoyan I am sorry I have no idea on this error. Are you working on a cluster instead of desktop/labtop?

[YuyanRuan]

@sqjin Hi! I got the same error. Yes, I am working on a HPC cluster. Looking forward to your reply

[VaiskiVemmel]

Hi. I also got this error. It worked fine while I was working on my full Seurat data but I tried to subset my samples based on treatment to see which pathways are active in different treatments and after subsetting control<-subset(data, subset = orig.ident == "control") I got this errorwhen I reached the cellchat <- computeCommunProb(cellchat) part.

[sqjin]

You may have to run future::plan("multiprocess", workers = 4) # do parallel options(future.globals.maxSize = 1000 * 1024^2)

On Fri, Jun 2, 2023 at 6:09 PM VaiskiVemmel @.***> wrote:

Hi. I also got this error. It worked fine while I was working on my full data but I tried to subset my samples based on treatment to see which pathways are active in different treatments and after subsetting control<-subset(data, subset = orig.ident == "control") I got this errorwhen I reached the cellchat <- computeCommunProb(cellchat) part.

— Reply to this email directly, view it on GitHub https://github.com/sqjin/CellChat/issues/326#issuecomment-1573486195, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHXF63BB7VJNAY3CGCXWPOTXJG3WNANCNFSM5IMV2JPA . You are receiving this because you were mentioned.Message ID: @.***>

[VaiskiVemmel]

Hi.

I tried that but apparently "multiprocess" is defunct and R suggest me to use "multisession" instead. In the earlier part of the code I have: CCobj<-subsetData(CCobj) future::plan("multisession", workers = 4) parallelly::supportsMulticore() CCobj<-identifyOverExpressedGenes(CCobj) CCobj<-projectData(CCobj, PPI.mouse) I also tried to add future::plan("multiprocess", workers = 4) # do parallel options(future.globals.maxSize = 1000 * 1024^2) before computeCommunProb but it still gave the same error.

UPDATE: I tried with the treated group and it runs with no problem. I would guess the empty cell group in control sample is causing the error. When I created the CellChat object using CCobj<-createCellChat(object = control, group.by = "clusterID") , there was 1 cell population that was not present in the control group and thus the group is empty.

As a solution for others suffering from the same error, I recommend to check that your data does not include empty groups and remove them if they do.