Report renv causes jupyterlab kernel die

[dfc-1931475]

Hello, thanks for developing this useful tool.

I got some problem when I use renv in jupyterlab:

When I ran a single cell project practice here, More specifically, the **find_de_MAST_RE**function in Chapter16.5.1. It is to use MAST package to fit model. However, once the code is coming to summaryCond <- summary(zlmCond, doLRT='groupstim.CD14_Monocytes')(within the find_de_MAST_RE function), my jupyterlab will report kernel died immediately, without any traceback information. The code will work well if I run renv::deactivate command, may you know why could this happen? Is there any method to make jupyterlab and renv work well with each other?

the find_de_MAST_RE function is like below:

  %%R
  find_de_MAST_RE <- function(adata_){
      # create a MAST object
      sca <- SceToSingleCellAssay(adata_, class = "SingleCellAssay")
      print("Dimensions before subsetting:")
      print(dim(sca))
      print("")
      # keep genes that are expressed in more than 10% of all cells
      sca <- sca[freq(sca)>0.1,]
      print("Dimensions after subsetting:")
      print(dim(sca))
      print("")
      # add a column to the data which contains scaled number of genes that are expressed in each cell
      cdr2 <- colSums(assay(sca)>0)
      colData(sca)$ngeneson <- scale(cdr2)
      # store the columns that we are interested in as factors
      label <- factor(colData(sca)$label)
      # set the reference level
      label <- relevel(label,"ctrl")
      colData(sca)$label <- label
      celltype <- factor(colData(sca)$cell_type)
      colData(sca)$celltype <- celltype
      # same for donors (which we need to model random effects)
      replicate <- factor(colData(sca)$replicate)
      colData(sca)$replicate <- replicate
      # create a group per condition-celltype combination
      colData(sca)$group <- paste0(colData(adata_)$label, ".", colData(adata_)$cell_type)
      colData(sca)$group <- factor(colData(sca)$group)
      # define and fit the model
      zlmCond <- zlm(formula = ~ngeneson + group + (1 | replicate), 
                     sca=sca, 
                     method='glmer', 
                     ebayes=F, 
                     strictConvergence=F,
                     fitArgsD=list(nAGQ = 0)) # to speed up calculations
      
      # perform likelihood-ratio test for the condition that we are interested in    
      summaryCond <- summary(zlmCond, doLRT='groupstim.CD14_Monocytes')
      # get the table with log-fold changes and p-values
      summaryDt <- summaryCond$datatable
      result <- merge(summaryDt[contrast=='groupstim.CD14_Monocytes' & component=='H',.(primerid, `Pr(>Chisq)`)], # p-values
                       summaryDt[contrast=='groupstim.CD14_Monocytes' & component=='logFC', .(primerid, coef)],
                       by='primerid') # logFC coefficients
      # MAST uses natural logarithm so we convert the coefficients to log2 base to be comparable to edgeR
      result[,coef:=result[,coef]/log(2)]
      # do multiple testing correction
      result[,FDR:=p.adjust(`Pr(>Chisq)`, 'fdr')]
      result = result[result$FDR<0.01,, drop=F]
  
      result <- stats::na.omit(as.data.frame(result))
      return(result)
  }