Question about MetacellsByGroups

[bprager5]

I have a seurat object that has been integrated with harmony, but when I try to run MetacellsByGroups using the harmony reduction, I get the following error:

Warning in (function (seurat_obj, name = "agg", ident.group = "seurat_clusters", :
Metacell failed
Error in h(simpleError(msg, call)) :
error in evaluating the argument 'x' in selecting a method for function 'as.data.frame': no applicable method for @ applied to an object of class "NULL"

traceback()
7: h(simpleError(msg, call))
6: .handleSimpleError(function (cond)
.Internal(C_tryCatchHelper(addr, 1L, cond)), "no applicable method for @ applied to an object of class "NULL"",
base::quote(x@misc))
5: FUN(X[[i]], ...)
4: lapply(metacell_list, function(x) {
x@misc$run_stats
})
3: do.call(rbind, lapply(metacell_list, function(x) {
x@misc$run_stats
}))
2: as.data.frame(do.call(rbind, lapply(metacell_list, function(x) {
x@misc$run_stats
})))
1: MetacellsByGroups(seurat_obj = x, reduction = "harmony", group.by = c("sample"),
k = 25, max_shared = 10, min_cells = 30, ident.group = "sample")

The 'harmony' reduction is in my seurat object and can be plotted, etc.

Reductions(x)
[1] "pca" "harmony" "umap"

It appears that MetacellsByGroups is looking for something in the @misc slot, which is empty in the newer seurat integration workflows. When I use other reductions it works, but there is very strong separation of metacells by sample, despite harmonizing metacells, so I am hesitant to proceed using the seurat object pca reduction.

Thanks

[smorabit]

Hi, could you please provide the code that you ran so that I can try to reproduce and diagnose the problem?

[bprager5]

Thanks, below is the code.

x <- SetupForWGCNA(
  x,
  gene_select = "fraction", 
  fraction = 0.05, 
  wgcna_name = "malig"

x <- MetacellsByGroups(
  seurat_obj = x,  
  reduction = 'harmony', 
  group.by = "sample", 
  k = 30, 
  min_cells=40,
  ident.group = 'sample' 
)

x <- NormalizeMetacells(x)
x <- ScaleMetacells(x, features=VariableFeatures(x))
x <- RunPCAMetacells(x, features=VariableFeatures(x))
x <- RunHarmonyMetacells(x, group.by.vars='sample')
x <- RunUMAPMetacells(x, reduction='harmony', dims=1:20)

[joebielanski]

By chance I'm running into the same problem today. Also getting the massive separation by sample when using other reductions. That said, I've found that tinkering with k and max_shared values have made the error come and go.

[crazyxiaoj]

I think you could try increasing the min_cells parameter to 60 or higher?

[bprager5]

Interesting, agreed, increasing min_cells to a certain threshold (in my dataset, 50) gets rid of the NULL error trying to reach x@misc$run_stats and lets me use the harmony reduction - thanks.