Error in FindClustersRecall when using both copula null methods

[Carolina-Toste]

Hi. Thank you so much for creating this method. I've been trying to implement it with my data and while running the following code:

recall_objs <- purrr::imap(seurat_objs, function(seurat, chip_name) {
  
  seurat <- FindClustersRecall(
    seurat,
    assay = "qumi_poilog_2.4",
    null_method = "ZIP-copula",
    resolution_start = 1.65,
    reduction_percentage = 0.05,
    dims = 1:30,
    verbose = FALSE
  )

  return(seurat)
  
}) 

I encountered the following error:

<error/purrr_error_indexed>
Error in `map2()`:
ℹ In index: 1.
ℹ With name: H2.
Caused by error in `$<-.data.frame`:
! replacement has 1 row, data has 1500

I tested both NB-copula and ZIP-copula which failed with the same error, but the other null methods worked.

Do you have any suggestions on what might be causing this or how to resolve it? I think there might be a bug in the creation of the sce object, but I'm not sure.

Thanks in advance!

[agdenadel]

Hi @Carolina-Toste, happy to try to fix this for you. Can you share the data that produces this error with me (privately is fine if this is unpublished data)?

[Carolina-Toste]

Hi . Thank you for getting back to me so quickly. I'm really sorry but I'm not authorised to share this data. I did however did a quick test with the pbmc3k dataset and it gave me the same error. Would this be enough to reproduce it on your end? Thank you in advance.

InstallData("pbmc3k")
pbmc3k <- SeuratData::LoadData("pbmc3k")
pbmc3k<- UpdateSeuratObject(pbmc3k)

pbmc3k<- pbmc3k %>% 
  NormalizeData(normalization.method = "LogNormalize", scale.factor = 10000) %>%
  FindVariableFeatures(selection.method = "vst", nfeatures = 2000) %>%
  ScaleData() %>%
  RunPCA(verbose = FALSE) %>%
  FindNeighbors(dims = 1:30, verbose = FALSE) %>%
  FindClusters(resolution = 0.5, verbose = FALSE) %>%
  RunUMAP(dims = 1:30, verbose = FALSE)

pbmc3k_recall <- FindClustersRecall(pbmc3k, null_method= "NB-copula", 
resolution_start = 2, reduction_percentage = 0.05, dims=1:30, verbose=FALSE)

Input Data Construction Start
Input Data Construction End
Start Marginal Fitting
Marginal Fitting End
Start Copula Fitting
Convert Residuals to Multivariate Gaussian
Converting End
Copula group 1 starts
Copula Fitting End
Start Parameter Extraction

Error in `$<-.data.frame`(`*tmp*`, "gene", value = list(PPBP = 0, S100A9 = 0, :
replacement has 1 row, data has 2000

[Carolina-Toste]

Hi.

After a bit of digging I found the issue in the extract_para() function of scdesign3() which is used inside your copula.R script.

The line data$gene <- count_mat[x, ] , returns a 1 row df instead of a vector causing the replacement error.

To fix it, I added the following to the top of your copula.R script:

trace("extract_para", where = asNamespace("scDesign3"), edit = TRUE)

After calling FindClustersRecall(), the editor opened and I changed line 23 of extract_para() to

data$gene <- as.numeric(count_mat[x, ]).

It's not the most elegant solution, but it's now working.