NA/NaN/Inf in foreign function call

[connerlambden]

Hi, when plotting with plot_cpdb3, I'm getting the following error. I've double checked and made sure that my sce (SingleCellExperiment) object has no nans or infs. All colsums/rowsums are > 1.

p <- plot_cpdb3(cell_type1 = 'seurat_cluster_0', cell_type2 = 'seurat_cluster_1', scdata = sce, idents = 'seurat_clusters_names', means = means, pvals = pvals, keep_significant_only = TRUE, standard_scale = TRUE, remove_self = TRUE ) p

Gives the following error:

Error in hclust(d): NA/NaN/Inf in foreign function call (arg 10) Traceback:

1. plot_cpdb3(cell_type1 = "seurat_cluster_0", cell_type2 = "seurat_cluster_1", . scdata = sce, idents = "seurat_clusters_names", means = means, . pvals = pvals, keep_significant_only = TRUE, standard_scale = TRUE, . remove_self = TRUE)
2. plot_cpdb(cell_type1 = cell_type1, cell_type2 = cell_type2, scdata = scdata, . idents = idents, split.by = split.by, means = means, pvals = pvals, . keep_significant_only = keep_significant_only, standard_scale = standard_scale, . return_table = TRUE, degs_analysis = degs_analysis, ...)
3. hclust(d)

Any ideas?

Additionally, when I try to make a dotplot using the following:

plot_cpdb(cell_type1 = "seurat_cluster_0", cell_type2 = "seurat_cluster_1", scdata = sce, idents = 'seurat_clusters_names', means = means, pvals = pvals, genes = genes_to_plot)

I get the following error:

Warning message in rep(yes, length.out = len): “'x' is NULL so the result will be NULL” ERROR while rich displaying an object: Error in ans[ypos] <- rep(yes, length.out = len)[ypos]: replacement has length zero

Thanks for the help!

[zktuong]

Hi @connerlambden, does plot_cpdb work with keep_significant_only = FALSE? Also are your files from deg_analysis mode in CellPhoneDB? If so, you may need to add the option deg_analysis =TRUE.

[connerlambden]

@zktuong thanks for the help! This is not deg_analysis.

For plot_cpdb, when I set keep_significant_only = TRUE, I get the following error:

Error in which(unlist(anysig)): argument to 'which' is not logical Traceback:

1. plot_cpdb(cell_type1 = "seurat_cluster_0", cell_type2 = "seurat_cluster_1", . scdata = sce, idents = "seurat_clusters_names", means = means, . pvals = pvals, keep_significant_only = TRUE, genes = genes_to_plot)
2. which(unlist(anysig))
3. which(unlist(anysig))

and when I set keep_significant_only = FALSE I get the same error as before (“'x' is NULL so the result will be NULL”)

I was able to get plot_cpdb3 to work!

However, plot_cpdb2 is giving the error:

Error in h(simpleError(msg, call)): error in evaluating the argument 'i' in selecting a method for function '[': error in evaluating the argument 'x' in selecting a method for function 'which': subscript contains invalid names Traceback:

1. plot_cpdb2(cell_type1 = "seurat_cluster_0", cell_type2 = "seurat_cluster_1", . scdata = sce, idents = "seurat_clusters_names", means = means, . pvals = pvals, deconvoluted = decon, desiredInteractions = list(c("seurat_cluster_0", . "seurat_cluster_1"), c("seurat_cluster_1", "seurat_cluster_0")), . interaction_grouping = interaction_annotation, edge_group_colors = c(Activating = "#e15759", . Chemotaxis = "#59a14f", Inhibitory = "#4e79a7", Intracellular trafficking = "#9c755f", . DC_development = "#B07aa1", Unknown = "#e7e7e7"), node_group_colors = c(seurat_cluster_0 = "red", . seurat_cluster_1 = "blue"), keep_significant_only = FALSE, . standard_scale = TRUE, remove_self = TRUE)

And plot_cpdb4 gives the following warning message and then returns <NA>: Warning message: “Since you have set group, you should better set grid.col as a named vector where sector names are the vector names (should contain all sectors).” Warning message in structure(rand_color(length(unnamed_grid)), names = unnamed_grid): “restarting interrupted promise evaluation” Warning message in structure(rand_color(length(unnamed_grid)), names = unnamed_grid): “internal error -3 in R_decompress1”

Thanks.

[zktuong]

So there's a couple of issues here.

1) plot_cpdb

not sure what's going on, but your command is causing the internal functions to not generate a proper dataframe. is genes = genes_to_plot malformed?

2) plot_cpdb2 is interaction_grouping = interaction_annotation and edge_group_colors = c(Activating = "#e15759", . Chemotaxis = "#59a14f", Inhibitory = "#4e79a7", Intracellular trafficking = "#9c755f", . DC_development = "#B07aa1", Unknown = "#e7e7e7")

directly from the README? You should create one that is more complete for your dataset as my dataframe is only a subset of interactions.

[saeedfc]

Had same error with plot_cpdb, plot_cpdb2 etc,

cdb <- plot_cpdb(cell_type1 = celltype1, cell_type2 = celltype2,
              scdata = SOT_sc,
              idents = 'Fine_Annotation', # column name where the cell ids are located in the metadata
              means = means,
              pvals = pvals
Error in hclust(d) : NA/NaN/Inf in foreign function call (arg 10)

Apparently if I block the part where hierarchcial clustering is applied on the distance metrix; it works fine. SO i think the cluster_rows = T being default in plot_cpdb, this could carry over to plot_cpdb2 or 3 as well as they use plot_cpdb internally.

 # if (nrow(means_mat) > 2) {
  #   d <- dist(as.data.frame(means_mat))
  #   h <- hclust(d)
  #   means_mat <- means_mat[h$order, , drop = FALSE]
  #   pvals_mat <- pvals_mat[h$order, , drop = FALSE]

I retrieved the distance matrix;

a part of it looks like this ; I guess entire vectors with NA is the issue. Maybe including a filter on NA and warnign on which interaction is NA and hence being removed, could solve the issue

CXCL12-CXCR3                                                                                             
OSMR-OSM                                                                                                 
TGFB1-AR                                                                                                 
TGFB1-integrin-aVb6-complex                                                                              
ACKR3-CXCL12                                                                                             
IFNG-Type-II-IFNR                                                                                        
DPP4-CXCL12                                                                                              
TNF-TNFRSF1B                                                                                             
CXCL12-CXCR4                                                                                             
TGFB1-TGFbeta-receptor2                                                                                  
EGFR-TGFB1                                                                                               
TNF-TNFRSF1A                                                                                             
TGFB1-VDR                                                                                                
TGFB1-TGFBR3                                                                                             
TGFB1-TGFbeta-receptor1       0.15030635                                                                 
TNFRSF1B-GRN                  1.76503428              1.36840589                                         
TNFRSF1A-GRN                  0.31918960              0.60351636   1.07538970                            
TNFRSF1A-FASLG                        NA                      NA           NA           NA               
TNFSF13-TNFRSF1A                      NA                      NA           NA           NA             NA

[zktuong]

thanks @saeedfc

I think before adding the filter, i would like to first find out what's causing those NAs - are these with 0 interactions across the cellphonedb output?? By default cellphonedb would remove that prior to generating the output but I suppose it can still happen if you (un)luckily to land on celltypes-celltypes combinations that happened to not express the ligand receptors.

So, if you expand celltype1/2 to have more celltypes, does the error go away?

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.