split_collect = TRUE fails to produce a unique legend scale for some genes in FeaturePlot_scCustom()

[CroixJeremy2]

Hello,

I have spotted a small bug: split_collect = TRUE fails to produce a unique legend scale for some genes in FeaturePlot_scCustom().

For some reason, for some genes, the legend can be unified, so I suspect something related to ranges of expression between the groups defined by the split.by. Alternatively, maybe the issue is related to recent ggplot2 updates that might have happened? In any case, can you correct the function and update the github dev branch please? Thanks in advance (I really like your package).

(Problem linked to #94 )

Here is a minimal example:

#Required libraries
	library(Seurat)
	library(scCustomize)
	library(SeuratData)
	# SeuratData::InstallData("pbmc3k")
	# SeuratData::InstallData("ifnb")

#Minimal example using pbmc3k dataset
	#Loading
		data(pbmc3k)
		pbmc3k = UpdateSeuratObject(object = pbmc3k)
		pbmc3k = Convert_Assay(seurat_object = pbmc3k, convert_to = "V5")

	#Quick pipeline
		pbmc3k = NormalizeData(pbmc3k)
		pbmc3k = FindVariableFeatures(pbmc3k)
		pbmc3k = ScaleData(pbmc3k)
		pbmc3k = RunPCA(pbmc3k)
		pbmc3k = RunUMAP(pbmc3k, dims = 1:30)

	#Graph that produces splitted legends despite asking for unified legend
		FeaturePlot_scCustom(seurat_object = pbmc3k,
							features = "GAPDH",
							na_cutoff = NA,
							split.by = "seurat_annotations",
							num_columns = 3,
							split_collect = TRUE)

---

	#Unsuccessful test with combine = FALSE
		FeaturePlot_scCustom(seurat_object = pbmc3k,
							features = "GAPDH",
							na_cutoff = NA,
							split.by = "seurat_annotations",
							num_columns = 3,
							split_collect = TRUE,
							combine = FALSE)

---

	#Graph with split_collect = FALSE, which works as intended
		FeaturePlot_scCustom(seurat_object = pbmc3k,
							features = "GAPDH",
							na_cutoff = NA,
							split.by = "seurat_annotations",
							num_columns = 3,
							split_collect = FALSE)

---

	#For some reasons, it works for some other genes
		FeaturePlot_scCustom(seurat_object = pbmc3k,
							features = "IL6",
							na_cutoff = NA,
							split.by = "seurat_annotations",
							num_columns = 3,
							split_collect = TRUE)

---

sessionInfo() output

> sessionInfo()
R version 4.5.3 (2026-03-11)
Platform: aarch64-apple-darwin20
Running under: macOS Tahoe 26.5

Matrix products: default
BLAS:   /Library/Frameworks/R.framework/Versions/4.5-arm64/Resources/lib/libRblas.0.dylib 
LAPACK: /Library/Frameworks/R.framework/Versions/4.5-arm64/Resources/lib/libRlapack.dylib;  LAPACK version 3.12.1

locale:
[1] C.UTF-8/C.UTF-8/C.UTF-8/C/C.UTF-8/C.UTF-8

time zone: Europe/Paris
tzcode source: internal

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] pbmc3k.SeuratData_3.1.4 ifnb.SeuratData_3.0.0   SeuratData_0.2.2.9002  
[4] scCustomize_3.2.4.9012  Seurat_5.5.0            SeuratObject_5.4.0     
[7] sp_2.2-1               

loaded via a namespace (and not attached):
  [1] RColorBrewer_1.1-3     rstudioapi_0.18.0      jsonlite_2.0.0        
  [4] shape_1.4.6.1          magrittr_2.0.5         spatstat.utils_3.2-3  
  [7] ggbeeswarm_0.7.3       farver_2.1.2           GlobalOptions_0.1.4   
 [10] vctrs_0.7.3            ROCR_1.0-12            spatstat.explore_3.8-0
 [13] paletteer_1.7.0        janitor_2.2.1          htmltools_0.5.9       
 [16] forcats_1.0.1          sctransform_0.4.3      parallelly_1.47.0     
 [19] KernSmooth_2.23-26     htmlwidgets_1.6.4      ica_1.0-3             
 [22] plyr_1.8.9             plotly_4.12.0          zoo_1.8-15            
 [25] lubridate_1.9.5        igraph_2.3.1           mime_0.13             
 [28] lifecycle_1.0.5        pkgconfig_2.0.3        Matrix_1.7-4          
 [31] R6_2.6.1               fastmap_1.2.0          fitdistrplus_1.2-6    
 [34] future_1.70.0          shiny_1.13.0           snakecase_0.11.1      
 [37] digest_0.6.39          colorspace_2.1-2       rematch2_2.1.2        
 [40] patchwork_1.3.2        tensor_1.5.1           RSpectra_0.16-2       
 [43] irlba_2.3.7            labeling_0.4.3         progressr_0.19.0      
 [46] spatstat.sparse_3.1-0  timechange_0.4.0       httr_1.4.8            
 [49] polyclip_1.10-7        abind_1.4-8            compiler_4.5.3        
 [52] withr_3.0.2            S7_0.2.2               fastDummies_1.7.6     
 [55] MASS_7.3-65            rappdirs_0.3.4         tools_4.5.3           
 [58] vipor_0.4.7            lmtest_0.9-40          otel_0.2.0            
 [61] beeswarm_0.4.0         httpuv_1.6.17          future.apply_1.20.2   
 [64] goftest_1.2-3          glue_1.8.1             nlme_3.1-168          
 [67] promises_1.5.0         grid_4.5.3             Rtsne_0.17            
 [70] cluster_2.1.8.2        reshape2_1.4.5         generics_0.1.4        
 [73] gtable_0.3.6           spatstat.data_3.1-9    tidyr_1.3.2           
 [76] data.table_1.18.4      spatstat.geom_3.7-3    RcppAnnoy_0.0.23      
 [79] ggrepel_0.9.8          RANN_2.6.2             pillar_1.11.1         
 [82] stringr_1.6.0          spam_2.11-3            RcppHNSW_0.6.0        
 [85] ggprism_1.0.7          later_1.4.8            circlize_0.4.18       
 [88] splines_4.5.3          dplyr_1.2.1            lattice_0.22-9        
 [91] survival_3.8-6         deldir_2.0-4           tidyselect_1.2.1      
 [94] miniUI_0.1.2           pbapply_1.7-4          gridExtra_2.3         
 [97] scattermore_1.2        matrixStats_1.5.0      stringi_1.8.7         
[100] lazyeval_0.2.3         codetools_0.2-20       tibble_3.3.1          
[103] cli_3.6.6              uwot_0.2.4             xtable_1.8-8          
[106] reticulate_1.46.0      Rcpp_1.1.1-1.1         globals_0.19.1        
[109] spatstat.random_3.4-5  mcprogress_0.1.1       png_0.1-9             
[112] ggrastr_1.0.2          spatstat.univar_3.2-0  parallel_4.5.3        
[115] ggplot2_4.0.3          dotCall64_1.2          listenv_0.10.1        
[118] viridisLite_0.4.3      scales_1.4.0           ggridges_0.5.7        
[121] purrr_1.2.2            crayon_1.5.3           rlang_1.2.0           
[124] cowplot_1.2.0         
> 

[CroixJeremy2]

Current workaround:

		FeaturePlot_scCustom(seurat_object = pbmc3k,
							features = "GAPDH",
							na_cutoff = NA,
							split.by = "seurat_annotations",
							num_columns = 3,
							split_collect = TRUE) &
			scale_color_gradientn(colours = viridis_plasma_dark_high,
									limits = range(FetchData(object = pbmc3k, vars = "GAPDH")$GAPDH),
									name = "GAPDH")

[samuel-marsh]

Hi @CroixJeremy2,

Thanks for posting issue I’ll take a look. Can you confirm for me that it is still issue with current release 3.3.0 (Updated on CRAN week and half ago). Dev branch is currently behind. If issue remains just let me know!

Thanks!
Sam

[CroixJeremy2]

After installation from CRAN 1min ago using install.packages("scCustomize"), the problem is still persistent unfortunately...

[samuel-marsh]

@CroixJeremy2 ok thanks for checking I'll dig into it!

[samuel-marsh]

Ok so I have isolated the issue to the use of na_cutoff = NA. By setting that parameter to numerical value the plot returns as it should. If you are just plotting gene expression you can leave na_cutoff alone because the default value is 1e-09 and there aren't any negative values in gene expression space. However, it does cause problems when plotting PC loadings or module scores which have/can have negative values and you want to plot the whole range.

I will work on fix but in meantime if plotting gene expression you can leave na_cutoff alone and if plotting PC loading or module score a slightly simpler workaround that what you posted above would be to set
na_cutoff = min(FetchData(obj, "SCORE_NAME")

Best,
Sam

edit: Unclear to me currently why it fails for some genes and not others but I think it's something funky happening when setting NA.

[CroixJeremy2]

Ah! I actually encountered this problem initially when I wanted to plot a score that was calculated using AddModuleScore(), so I really needed to use na_cutoff = NA because of negative values, hence the fact that I have let this parameter inside my code on this issue I guess (but it is true that it is not a "minimal example" with this parameter; a welcomed mistake from my part I guess). In any case, thanks for the workaround, I find your way slightly more elegant than my previous workaround.

Thank you in advance for elaborating a better fix in the future (it is not an important issue at the moment, just slightly "annoying", but there are easy and different workarounds, so take your time, no rush). Thanks for having checked the issue quickly.

PS: I let the issue open for the moment.