BUG: DimPlot ignores the reduction parameter if multiple reductions have the same key

[nigiord]

Issue Description

When using DimPlot on a SeuratObject with multiple reductions that have the same key (for instance, PC_), the reduction parameter is ignored and the last inserted reduction is always the one showed.

Reproducing Code Example

library(Seurat)

genes <- paste0("Gene", 1:10)
counts <- matrix(rpois(40, lambda = 10), ncol = 4)
rownames(counts) <- genes
colnames(counts) <- paste0("cell", 1:4)

dummyobj <- CreateSeuratObject(counts = counts)

pcs <- prcomp(x = (dummyobj[["RNA"]]$counts))
rev_pcs <- prcomp(x = t(dummyobj[["RNA"]]$counts))

pca <- CreateDimReducObject(
  embeddings = pcs$rotation,
  loadings = pcs$x,
  stdev = pcs$sdev,
  key = "PC_",
  assay = "RNA"
)

rev_pca <- CreateDimReducObject(
  embeddings = rev_pcs$x,
  loadings = rev_pcs$rotation,
  stdev = rev_pcs$sdev,
  key = "PC_",  ## same key as above
  assay = "RNA"
)

dummyobj@reductions[["PCA"]] <- pca
dummyobj@reductions[["PCA_REV"]] <- rev_pca

DimPlot(dummyobj, reduction = "PCA")
DimPlot(dummyobj, reduction = "PCA_REV") # Plot exactly the same thing

Error Message

No error message, but only PCA_REV is plotted whichever reduction you choose, even if the reductions are about different assays.

Additional Comments

Bug comes from the use of FetchData there:

seurat/R/visualization.R

Lines 909 to 914 in 9354a78

	data <- FetchData(
	object = object,
	vars = c(dims, group.by),
	cells = cells,
	clean = 'project'
	)

The function does not take into account the reduction, only the name of the key, so it does not work as expected when you have multiple reductions with the same key.

Session Info

R version 4.3.3 (2024-02-29)
Platform: x86_64-apple-darwin13.4.0 (64-bit)
Running under: macOS Big Sur ... 10.16

Matrix products: default
BLAS/LAPACK: /Users/giordano-n/micromamba/envs/muscadet/lib/libopenblasp-r0.
3.29.dylib;  LAPACK version 3.12.0

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

time zone: Europe/Paris
tzcode source: internal

attached base packages:
[1] stats4    stats     graphics  grDevices utils     datasets  methods  
[8] base     

other attached packages:
 [1] glue_1.8.0                  stringr_1.5.1              
 [3] dplyr_1.1.4                 Seurat_5.1.0               
 [5] SeuratObject_5.0.2          sp_2.2-0                   
 [7] SummarizedExperiment_1.32.0 Biobase_2.62.0             
 [9] GenomicRanges_1.54.1        GenomeInfoDb_1.38.1        
[11] IRanges_2.36.0              S4Vectors_0.40.2           
[13] BiocGenerics_0.48.1         MatrixGenerics_1.14.0      
[15] matrixStats_1.2.0           muscadet_0.1.0             
[17] logger_0.2.2                nvimcom_0.9-167            

loaded via a namespace (and not attached):
  [1] RcppAnnoy_0.0.22         splines_4.3.3           
  [3] later_1.4.1              prettyGraphs_2.1.6      
  [5] bitops_1.0-9             tibble_3.2.1            
  [7] polyclip_1.10-7          fastDummies_1.7.5       
  [9] lifecycle_1.0.4          fastcluster_1.2.6       
 [11] doParallel_1.0.17        globals_0.16.3          
 [13] lattice_0.22-6           prabclus_2.3-4          
 [15] MASS_7.3-60.0.1          dendextend_1.17.1       
 [17] magrittr_2.0.3           plotly_4.10.4           
 [19] httpuv_1.6.15            sctransform_0.4.1       
 [21] spam_2.11-1              flexmix_2.3-20          
 [23] spatstat.sparse_3.1-0    reticulate_1.41.0.1     
 [25] cowplot_1.1.3            pbapply_1.7-2           
 [27] RColorBrewer_1.1-3       ade4_1.7-23             
 [29] facets_0.6.2             abind_1.4-5             
 [31] zlibbioc_1.48.0          Rtsne_0.17              
 [33] purrr_1.0.4              RCurl_1.98-1.16         
 [35] nnet_7.3-20              circlize_0.4.16         
 [37] GenomeInfoDbData_1.2.11  ggrepel_0.9.6           
 [39] irlba_2.3.5.1            listenv_0.9.1           
 [41] spatstat.utils_3.1-3     goftest_1.2-3           
 [43] RSpectra_0.16-2          spatstat.random_3.3-2   
 [45] fitdistrplus_1.2-2       parallelly_1.42.0       
 [47] leiden_0.4.3.1           codetools_0.2-20        
 [49] DelayedArray_0.28.0      tidyselect_1.2.1        
 [51] shape_1.4.6.1            farver_2.1.2            
 [53] viridis_0.6.5            spatstat.explore_3.3-4  
 [55] jsonlite_1.9.1           GetoptLong_1.0.5        
 [57] e1071_1.7-16             progressr_0.15.1        
 [59] ggridges_0.5.6           survival_3.8-3          
 [61] iterators_1.0.14         foreach_1.5.2           
 [63] tools_4.3.3              ica_1.0-3               
 [65] Rcpp_1.0.14              gridExtra_2.3           
 [67] SparseArray_1.2.2        withr_3.0.2             
 [69] alluvial_0.1-2           fastmap_1.2.0           
 [71] SNFtool_2.3.1            caTools_1.18.2          
 [73] digest_0.6.37            R6_2.6.1                
 [75] mime_0.13                colorspace_2.1-1        
 [77] scattermore_1.2          tensor_1.5              
 [79] gtools_3.9.5             spatstat.data_3.1-6     
 [81] diptest_0.77-1           tidyr_1.3.1             
 [83] generics_0.1.3           data.table_1.14.10      
 [85] robustbase_0.99-4-1      class_7.3-23            
 [87] httr_1.4.7               htmlwidgets_1.6.4       
 [89] S4Arrays_1.2.0           uwot_0.2.3              
 [91] pkgconfig_2.0.3          gtable_0.3.6            
 [93] modeltools_0.2-23        ComplexHeatmap_2.18.0   
 [95] lmtest_0.9-40            XVector_0.42.0          
 [97] htmltools_0.5.8.1        dotCall64_1.2           
 [99] clusterSim_0.51-5        clue_0.3-66             
[101] scales_1.3.0             png_0.1-8               
[103] spatstat.univar_3.1-2    reshape2_1.4.4          
[105] rjson_0.2.23             nlme_3.1-167            
[107] proxy_0.4-27             zoo_1.8-13              
[109] GlobalOptions_0.1.2      KernSmooth_2.23-26      
[111] parallel_4.3.3           miniUI_0.1.1.1          
[113] ExPosition_2.8.23        RcppZiggurat_0.1.6      
[115] pillar_1.10.1            grid_4.3.3              
[117] vctrs_0.6.5              RANN_2.6.2              
[119] promises_1.3.2           xtable_1.8-4            
[121] cluster_2.1.8.1          cli_3.6.4               
[123] compiler_4.3.3           rlang_1.1.5             
[125] crayon_1.5.3             pctGCdata_0.3.0         
[127] future.apply_1.11.3      labeling_0.4.3          
[129] mclust_6.1.1             plyr_1.8.9              
[131] stringi_1.8.4            deldir_2.0-4            
[133] viridisLite_0.4.2        munsell_0.5.1           
[135] lazyeval_0.2.2           spatstat.geom_3.3-5     
[137] Matrix_1.6-4             RcppHNSW_0.6.0          
[139] patchwork_1.3.0          sparseMatrixStats_1.14.0
[141] future_1.34.0            ggplot2_3.5.1           
[143] fpc_2.2-13               shiny_1.10.0            
[145] kernlab_0.9-33           ROCR_1.0-11             
[147] Rfast_2.1.0              igraph_1.5.1            
[149] RcppParallel_5.1.9       DEoptimR_1.1-3-1

[igrabski]

Hi, thanks for your question! We intend different keys to be used for different reductions, so we recommend avoiding the situation of trying to use the same key for multiple reductions.

[nigiord]

Hi @igrabski,
Thank you for your response. In this case, I believe it would be better to enforce this behavior when a user creates a new reduction—either by throwing an error or renaming the key with a warning. The current implementation can lead to unexpected behavior and misinterpretation. For example, I spent quite some time wondering why my plot looked exactly the same; it would have been easy to overlook and continue the analysis with incorrect assumptions.

Cheers,
−Nils