Why SCTransform converts the gene expression levels to integers or some categorized values?

[ElyasMo]

I am trying to analyze Alzheimer and healthy (in case of Alzheimer) human brain slices. After loading the data and normalizing the data through default options of SCTransform command, when I try to plot the the expression levels of some genes through vln command I see that the expression values are changed to integers or they are categorized in some defined levels and not a continues period of numbers.

I have rechecked this issue with the default mouse brain tutorial and I have got the same problem. Here is vln plot from the mouse brain dataset which is provided by Seurat as a tutorial. So it could be reproduceable for you as well.

brain<-Load10X_Spatial(Directory,filename = "filtered_feature_bc_matrix.h5",
                assay = "Spatial",
                slice = "slice1",
                filter.matrix = TRUE,
                to.upper = FALSE)
brain <- SCTransform(brain, assay = "Spatial", verbose = FALSE, do.scale=TRUE)
VlnPlot(brain, features = 'APP')

here is the result of my sessioninfo():

R version 4.0.2 (2020-06-22)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows >= 8 x64 (build 9200)

Matrix products: default

locale:
[1] LC_COLLATE=English_United States.1252  LC_CTYPE=English_United States.1252    LC_MONETARY=English_United States.1252 LC_NUMERIC=C                          
[5] LC_TIME=English_United States.1252    

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

other attached packages:
[1] sctransform_0.3.2         dplyr_1.0.2               patchwork_1.1.1           ggplot2_3.3.2             stxBrain.SeuratData_0.1.1 panc8.SeuratData_3.0.2   
[7] SeuratData_0.2.1          Seurat_3.2.3             

loaded via a namespace (and not attached):
  [1] nlme_3.1-148          matrixStats_0.57.0    bit64_4.0.5           RcppAnnoy_0.0.16      RColorBrewer_1.1-2    httr_1.4.2            tools_4.0.2          
  [8] R6_2.5.0              irlba_2.3.3           rpart_4.1-15          KernSmooth_2.23-17    uwot_0.1.10           mgcv_1.8-31           lazyeval_0.2.2       
 [15] colorspace_2.0-0      withr_2.3.0           tidyselect_1.1.0      gridExtra_2.3         bit_4.0.4             compiler_4.0.2        cli_2.2.0            
 [22] hdf5r_1.3.3           plotly_4.9.2.2        labeling_0.4.2        scales_1.1.1          lmtest_0.9-38         spatstat.data_1.7-0   ggridges_0.5.2       
 [29] pbapply_1.4-3         rappdirs_0.3.1        spatstat_1.64-1       goftest_1.2-2         stringr_1.4.0         digest_0.6.27         spatstat.utils_1.17-0
 [36] pkgconfig_2.0.3       htmltools_0.5.1.1     parallelly_1.22.0     fastmap_1.0.1         htmlwidgets_1.5.3     rlang_0.4.9           rstudioapi_0.13      
 [43] shiny_1.5.0           farver_2.0.3          generics_0.1.0        zoo_1.8-8             jsonlite_1.7.2        ica_1.0-2             magrittr_2.0.1       
 [50] Matrix_1.2-18         fansi_0.4.1           Rcpp_1.0.5            munsell_0.5.0         abind_1.4-5           reticulate_1.18       lifecycle_0.2.0      
 [57] stringi_1.5.3         yaml_2.2.1            MASS_7.3-51.6         Rtsne_0.15            plyr_1.8.6            grid_4.0.2            parallel_4.0.2       
 [64] listenv_0.8.0         promises_1.1.1        ggrepel_0.9.0         crayon_1.3.4          miniUI_0.1.1.1        deldir_0.2-3          lattice_0.20-41      
 [71] cowplot_1.1.0         splines_4.0.2         tensor_1.5            pillar_1.4.7          igraph_1.2.6          future.apply_1.6.0    reshape2_1.4.4       
 [78] codetools_0.2-16      leiden_0.3.6          glue_1.4.2            data.table_1.13.4     png_0.1-7             vctrs_0.3.6           httpuv_1.5.4         
 [85] gtable_0.3.0          RANN_2.6.1            purrr_0.3.4           polyclip_1.10-0       tidyr_1.1.2           assertthat_0.2.1      scattermore_0.7      
 [92] future_1.21.0         rsvd_1.0.3            mime_0.9              xtable_1.8-4          RSpectra_0.16-0       later_1.1.0.1         survival_3.1-12      
 [99] viridisLite_0.3.0     tibble_3.0.4          cluster_2.1.0         globals_0.14.0        fitdistrplus_1.1-3    ellipsis_0.3.1        ROCR_1.0-11          . -->

[ChristophH]

By default VlnPlot pulls the expression values from the data slot. See the documentation here. Citing from the SCTransform documentation:

Returns a Seurat object with a new assay (named SCT by default) with counts being (corrected) counts, data being log1p(counts), scale.data being pearson residuals

So in your example above you are plotting the log1p-transformed corrected (discrete) counts. To plot the Pearson residuals set slot = "scale.data" in VlnPlot