Assay in `SingleCellExperiment` not being transposed

[ekageyama]

Hello, I have a SingleCellExperiment of 3 samples and 3 genes, when trying to convert it to annData i get the following:

> sce
class: SingleCellExperiment
dim: 3 3
metadata(1): .internal
assays(2): counts X
rownames(3): 3xLinker MIR1302-2HG FAM138A
rowData names(1): Symbol
colnames(3): SAM24451293_rep1-AAACCCAAGAAATCCA
  SAM24451293_rep1-AAACCCAAGACAGTCG SAM24451293_rep1-AAACCCACAGACCCGT
colData names(2): Sample Barcode 
reducedDimNames(0):
mainExpName: NULL
altExpNames(0):
> as_AnnData(sce)
Error in `private$.validate_aligned_mapping()`:
! `rownames(layers[['counts']])` is not as expected
ℹ Expected row names: "SAM24451293_rep1-AAACCCAAGAAATCCA", "SAM24451293_rep1-AAACCCAAGACAGTCG", "SAM24451293_rep1-AAACCCACAGACCCGT"
ℹ Provided row names: "3xLinker", "MIR1302-2HG", "FAM138A"

Not sure why the assay is not being transposed, or why it infers it shouldnt be transposed

[lazappi]

This may be an issue with the check rather than the matrix itself. Could you please share a minimal version of the dataset that reproduces the issue? It's fine to replaces/remove any data you want to keep private as long as you still see the error.

It would also be useful to know what version of {anndataR} you are using.

[ekageyama]

sorry for the late reply, so the reason i assume is a transposition problem is that I only get an error witha. square matrix, but with a rectangular dimensions dont match.

> sce<-readRDS("sce_trans-prob.rds")
> sce
Loading required namespace: SingleCellExperiment
class: SingleCellExperiment
dim: 3 2
metadata(3): .internal source cropflow
assays(1): counts
rownames(3): ENSG00000237491 ENSG00000228794 ENSG00000188976
rowData names(11): source gene_version ... Symbol Type
colnames(2): SAM24480943-ACCGAAAGTGCGGAAT_50461
  SAM24480943-ACGGACCCATTAGTGA_50913
colData names(24): original_cell_id perturbation ... Sunrise_outs
  Cells_loaded
> library(anndataR)
> as_AnnData(sce)
Error in `private$.validate_aligned_mapping()`:
! Unexpected shape for layers[['counts']]
ℹ Expected [2, 3], got [3, 2]
> sessionInfo()
R version 4.5.1 (2025-06-13)
Platform: aarch64-apple-darwin20
Running under: macOS Sonoma 14.7.4

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] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

time zone: America/Los_Angeles
tzcode source: internal

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

other attached packages:
[1] anndataR_1.0.0

loaded via a namespace (and not attached):
 [1] vctrs_0.6.5                 cli_3.6.5
 [3] rlang_1.1.6                 png_0.1-8
 [5] purrr_1.2.0                 generics_0.1.4
 [7] jsonlite_2.0.0              DelayedArray_0.36.0
 [9] glue_1.8.0                  SummarizedExperiment_1.40.0
[11] S4Vectors_0.48.0            stats4_4.5.1
[13] MatrixGenerics_1.22.0       Biobase_2.70.0
[15] grid_4.5.1                  Seqinfo_1.0.0
[17] abind_1.4-8                 SingleCellExperiment_1.32.0
[19] IRanges_2.44.0              lifecycle_1.0.4
[21] compiler_4.5.1              Rcpp_1.1.0
[23] XVector_0.50.0              lattice_0.22-7
[25] R6_2.6.1                    reticulate_1.44.1
[27] SparseArray_1.10.4          pillar_1.11.1
[29] GenomicRanges_1.62.0        magrittr_2.0.4
[31] Matrix_1.7-4                tools_4.5.1
[33] matrixStats_1.5.0           S4Arrays_1.10.1
[35] BiocGenerics_0.56.0

[sce_trans-prob.rds.zip](https://github.com/user-attachments/files/23971027/sce_trans-prob.rds.zip)

[lazappi]

Thanks for the update. I can't seem to download the example file though, could you please try uploading it again?

Could you also try running this example:

counts <- matrix(1:6, nrow = 2, ncol = 3)
counts
#>      [,1] [,2] [,3]
#> [1,]    1    3    5
#> [2,]    2    4    6
sce <- SingleCellExperiment::SingleCellExperiment(assays = list(counts = counts))
sce
#> class: SingleCellExperiment 
#> dim: 2 3 
#> metadata(0):
#> assays(1): counts
#> rownames: NULL
#> rowData names(0):
#> colnames: NULL
#> colData names(0):
#> reducedDimNames(0):
#> mainExpName: NULL
#> altExpNames(0):
as_AnnData(sce)
#> InMemoryAnnData object with n_obs × n_vars = 3 × 2
#>     layers: 'counts'

[ekageyama]

That example works fine on my end, the object is saved with saveRDS so load will give an error but readRDS should work

sce_trans-prob.rds.zip

[myushen]

I had the same transpose error for a delayed matrix object. Change to matrix worked. Might be matrix type check issue?

counts = x |> counts()
sce <- SingleCellExperiment::SingleCellExperiment(assays = list(counts = counts))
counts |> class()
[1] "DelayedMatrix"
attr(,"package")
[1] "DelayedArray"
write_h5ad(sce, "~/scratch/cache_temp/data.h5ad", compression = "gzip", mode = "w")
Error in `private$.validate_aligned_mapping()`:
! Unexpected shape for layers[['counts']]
ℹ Expected [6, 59236], got [59236, 6]
Run `rlang::last_trace()` to see where the error occurred.

counts2 = x |> counts() |> as.matrix()
sce2 <- SingleCellExperiment::SingleCellExperiment(assays = list(counts = counts2))
write_h5ad(sce2, "~/scratch/cache_temp/data2.h5ad", compression = "gzip", mode = "w")

[lazappi]

Thanks for the example data @ekageyama! That helped me work out the issue. You have a SVT_SparseMatrix object in counts() which we don't currently support but plan to in the future. @myushen The same happens with the DelayedArray.

As a temporary improvement until we support these I have opened a PR that adds a warning so at least the user should see what the issue is.

as_AnnData(sce)
# Error in `private$.validate_aligned_mapping()` at anndataR/R/InMemoryAnnData.R:80:9:
# ! Unexpected shape for layers[['counts']]
# ℹ Expected [2, 3], got [3, 2]
# Run `rlang::last_trace()` to see where the error occurred.
# Warning message:
# Converting unknown object of class <SVT_SparseMatrix> from <SingleCellExperiment>
# ! It will be returned as is but this may result in other errors
# ℹ This may be solved by converting to a <matrix>/<Matrix>, <data.frame>/<DataFrame>, or
#  <list>/.<SimpleList> 

[HelenaLC]

Not 100% sure whether this is related, but I was having this error working with HDF5Array ... converting to dgCMatrix did the trick, however, in working up a reproducible example I am now really confused, as assayNames seems to have an effect (independent of being delayed or not). See here:

suppressPackageStartupMessages({
    library(anndataR)
    library(HDF5Array)
    library(SingleCellExperiment)
})

mx <- array(runif(20), c(4, 5))
mx <- as(mx, "dgCMatrix")

# this works...
# (unnamed list as assays)

se <- SingleCellExperiment(assays=list(mx))
as_AnnData(se)
# InMemoryAnnData object with n_obs × n_vars = 5 × 4

h5 <- tempfile()
saveHDF5SummarizedExperiment(se, h5)
sf <- loadHDF5SummarizedExperiment(h5)
as_AnnData(sf)
# InMemoryAnnData object with n_obs × n_vars = 5 × 4

# ...this doesn't?
# (named list as assays)

se <- SingleCellExperiment(assays=list(counts=mx))
as_AnnData(se)
# InMemoryAnnData object with n_obs × n_vars = 5 × 4
#     layers: 'counts'

h5 <- tempfile()
saveHDF5SummarizedExperiment(se, h5)
sf <- loadHDF5SummarizedExperiment(h5)
as_AnnData(sf)
# Error in `private$.validate_aligned_mapping()`:
# ! Unexpected shape for layers[['counts']]
# ℹ Expected [5, 4], got [4, 5]

[Artur-man]

I think the getters of the package are not DelayedArray compatible yet. See here: #387