hdf5_adata$as_Seurat fails when features contain `_`

[doliv071]

Description

I'm trying to convert an anndata h5ad to Seurat object via read_h5ad. The object contains feature names with _ in them e.g. gene_1, gene_2, etc. This apparently is not allowed in the seurat object: Warning: Feature names cannot have underscores ('_'), replacing with dashes ('-') ... however, Seurat does allow underscore in the assay rownames?!?! The result is that as_Seurat fails because the sanity check indicates some insanity Error: No feature overlap between existing object and new layer data

The underlying object is okay, because the conversion to SingleCellExperiment works just fine.

> system.file("ext/", "demo_data.h5ad", package = "rcna") |> 
+     read_h5ad(as = "Seurat")
Warning: Feature names cannot have underscores ('_'), replacing with dashes ('-')
Warning: Data is of class matrix. Coercing to dgCMatrix.
Error: No feature overlap between existing object and new layer data
In addition: Warning message:
No "counts" or "data" layer found in `names(layers_mapping)`, this may lead to unexpected
results when using the resulting <Seurat> object. 
> traceback()
9: stop("No feature overlap between existing object and new layer data", 
       call. = FALSE)
8: `LayerData<-.Assay5`(object = `*tmp*`, layer = layer, ..., value = value)
7: `LayerData<-`(object = `*tmp*`, layer = layer, ..., value = value)
6: `LayerData<-.Seurat`(`*tmp*`, layer = dummy_counts, value = counts)
5: SeuratObject::`LayerData<-`(`*tmp*`, layer = dummy_counts, value = counts)
4: .as_Seurat_create_object_with_layers(adata, layers_mapping, object_metadata, 
       assay_name)
3: as_Seurat(self, assay_name = assay_name, x_mapping = x_mapping, 
       layers_mapping = layers_mapping, object_metadata_mapping = object_metadata_mapping, 
       assay_metadata_mapping = assay_metadata_mapping, reduction_mapping = reduction_mapping, 
       graph_mapping = graph_mapping, misc_mapping = misc_mapping)
2: hdf5_adata$as_Seurat(...)
1: read_h5ad(system.file("ext/", "demo_data.h5ad", package = "rcna"), 
       as = "Seurat")

Minimal Reproducible Example

Session Information

Session info

``` > sessionInfo() R version 4.5.0 (2025-04-11) Platform: x86_64-pc-linux-gnu Running under: Ubuntu 24.04.4 LTS

Matrix products: default
BLAS/LAPACK: /usr/lib/x86_64-linux-gnu/openblas-pthread/libopenblasp-r0.3.26.so; LAPACK version 3.12.0

locale:
[1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C LC_TIME=en_US.UTF-8
[4] LC_COLLATE=en_US.UTF-8 LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
[7] LC_PAPER=en_US.UTF-8 LC_NAME=C LC_ADDRESS=C
[10] LC_TELEPHONE=C LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C

time zone: America/New_York
tzcode source: system (glibc)

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

other attached packages:
[1] ggbeeswarm_0.7.3 anndataR_1.0.2 Seurat_5.5.1 SeuratObject_5.4.0
[5] sp_2.2-1 ggplot2_4.0.2 dplyr_1.2.1 rcna_0.1.0

loaded via a namespace (and not attached):
[1] RColorBrewer_1.1-3 rstudioapi_0.18.0 jsonlite_2.0.0
[4] magrittr_2.0.5 spatstat.utils_3.2-2 farver_2.1.2
[7] fs_2.0.1 vctrs_0.7.3 ROCR_1.0-12
[10] memoise_2.0.1 spatstat.explore_3.8-0 S4Arrays_1.10.1
[13] htmltools_0.5.9 usethis_3.2.1 Rhdf5lib_1.32.0
[16] SparseArray_1.10.10 rhdf5_2.54.1 sctransform_0.4.3
[19] parallelly_1.46.1 KernSmooth_2.23-26 htmlwidgets_1.6.4
[22] desc_1.4.3 ica_1.0-3 plyr_1.8.9
[25] plotly_4.12.0 zoo_1.8-15 cachem_1.1.0
[28] igraph_2.2.3 mime_0.13 lifecycle_1.0.5
[31] pkgconfig_2.0.3 Matrix_1.7-5 R6_2.6.1
[34] fastmap_1.2.0 MatrixGenerics_1.22.0 fitdistrplus_1.2-6
[37] future_1.70.0 shiny_1.13.0 digest_0.6.39
[40] S4Vectors_0.48.1 patchwork_1.3.2 rprojroot_2.1.1
[43] tensor_1.5.1 RSpectra_0.16-2 irlba_2.3.7
[46] pkgload_1.5.1 GenomicRanges_1.62.1 beachmat_2.26.0
[49] progressr_0.19.0 spatstat.sparse_3.1-0 httr_1.4.8
[52] polyclip_1.10-7 abind_1.4-8 compiler_4.5.0
[55] withr_3.0.2 S7_0.2.1 BiocParallel_1.44.0
[58] fastDummies_1.7.5 pkgbuild_1.4.8 MASS_7.3-65
[61] DelayedArray_0.36.1 sessioninfo_1.2.3 tools_4.5.0
[64] vipor_0.4.7 lmtest_0.9-40 otel_0.2.0
[67] beeswarm_0.4.0 httpuv_1.6.17 future.apply_1.20.2
[70] goftest_1.2-3 glue_1.8.0 rhdf5filters_1.22.0
[73] nlme_3.1-169 promises_1.5.0 grid_4.5.0
[76] Rtsne_0.17 cluster_2.1.8.2 reshape2_1.4.5
[79] generics_0.1.4 gtable_0.3.6 spatstat.data_3.1-9
[82] tidyr_1.3.2 data.table_1.18.2.1 XVector_0.50.0
[85] xml2_1.5.2 BiocGenerics_0.56.0 spatstat.geom_3.7-3
[88] RcppAnnoy_0.0.23 ggrepel_0.9.8 RANN_2.6.2
[91] pillar_1.11.1 stringr_1.6.0 spam_2.11-3
[94] RcppHNSW_0.6.0 later_1.4.8 splines_4.5.0
[97] lattice_0.22-9 survival_3.8-6 deldir_2.0-4
[100] tidyselect_1.2.1 SingleCellExperiment_1.32.0 scuttle_1.20.0
[103] miniUI_0.1.2 pbapply_1.7-4 knitr_1.51
[106] gridExtra_2.3 Seqinfo_1.0.0 IRanges_2.44.0
[109] SummarizedExperiment_1.40.0 scattermore_1.2 stats4_4.5.0
[112] xfun_0.57 Biobase_2.70.0 devtools_2.5.0
[115] matrixStats_1.5.0 stringi_1.8.7 lazyeval_0.2.3
[118] evaluate_1.0.5 codetools_0.2-20 tibble_3.3.1
[121] cli_3.6.6 uwot_0.2.4 xtable_1.8-8
[124] reticulate_1.46.0 roxygen2_7.3.3 dichromat_2.0-0.1
[127] Rcpp_1.1.1 globals_0.19.1 spatstat.random_3.4-5
[130] png_0.1-9 spatstat.univar_3.1-7 parallel_4.5.0
[133] ellipsis_0.3.3 dotCall64_1.2 listenv_0.10.1
[136] viridisLite_0.4.3 scales_1.4.0 ggridges_0.5.7
[139] purrr_1.2.2 rlang_1.2.0 cowplot_1.2.0

</details>

[LouiseDck]

As you can see in the following reprex, Seurat indeed does not allow underscores. This issue has been raised on the Seurat github page before.

library(Seurat)
#> Loading required package: SeuratObject
#> Loading required package: sp
#> 
#> Attaching package: 'SeuratObject'
#> The following objects are masked from 'package:base':
#> 
#>     intersect, t

counts <- matrix(
  rpois(10 * 20, lambda = 5),
  nrow = 10,
  ncol = 20
)

rownames(counts) <- paste0("Gene_", 1:10)
colnames(counts) <- paste0("Cell_", 1:20)

seurat_obj <- CreateSeuratObject(counts = counts)
#> Warning: Feature names cannot have underscores ('_'), replacing with dashes
#> ('-')
#> Warning: Data is of class matrix. Coercing to dgCMatrix.

Created on 2026-07-09 with [reprex v2.1.1](https://reprex.tidyverse.org/)

The easiest way to get around this would be to change the featurenames of the object before converting.
I feel it would be out of scope for anndataR to detect this change from underscore to dashes and apply it to the other layers in the object.

[doliv071]

Thanks for pointing out the issue, I didn't see it when I initially checked their github.

I guess my question then is whether the last 3 calls in the traceback (LayerData<-*) are anndataR functions or Seurat/SeuratObject?

[lazappi]

Sorry we didn't get back to you earlier. Steps 6-9 in the traceback are from {Seurat}, 1-5 are from {anndataR}.