Cannot retrieve matrix and metadata after reading a h5ad data with read_h5ad

[joonlee3]

Hi,

I was trying to convert a h5ad file (anndata) to a Seurat object. It seems that read_h5ad read my h5ad file in R.

> COPD.anndata <- read_h5ad("COPD.h5ad")
> COPD.anndata

AnnData object with n_obs × n_vars = 918486 × 5101
    obs: 'cell_id', 'transcript_counts', 'control_probe_counts', 'genomic_control_counts', 'control_codeword_counts', 'unassigned_codeword_counts', 'deprecated_codeword_counts', 'total_counts', 'cell_area', 'nucleus_area', 'nucleus_count', 'z_level', 'cell_labels', 'nucleus_ratio', 'n_genes_by_counts', 'log1p_n_genes_by_counts', 'log1p_total_counts', 'pct_counts_in_top_10_genes', 'pct_counts_in_top_25_genes', 'pct_counts_in_top_50_genes', 'pct_counts_in_top_100_genes'
    var: 'gene_ids', 'feature_types', 'genome'
    uns: 'spatialdata_attrs'
    obsm: 'spatial'

However, when I retrieved matrix data and metadata in the object, I've got the following errors

> COPD.anndata$X

Error in as.integer(as_r_value(x$indices)) : `
  cannot coerce type 'environment' to vector of type 'integer'

> COPD.anndata$obs

Error in py_convert_pandas_df(x) : `
  INTEGER() can only be applied to a 'integer', not a 'environment'

Please let me know if there is any solution. Thank you.

Best,
Joon

[rcannood]

Hi @joonlee3 !

Could you send me the COPD.h5ad or create a minimal reproducible example?

Thanks!

Robrecht

[Hato1713]

Hi ,
I met the same issue when reading an Xenium ovarian data with read_h5ad:

ad <- import('anndata')
adata1 <- ad$read_h5ad('/PATH/data/Xenium/ovarian/adata_subtest_.h5ad')

adata1$obs

Error in py_convert_pandas_df(x): INTEGER() can only be applied to a 'integer', not a 'environment'
Traceback:

1. py_to_r_ifneedbe(private$.anndata$obs)
2. private$.anndata$obs
3. `$.python.builtin.object`(private$.anndata, obs)
4. py_get_attr_or_item(x, name, TRUE)
5. py_maybe_convert(object, py_has_convert(x))
6. py_to_r(x)
7. py_to_r.pandas.core.frame.DataFrame(x)
8. py_convert_pandas_df(x)
9. .handleSimpleError(function (cnd) 
 . {
 .     watcher$capture_plot_and_output()
 .     cnd <- sanitize_call(cnd)
 .     watcher$push(cnd)
 .     switch(on_error, continue = invokeRestart("eval_continue"), 
 .         stop = invokeRestart("eval_stop"), error = NULL)
 . }, "INTEGER() can only be applied to a 'integer', not a 'environment'", 
 .     base::quote(py_convert_pandas_df(x)))

The anndata version was 0.11.4, r-anndata version was 0.7.5.6, under R version 4.3.0 and python version 3.12.2

The data was a public data，although had undergone a preprocessing:
https://www.10xgenomics.com/datasets/xenium-prime-ffpe-human-ovarian-cancer

During the debugging process, I found that NaN values existed in the adata.obs:

sum([sum(obs[i].isna()) for i in obs.columns])
# result: 470

I am wondering if this was the cause of the issue

Thank you.

Hato1713 2025/4/28

[Hato1713]

Hi , I met the same issue when reading an Xenium ovarian data with read_h5ad:

ad <- import('anndata')
adata1 <- ad$read_h5ad('/PATH/data/Xenium/ovarian/adata_subtest_.h5ad')

adata1$obs

Error in py_convert_pandas_df(x): INTEGER() can only be applied to a 'integer', not a 'environment'
Traceback:

1. py_to_r_ifneedbe(private$.anndata$obs)
2. private$.anndata$obs
3. `$.python.builtin.object`(private$.anndata, obs)
4. py_get_attr_or_item(x, name, TRUE)
5. py_maybe_convert(object, py_has_convert(x))
6. py_to_r(x)
7. py_to_r.pandas.core.frame.DataFrame(x)
8. py_convert_pandas_df(x)
9. .handleSimpleError(function (cnd) 
 . {
 .     watcher$capture_plot_and_output()
 .     cnd <- sanitize_call(cnd)
 .     watcher$push(cnd)
 .     switch(on_error, continue = invokeRestart("eval_continue"), 
 .         stop = invokeRestart("eval_stop"), error = NULL)
 . }, "INTEGER() can only be applied to a 'integer', not a 'environment'", 
 .     base::quote(py_convert_pandas_df(x)))

The anndata version was 0.11.4, r-anndata version was 0.7.5.6, under R version 4.3.0 and python version 3.12.2

The data was a public data，although had undergone a preprocessing: https://www.10xgenomics.com/datasets/xenium-prime-ffpe-human-ovarian-cancer

During the debugging process, I found that NaN values existed in the adata.obs:

sum([sum(obs[i].isna()) for i in obs.columns])

result: 470

I am wondering if this was the cause of the issue

Thank you.

Hato1713 2025/4/28

This issue is solved by degrading numpy version from 2.* to 1.26.*

Hato1713 2025/4/28