`UnicodeDecodeError` when reading H5AD in Python

[lazappi]

Thanks so much for your reply @lazappi . I tried to upgrade to Seurat v5 and upgraded the v4 data to the v5 structure. After exporting it to h5ad format using write_h5ad(), when I read this file in Python, an error occurred:

library(anndataR)
anndataR::write_h5ad(
  object = d2m1_all_v5,
  path = "anndataR_v5.h5ad")

adata1 = ad.read_h5ad("anndataR_v5.h5ad")

UnicodeDecodeError: 'ascii' codec can't decode byte 0xc3 in position 4: ordinal not in range(128).

I thought it was a problem with my data, but when I used another tool to convert my data, there was no issue when reading it in Python:

library(convert2anndata)
library(anndata)
sce <- convert2anndata::convert_seurat_to_sce(d2m1_all_v5)
ad <- convert2anndata::convert_to_anndata(
  sce,
  assayName = "RNA_counts", 
  useAltExp = T )
anndata::write_h5ad(ad, "convert2anndata_v5_counts.h5ad")

adata2 = ad.read_h5ad("convert2anndata_v5.h5ad")
adata2

AnnData object with n_obs × n_vars = 633056 × 24591
obs: 'orig.ident', 'nCount_RNA', 'nFeature_RNA', 'percent.mt', 'Patient_ID', 'Cell_cluster_major', 'Cell_cluster_minor', 'Tumor_Capsule_Integrity', 'Age', 'Gender', 'AFP_μg_L', 'Tumor_Stage', 'Size_cm', 'Vascular_Invasion', 'Etiology', 'Cirrhosis', 'GS_G', 'GS_S', 'ALB_mutation', 'APOB_mutation', 'ARID1A_mutation', 'ATRX_mutation', 'AXIN1_mutation', 'CPS1_mutation', 'CTNNB1_mutation', 'KEAP1_mutation', 'KMT2C_mutation', 'RB1_mutation', 'SMARCA2_mutation', 'TP53_mutation', 'TSC2_mutation', 'Patient', 'Sample', 'Tissue_Source', 'Diagnosis', 'Tumor_Type', 'Platform', 'Sample_ID', 'log10GenesPerUMI', 'percent.rb', 'percent.hb', 'scDblFinder.class', 'scDblFinder.score', 'Contamination', 'TNM', 'Site_Major', 'Site_Minor', 'Distant_Metastasis', 'CEA', 'CA199', 'Child_Pugh', 'BCLC_Stage', 'Number_of_Lesions', 'RNA_snn_res.0.01', 'RNA_snn_res.0.05', 'RNA_snn_res.0.08', 'RNA_snn_res.0.1', 'RNA_snn_res.0.2', 'RNA_snn_res.0.5', 'RNA_snn_res.0.8', 'RNA_snn_res.1', 'Cell_cluster_res.1', 'Cell_cluster_res.0.05'
var: 'vst.mean', 'vst.variance', 'vst.variance.expected', 'vst.variance.standardized', 'vst.variable'
obsm: 'X_harmony', 'X_pca', 'X_tsne', 'X_tsne_naive', 'X_umap', 'X_umap_naive'

Originally posted by @rainbowkiva in #399

[lazappi]

@rainbowkiva This looks like a problem with how we have written a non-standard character to the file, maybe something with an accent. Do you have any ideas what that might be in your original object? If you could try to find the character that is causing the problem that would be great. Using the mapping arguments to write a file with only some slots might help (e.g. setting write_h5ad(..., layers_mapping = FALSE) will write a file without the layers data.