ValueError: Data matrix has wrong shape (20000, 28649), need to be (20000, 136125)

[EYNOCATTFOS]

scar /data/R04/usr/joint_nuclei/02_F18W/outs/filtered_feature_bc_matrix.h5 -ft mRNA -o /data/R04/usr/joint_nuclei/03_analysis/scar

/data/R04/usr/anaconda3/envs/scar/lib/python3.12/site-packages/anndata/_core/anndata.py:1793: UserWarning: Variable names are not unique. To make them unique, call .var_names_make_unique.
utils.warn_names_duplicates("var")
2025-08-20 17:48:57|INFO|scar|unprocessed data contains: 20000 cells and 139761 genes
2025-08-20 17:48:57|INFO|scar|filter out features of zero counts, remaining data contains: 20000 cells and 136125 genes
/data/R04/usr/anaconda3/envs/scar/lib/python3.12/site-packages/anndata/_core/anndata.py:1793: UserWarning: Variable names are not unique. To make them unique, call .var_names_make_unique.
utils.warn_names_duplicates("var")
2025-08-20 17:48:57|INFO|scar|modalities to denoise: Gene Expression
2025-08-20 17:48:57|INFO|scar|feature_type: mRNA
2025-08-20 17:48:57|INFO|scar|count_model: binomial
2025-08-20 17:48:57|INFO|scar|output_dir: /data/R04/zhangchao/joint_nuclei/03_analysis/scar
2025-08-20 17:48:57|INFO|scar|count_matrix_path: /data/R04/zhangchao/joint_nuclei/02_F18W/outs/filtered_feature_bc_matrix.h5
2025-08-20 17:48:57|INFO|scar|ambient_profile_path: None
2025-08-20 17:48:57|INFO|scar|expected data sparsity: 0.90
2025-08-20 17:48:57|INFO|model|cuda is detected and will be used.
2025-08-20 17:48:57|INFO|model|Ambient profile not found in AnnData.uns['ambient_profile'], estimating it by averaging pooled cells...
2025-08-20 17:48:57|INFO|model| Evaluate ambient profile from cells
2025-08-20 17:48:58|INFO|VAE|Running VAE using the following param set:
2025-08-20 17:48:58|INFO|VAE|...denoised count type: mRNA
2025-08-20 17:48:58|INFO|VAE|...count model: binomial
2025-08-20 17:48:58|INFO|VAE|...num_input_feature: 28649
2025-08-20 17:48:58|INFO|VAE|...NN_layer1: 150
2025-08-20 17:48:58|INFO|VAE|...NN_layer2: 100
2025-08-20 17:48:58|INFO|VAE|...latent_space: 15
2025-08-20 17:48:58|INFO|VAE|...dropout_prob: 0.00
2025-08-20 17:48:58|INFO|VAE|...expected data sparsity: 0.90
2025-08-20 17:49:11|INFO|model|kld_weight: 1.00e-05
2025-08-20 17:49:11|INFO|model|learning rate: 1.00e-03
2025-08-20 17:49:11|INFO|model|lr_step_size: 5
2025-08-20 17:49:11|INFO|model|lr_gamma: 0.97
Training: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████| 800/800 [39:18<00:00, 2.95s/it, Loss=1.4234e+03]
2025-08-20 18:34:36|INFO|scar|Saving results...
/data/R04/usr/anaconda3/envs/scar/lib/python3.12/site-packages/anndata/_core/anndata.py:1793: UserWarning: Variable names are not unique. To make them unique, call .var_names_make_unique.
utils.warn_names_duplicates("var")
Traceback (most recent call last):
File "/data/R04/usr/anaconda3/envs/scar/bin/scar", line 7, in
sys.exit(main())
^^^^^^
File "/data/R04/usr/anaconda3/envs/scar/lib/python3.12/site-packages/scar/main/main.py", line 212, in main
denoised_adata.X = scar_model.native_counts
^^^^^^^^^^^^^^^^
File "/data/R04/usr/anaconda3/envs/scar/lib/python3.12/site-packages/anndata/_core/anndata.py", line 675, in X
raise ValueError(msg)
ValueError: Data matrix has wrong shape (20000, 28649), need to be (20000, 136125).

[EYNOCATTFOS]

i use cellranger-arc count outputs filtered_feature_bc_matrix.h5, it includes:
adata
AnnData object with n_obs × n_vars = 20000 × 32285
var: 'gene_ids', 'feature_types', 'genome', 'interval'
but scar detect "scar|unprocessed data contains: 20000 cells and 139761 genes". It confused me.

[EYNOCATTFOS]

The origin codes :

adata = sc.read_10x_h5(count_matrix_path, gex_only=False)
main_logger.info(
"unprocessed data contains: {0} cells and {1} genes".format(
adata.shape[0], adata.shape[1]
)

But when I print "adata.shape[0], adata.shape[1]", it truely output:

adata.shape[0]
20000
adata.shape[1]
32285

So why scar detect 139761 genes? thx so much!

[EYNOCATTFOS]

i have solved the problem, exm.
adata = sc.read_10x_h5(count_matrix_path, gex_only=False), "gex_only=False " it means read RNA and peaks data, so the shape[1] equals the number of genes and peaks.
so I recommand using cellranger count rather cellranger-arc count to generate outputs

[EYNOCATTFOS]

I want to say: why must gex_only=False ? there is no need! if i use the genes and peaks together, how to use ?