Is n_obs and n_var possibly mixed in some test data?

[uzulim]

Hi, I'm a postdoc working on topological data analysis.

I might be mixing things up but I observed that n_obs and n_var might be mixed up in a test dataset.

I ran:

adata = dyn.sample_data.pancreatic_endocrinogenesis()
preprocessor = Preprocessor()
preprocessor.preprocess_adata(adata, recipe="monocle")
dyn.tl.dynamics(adata)
dyn.tl.reduceDimension(adata)

print(adata.obsm['X_umap'].shape)
print(adata.var['nCells'])
print(adata.obs['nGenes'])

and the output is:

(3696, 2)

index
Xkr4          6
Gm37381       0
Rp1           0
Rp1-1         0
Sox17         0
           ... 
Gm28672       0
Gm28670       0
Gm29504       0
Gm20837       0
Erdr1      1200
Name: nCells, Length: 27998, dtype: int64

index
AAACCTGAGAGGGATA    2356
AAACCTGAGCCTTGAT    2613
AAACCTGAGGCAATTA    1984
AAACCTGCATCATCCC    2860
AAACCTGGTAAGTGGC    2211
                    ... 
TTTGTCAAGTGACATA    3283
TTTGTCAAGTGTGGCA    2605
TTTGTCAGTTGTTTGG    2168
TTTGTCATCGAATGCT    3155
TTTGTCATCTGTTTGT    2551
Name: nGenes, Length: 3696, dtype: int64

This means that there are 27998 cells and 3696 genes.
However, note that the UMAP-reduced array has 3696 data points.
Each data point then represents one gene, but it is supposed to represent one cell.

Am I missing something, or is the original gene expression array unintentionally transposed so that the row and column have been mixed up for the UMAP projection?