Means for reciprocal cell type comparisons are different

[manf1984]

Dear authors of cellphoneDB, first of all, thanks for this very very useful tool!
I am using it to evaluate ligand-receptor interactions among different annotated cell type a scRNAseq dataset.
I am using cellphonedb 5.0.0 in a python 3.10 dedicated conda environment.

Checking the output files generated I noticed one discrepancy. The means for an interaction (e.g. GENE A - GENE B) are not the same between reciprocal/mirror cell type pairs (e.g. "cell type A | cell type B" and "cell type B | cell type A"). How is this possible?

Code below

from cellphonedb.src.core.methods import cpdb_statistical_analysis_method

cpdb_results = cpdb_statistical_analysis_method.call(
         cpdb_file_path = "/path/to/cellphonedb.zip",
         meta_file_path = "/path/to/Metadata_for_cellphoneDB.csv",
         counts_file_path = "/path/to/AnnData_for_cellphoneDB.h5ad",
         counts_data = 'hgnc_symbol',
         output_path = "/path/to/output")

Then I check the txt output files with "means" and "significant means"

result_file_path = "/path/to/statistical_analysis_means_06_11_2024_124047.txt"
resultsDf = pd.read_csv(result_file_path, sep = "\t", header = 0)


print(resultsDf.loc[288, "Cell type A|Cell type B"]) #mean is 2.279
print(resultsDf.loc[288, "Cell type B|Cell type A"]) #mean is 0.006 

The values are different. But the cells are exactly the same, just the comparison is reciprocal. Why are the values different? Is that expected?
Thanks in advance for your help with this!

[ktroule]

Hi.

They are different because you are looking at the same pair of cells and interacting proteins but with the proteins swapped:

• Cell type A (Gene A) | Cell type B (Gene B)
• Cell type A (Gene B) | Cell type B (Gene A)

[manf1984]

Ohh I see!
That explains it! Thanks very much for your quick reply!