Subject: Clarification on input data format: which preprocessing pipeline is expected?

[Chenchen3Chen]

Hello,

I am currently preparing my single-cell RNA-seq data for analysis with CellPhoneDB and would like to seek clarification regarding the required input data format.

In the documentation, the parameter counts_file_path is described as "paths to normalized counts file (not z-transformed)", with an example filename normalised_log_counts.h5ad. This suggests that the input data should be log-transformed.

However, I have also seen community discussions suggesting that log-transformation should not be applied, and that only library-size normalization is acceptable. This has left me confused about which preprocessing pipeline is actually recommended.

Specifically, I would like to confirm which of the following two approaches is the preferred one for CellPhoneDB:

Option A: Library-size normalization only

 import scanpy as sc
sc.pp.normalize_total(adata, target_sum=10000)
# Do NOT apply log1p

Option B: Library-size normalization + log-transformation

import scanpy as sc
sc.pp.normalize_total(adata, target_sum=10000)
sc.pp.log1p(adata)

Could you please clarify:

1. Which option is the recommended/preferred input for CellPhoneDB?
2. If Option A is recommended, would you consider updating the example filename (normalised_log_counts.h5ad) in the documentation to avoid confusion for future users?

I would greatly appreciate your guidance on this matter. Thank you for your time and for developing this wonderful tool!

Many thanks,
Chen