Hello, I have 136 cell surface protein label data in my scRNA seq data. I normalized the protein data with "CLR", I have 8 samples in each treatment. I understand I need do pseudobulk analysis before the differential expression of Gene analysis. My questions is, for the small number of Protein, should I still need to do the pseudobulk analysis before I do the differential expression on the protein? I tried pseudobulk analysis before I do the protein differential analysis, no significant protein was found, I want to know if I can do 136 protein differential analysis without pseudobulk analysis? is it acceptable in statistics? I really appreciate if any professional suggestions.

Best, Lei

Code should be placed in three backticks as shown below

# include your problematic code here with any corresponding output 
# please also include the results of running the following in an R session 

sessionInfo( )

There is no "MUST", but pseudobulk aggregation is beneficial to a) enable use of tested and robust methods such as limma (though it could use cells as replicates), b) and to avoid inflated p-values due to pseudoreplication and c) avoid that individual donors contribute overly much to the inference if no proper even sampling (same number of cells per donor) was done. You can use for example limma here given that you can robustly normalize the data somehow. See for example this thread on using limma with a small number of features LIMMA on RT-PCR data and the section in the OSCA book towards surface features: https://bioconductor.org/books/release/OSCA.advanced/integrating-with-protein-abundance.html