When should I normalize and subset adata objects in scanpy?

Question

[bioinfo]

Hello,

I have some ingle cell samples that I have merged and done QC, hashtag demultiplexing and then doublet removal. I have 2 different genotypes with big chromosomal differences so I decided to analyze all the samples together and then also separately. I was wondering when would be the correct time to subset my matrix to the different genotypes. Should I do it before normalizing and doing log on the object with all the samples or should I normalize and log after subsetting the adata object based on the genotype?

From what I understand when running sc.pp.normalize_total(adata, target_sum=1e4) it normalized per cell so both ways should yield the same results right?

Thank you

Answer 1

[LChart]

The appropriate fork for all together vs. separated analysis is computing highly variable genes. It's the first step where cross-genotype differences could impact the results. Depending on your flavor= and subset= flags, this could be before or after normalization.