Should I normalize individually or after merging all the data?

Question

[SupBioInformatics]

Hi,

I am still practicing the Seurat workflow. I wonder how the pipeline will be the best practice for large datasets. Specifically, I am not sure if I should normalize prior to or after the merging. I list the questions in my mind:

• If I have 6 patients and each patient has two conditions, I will have total 12 datasets
• Should I normalize each data individually and then Merge. Or I can merge all together at the beginning and do rest of QC, i.e subset, NormalizeData/SCTtransform, ScaleData, etc..
• What perspectives needed to consider picking which normalization method (log vs SCT) I should use?
• Do I still need to use NormalizeData/ScaleData if I decide to use SCTransform?

Thank you in advance.

Answer 1

[rpolicastro]

In the preferred Seurat integration workflow when working with multiple samples, samples are first normalized separately using SCTransform, and then combined during integration. Refer to their SCTransform vignette here for more info.