scRNA samples: merge vs. integration

Question

[jhb80]

Hello!

I was hoping to catch a quick collection of opinions on the subject of sample batch handling in single cell RNA-seq analyses. For highly related sample batches (e.g. replicates from the same cell population, or same cell type populations from different donors), would you:

a) Do a simple 'merge' on raw counts and process downstream?

b) Use one of the newer 'integration' approaches (e.g. as implemented in Seurat's IntegrateData (Butler et al.)?

Thanks!