Batch correction and Integration: For 2 conditions, 3 replicates(n=3).

[dmsalsgh97]

Hi, first, thanks for developing and maintaining the Seurat package!!

I'm analyzing the scRNA-seq data with 2 conditions, 3 replicates(n=3), and 3 tissues( total 18samples).
In my think, my replicates are showing minor batch effects (3 replicates with the same condition):

Now, I want to compare 2 conditions on the same tissue for doing downstream analysis. I think I have 3 possible cases:
1) Just merge 3 replicates for each condition, and integrate 2conditions.
2) Remove batch effects between 3 replicates using FastMNN and integrate the batch-removed objects.
3) Just merge all 6 samples

Which one is appropriate for comparing the 2 conditions with 3 replicates(n=3)?

My own understandings of "Integration" function is to avoid generating condition-specific clusters that hinder comparing.
But, When I tried case3 and case1, It dosen't seems make bing difference...