Seurat integrated analysis with replicates

Question

[from the mountains]

I have done some exploring with Seurat integrated analysis and was wondering how I can incorporate biological replicates. So in a hypothetical experimental design, I would have 2 biological replicates for untreated samples and 2 biological replicates (same platform for everything). How would I go about doing this comparison? Would I first merge the replicates, then compare the groups? I am slightly confused about what situations one would use merge function. Is it similar to cellranger aggr?

Answer 1

[jared.andrews07]

Merge is a flat cbind/rbind scenario across the samples. It's not normalizing anything, just concatenating everything into a single Seurat object. If you don't think you have batch effects or anything, that'd be what I'd do. Then just compare between the groups.

Seurat's integration method is quite heavy handed in my experience, so if you decide to go the integration route, I'd recommend using the SeuratWrapper around the fastMNN function from batchelor.

[from the mountains]

Thank you for taking the time to answer my question. But then what is the purpose of Seurat's merge?

[jared.andrews07]

To merge multiple samples into one data object for further analysis.

[from the mountains]

i guess my question is, will merge do any normalization or removal of batch effects?

[jared.andrews07]

And as my answer says, no. It's just throwing all the data in the same object. It is not meant for normalizing for batch effects, but rather for scenarios in which batch effects are not a concern.

[from the mountains]

got it, i think i didn't see your edit after you initially answered the question. Thank you.