Data Normalization after merging two Seurat objects.

[nitrato]

Hi,

I am a Seurat rookie and I have a question regarding data normalization. I have two individual Seurat Objects, let us say A and B, for which the features in the RNA assay are not identical. I have merged A and B into a combined Seurat Object, and then proceeded with Seurat workflow. When you run NormalizeData, counts from A and B are normalized separately.

My Question is: Is there a way to normalize A B data together? If not, would it be wrong to create, from the start, a single AB dataset where genes present in A but not in B, and vice versa, would have 0 counts in all cells from the set in which they are not accountable?

Please note I used filtered data from CellRanger to create the Seurat object. Alternatively, I guess I could use the raw data, which brings up more questions, but I'd rather leave it like this by now. My post is long enough as it is.

Thanks for your help!

Regards,

David

[rharao]

I have two individual Seurat Objects, let us say A and B, for which the features in the RNA assay are not identical.

Can you tell us more about the difference between the objects and why they don't have the same feature sets?

[nitrato]

Hi,
Thanks for your reply. I think I might have got it and how to come around it.
to get A and B objects I run Read10X() and then CreateSeuratObject() on each dataset. The problem came from the arguments under the latter, i.e. min.cells = 3, min.features = 200, which yielded, obviously, different amount of genes o neither object.
Thus, What I guess I can do is just cbind() before CreateSeuratObject() and then use CreateSeuratObject() as before.
Am I right?
PS: When binding I guess I should tag Barcodes with A and B, even though there is not repeated barcodes between A and B
Regards,

David

[rharao]

It's good that you've identified the source of the disparity as the min.cells argument, rather than, for example, different reference genomes in mapping. What is the relation between the datasets, though? The experimental design does effect the best way to proceed.

• Are they totally different conditions?
• Are they biological replicates?
• Are they from the same biological sample but different GEM wells?

Thanks for posting!

[nitrato]

Hi Rharao,
Thanks for your reply. You are right I should have pointed out that A and B correspond to cells from two different genotypes that have undergone the same treatment. The goal of the study is to see the differences in cells between those two genotypes upon treatment. I must add that the experimental design is not optimal, in my opinion. For instance: cells come from eight different mice of each genotype. However, once the cells of interest were isolated from each individual, they were pooled before the library synthesis, so that only two libraries were synthesized, the famous A and B. It was certainly cheaper, but not optimal.

Thanks!

David