Correct normalization and merge approach for Spatial Transcriptomics

[schmerlain14]

Hi!

We have a very similar use case as the shown demo "brain" dataset but with 2 individuals and 2 conditions, so 4 slices in total. We need to merge them correctly, which is not easy after the tutorial and reading the reference.

The question is: What is the correct approach: Normalize each slice individually then merge OR merge the 4 slices then normalize OR normalize then merge then normalize?

In the tutorial the two slices are normalized and then merged, according to the vignette about merging this is not the correct way to do it: "By default, merge() will combine the Seurat objects based on the raw count matrices, erasing any previously normalized and scaled data matrices. If you want to merge the normalized data matrices as well as the raw count matrices, simply pass merge.data = TRUE. This should be done if the same normalization approach was applied to all objects." https://satijalab.org/seurat/articles/merge_vignette.html#merge-based-on-normalized-data-1
Also it was already discussed at this Question #4430 without a proper answer.

I tried it with merge then normalize and normalize - merge -normalize and got slightly different results which means that the normalization is not deleted despite stated in the vignette. So we don't know what the correct approach is.