I have 4 scRNAseq samples that I create Seurat objects for. I do the QC on each of these samples and then I merge them using the merge function from Seurat. I then normalize, scale, and run PCA on the merged dataset. Then, I take the normalized count matrix from Seurat and input it into Harmony. Do I have to normalize and scale each sample individually before merging it and normalize it again? or do I normalize only after I merge the four samples together. Any guidance would be appreciated!