Seurat Integration Post-Normalization and Cluster Identification or Pre those Steps

Question

[fouerghi20]

I have two samples that I have analyzed separately using the Seurat pipeline, which means I currently have two Seurat objects that I have run the normalization on. I am currently thinking about integrating the two samples, so I can run the analysis on it. In my current workflow, I integrate the two samples after having run the normalization, scaling and assigned cell types by identifying the clusters through the marker genes. I am wondering if it is a better practice if I integrate the two samples, and then normalize them, find the variable features, run PCA to find the appropriate PCs for dimensionality reduction and then run UMAP/PHATE/tSNE etc. Any advice on which practice is best would be greatly appreciated.

[fouerghi20]

Still looking for help! Thank you.