I have two datasets I am trying to run a Seurat integration on. I am following the vignette on the Seurat website. One of my dataset has a bunch of Nan row names (the genes columns), and if I drop them before the integration I get one set of results. If I do not drop the nan rows, but just find the common genes between my two datasets and subset the row names by the common genes, I am getting another result. Which approach is correct?