Its probably best practice to read all samples into the same seurat object, then do the clustering and asign the cell type labels. Anyways if you got 2 seurat objects and want to merge them Seurat has functions to do that: https://satijalab.org/seurat/articles/integration_introduction.html . You can add a new metadata column as you mention ("experimental"-"control"), maybe a second column pasting condition+cell_type and do the comparison all cell_type_1_control vs cell_type_1_experimental with FindMarkers function as usual. Alternatively if you want to use DEseq2 you can create pseudobulk profiles for each sample and cell type with AggregateExpression function, then do the comparison with DESeqDataSetFromMatrix. Although comparison between pseudobulks derived from very different cell numbers will probably be problematic.