Good comprehensive video. One comment regarding differential expression analysis post-integration (starting at 56:27). In either bulk or single-cell sequencing, most tools highly recommend NOT doing DE analysis on corrected data, exactly because of the concerns regarding under- or over-correction and removing potential biological variation. If you read the literature around older batch correction tools such as ComBat, most recommend that DE analysis correct for batch effects within the statistical model (i.e. including a model of ~ Batch + Condition). Obviously, Seurat doesn't have that functionality, but if you perform pseudobulk aggregation for DE analysis (literature suggests this is more statistically sound), then tools capable of including batch effects in the regression model (DESeq2, edgeR, limma-voom, etc.) are available.
Good comprehensive video. One comment regarding differential expression analysis post-integration (starting at 56:27). In either bulk or single-cell sequencing, most tools highly recommend NOT doing DE analysis on corrected data, exactly because of the concerns regarding under- or over-correction and removing potential biological variation. If you read the literature around older batch correction tools such as ComBat, most recommend that DE analysis correct for batch effects within the statistical model (i.e. including a model of ~ Batch + Condition). Obviously, Seurat doesn't have that functionality, but if you perform pseudobulk aggregation for DE analysis (literature suggests this is more statistically sound), then tools capable of including batch effects in the regression model (DESeq2, edgeR, limma-voom, etc.) are available.