DESeq2 comparison of normalized counts from different experiments

Question

[Matthias]

Hi,

I'm putting together a database to compare gene expression across many different RNA-seq experiments. I have multiple DESeq2 normalized counts matrices from the different experiments.

My question is am I able to merge these normalized counts matrices by gene names and directly compare the gene expression across all of these experiments? Or will I have to re-run DESeq2 with all samples together to have the normalization done together?

Thanks, Matt

[piyushjo]

The DeSeq2 normalizes counts with respect to library size. So in principal it should be able to compare different sequencing runs together even with different library sizes.

Answer 1

[jared.andrews07]

You'll want to run them all together, otherwise the gene-dispersion estimates won't be accurate, meaning your counts won't be comparable across experiments. You'll likely still run into batch effects, but at least you can try to account for those.

[Kevin Blighe]

Moved your comment to an answer, jared.