I am analyzing a time series of a metatranscriptome. In metatranscriptomics, we have to do the regular normalizations found in RNA-Seq (ie. gene sequence length, and sample-differences in total expression). However, there is an additional normalization required to account for changes in species abundance.

I would like to analyze differences in species-level transcript abundance, so I've done all these normalizations. But to then analyze differential expression with DeSeq2, it seems I shouldn't put the normalized values as input. Is there a recommended way around this?

Any tips or advice is appreciated!

So DESeq2 does it's own normalization such that you are comparing differences between counts after normalizing away any global shifts (it's not possible to separate these from sequencing depth without making more assumptions or including prior information about housekeeping genes etc). That's the RNAseq context. The short answer is that the methods only work with unnormalized counts but the software is flexible such that you can modify the normalization parameters how you like (see vignette on sizeFactors and normalizationFactors).