Say you have two metagenomic samples. You want to quantify the differential abundance of 15 genes between the two samples. What's the most straight-forward way to do this, while properly accounting for the sequencing depth of each library?
You'll have to find a normalizing gene, a gene that should have the same levels in both samples. It would be very hard to normalize otherwise. Another approach is to have conclusions in the form: gene A is more abundant than gene B in sample 1 (or condition 1) but less in sample 2. It would be much stronger if you'll look at pathways or modules etc.