Normalization to Read Depth After Alignment to Database

Question

[Nat]

Hi, first time poster and very new to bioinformatics!

I have microbial metagenomic data from two different patients that I want to identify what genes are differentially abundant in each patient (note: this is all DNA). I aligned my data to IGC (a microbial reference database utilizing KEGG Orthology) with bowtie2. I figured out how to get the total number of reads aligning to each KO accession number but am unsure how to normalize this. I looked into DESeq2 but saw I need a gtf (annotated genome file) to run FeatureCounts to plug into DESeq2. I was unsure if I could just divide read counts by sequencing depth (essentially CPM) or if more rigorous normalization is expected. Thanks for any advice you can provide!