bakta VS metaGEM

If you need more accurate ORF(CDS) prediction including functional annotation, I recommend using CLI tools such as prokka, bakta, or DFAST (DFAST is also available in a web version).

Maybe this workflow will be interesting to you: it assembles metagenomes, generates bins, and also creates metabolic models for predicting interactions within microbiomes 💎 https://github.com/franciscozorrilla/metaGEM

For example, I started developing metaGEM for my masters thesis, although it took another 2 years of work to get the publication and GitHub repo ready 💎 Although it’s not multi-omics, it spans across metagenomic analysis and metabolic modeling.

shameless self plug https://github.com/franciscozorrilla/metaGEM it helps to get feedback from your colleagues, especially the more design-minded ones. here's what mine looked like before feedback https://github.com/franciscozorrilla/metaGEM/wiki

TLDR you get higher genomic resolution compared to 16S. Also consider the fact that there is a lot of strain level variation within species, which you completely miss out on without shotgun or long read sequencing. Self plugging our workflow that takes in shotgun sequencing reads, assembles MAGs and then reconstructs metabolic models that can be used for flux balance analysis simulations https://github.com/franciscozorrilla/metaGEM

In my experience I haven’t found metaquast or other assembly evaluation tools very useful precisely because they are geared toward reference genome based assessment. I don’t think there is a standardized way of assessing your assemblies (someone please correct me if I’m wrong), but it helps to look at the distribution of contig lengths. For example, an assembly with a distribution peak around 10kbp is much better than a peak around 1kbp. In your case you probably want to bin the assembled contigs into MAGs and then assess the quality of those genomes using a tool like CheckM or BUSCO. If you want to get an idea of tools/workflows you can use then maybe check out the metaGEM pipeline on GitHub or read the paper

One thing you could try is generating metagenome assembled genomes (MAGs) from each metagenome, use those MAGs to automatically reconstruct genome scale metabolic models (GEMs), and then do flux balance analysis (FBA) based simulations to compare the predicted metabolism across different treatments. We developed the metaGEM pipeline for exactly this purpose, you can read more about it here or check it out on GitHub

At the risk of self promotion, I would invite you to check out the metaGEM 💎 GitHub repo/paper. This was originally my MS thesis which I published at the start of my PhD. It is a Snakemake workflow for generating context specific genome-scale metabolic models and predicting metabolic interactions within microbial communities directly from short read metagenomic data, i.e. bioinformatics + systems biology. If this is interesting, you may also want to check out my other pinned repos which include related tutorials and resources.

Paper: https://academic.oup.com/nar/article/49/21/e126/6382386 GitHub: https://github.com/franciscozorrilla/metaGEM

Perhaps you may be interested in contributing to the development of the metaGEM pipeline? There are a number of ideas/modifications that I wanted to explore but have not had the time to do yet, you can find these in the issues sections with the "method" flag (https://github.com/franciscozorrilla/metaGEM/labels/method). You can have a look and see if anything piques your interest, in particular I think this would be a good one to address: https://github.com/franciscozorrilla/metaGEM/issues/31 . Here is the paper if you want to get more info about the pipeline itself: doi.org/10.1093/nar/gkab815

GitHub: https://github.com/franciscozorrilla/metaGEM Paper: https://academic.oup.com/nar/advance-article/doi/10.1093/nar/gkab815/6382386