Metabolic Prediction from Shotgun Sequences without Abundance


Instead of predicting metabolic output based on 16S or shotgun sequences from an actual sample, which would have information on relative abundance, I am interested in whether I can use MelonnPan to predict the metabolic output of hypothetical microbiota communities. That is, I would have a library of shotgun sequences from databases only and would examine the effect of different combinations on predicted metabolic output, to identify combinations of interest to then grow in vitro/colonize mice with. Do you know if there is a way to accomplish this with MelonnPan? Or do you know of another tool that can? As the community is hypothetical, I also do not have metabolomics data.

Thank you very much!


This is not something MelonnPan is suited to (since it needs real mbx data for training). It sounds like you’re interested in something more akin to Flux Balance Analysis (i.e. predicting metabolic potential from a given reaction set)?

Hi Frazosa,

Thanks for your reply! I would interested in Flux Balance Analysis but the given reaction set would have to be all of the metabolites present in a human or mouse intestine. I want to predict how the metabolic output of x vs. y vs z consortia of bacteria would differ under physiological conditions. If that is not possible, I have metabolomics on the media I use (brain heart infusion), so could use that as the input metabolites, but it would still be a lot of metabolites. Do you know anything suitable for this? Thanks! Noah

That is, I want to find a community (among many possibilities) that maximizes metabolic output along some pathways and minimizes output along others, and then colonize a GF/antibiotic treated mouse with that community

Maximize X and minimize Y is exactly the sort of thing that FBA seeks to accomplish via its objective functions, so you could look into the literature / software in that area. It’s not something we currently cover in the bioBakery suite.

Will do, thanks for your help!