Prediction of butyrate production in microbial communities using machine learning
Abstract
Butyrate is a beneficial metabolite produced by the gut microbiome with disproportionate in vivo concentrations linked to disease, particularly inflammation-based disorders. We trained a machine learning model to predict microbial community butyrate production using published in vitro data. We explored different training and testing regimes, demonstrating that training on communities of intermediate complexity outperformed training on simple or highly complex communities while training on data distributed across communities of varying complexity was the most effective strategy. Our work aims to optimize experimental data collection strategies. Future modeling work will incorporate mechanistic metabolic information into predictive models to improve performance.
Primary Faculty Mentor Name
David Bernstein
Status
Undergraduate
Student College
College of Engineering and Mathematical Sciences
Program/Major
Biomedical Engineering
Primary Research Category
Engineering and Math Science
Prediction of butyrate production in microbial communities using machine learning
Butyrate is a beneficial metabolite produced by the gut microbiome with disproportionate in vivo concentrations linked to disease, particularly inflammation-based disorders. We trained a machine learning model to predict microbial community butyrate production using published in vitro data. We explored different training and testing regimes, demonstrating that training on communities of intermediate complexity outperformed training on simple or highly complex communities while training on data distributed across communities of varying complexity was the most effective strategy. Our work aims to optimize experimental data collection strategies. Future modeling work will incorporate mechanistic metabolic information into predictive models to improve performance.
