Explainable Machine Learning for Longitudinal Multi-Omic Microbiome

Over the years, research studies have shown there is a key connection between microbial community in gut, genes, and immune system. Understanding this association may help discover cause of complex chronic idiopathic disorders such as inflammatory bowel disease. Even though important efforts been put into field, functions, dynamics, causation dysbiosis state performed by remains unclear. Machine learning models can elucidate connections relationships microbes human host. Our study aims to extend current knowledge associations microbiome health disease through application dynamic Bayesian networks describe temporal variation gut microbiota taxonomic entities clinical variables. We develop set preprocessing steps clean, filter, select, integrate, model informative metagenomics, metatranscriptomics, metabolomics longitudinal data from Human Microbiome Project. This accomplishes novel network with satisfactory predictive performance (accuracy = 0.648) for each state, validating framework developing interpretable understand basic ways different biological (taxa, metabolites) interact other given environment (human gut) over time. These findings serve starting point advance discovery therapeutic approaches new biomarkers precision medicine.