Reconstruction and analysis of gut microbiota metabolic networks in the context of personalized nutrition

Resumen

The gut microbiota refers to the collection of microorganisms coexisting in the human gut. It is well-established that various species within our gut microbiota metabolize nutrients derived from the diet, transforming them into key compounds that regulate human health. In particular, the composition of the intestinal microbiota is directly linked to the development of relevant diseases. Understanding the relationship between microbiota, diet and human health constitutes a primary objective in biomedicine that can be investigated from different approaches. Genome-scale metabolic reconstructions, developed in the field of Systems Biology, have grown enormously in recent years and constitute a very promising tool to develop computational models of the gut microbiota. However, this methodology has not reached an adequate level of maturity, making it necessary to improve existing reconstructions and algorithms for different applications. The main objective of this doctoral thesis is to improve existing metabolic reconstructions of the human gut microbiota, specifically integrating degradation pathways of diet-derived compounds of great relevance for human health, as well as to develop new algorithms that allow predicting and analyzing metabolic alterations of the gut microbiota in different contexts. This work is primarily divided into two parts. The first part focuses on improving existing metabolic reconstructions of the human gut microbiota. In this section, we introduce AGREDA, a repository of microbial metabolic models that includes metabolic pathways for a wide range of dietary compounds not represented in previous reconstructions. Among the different families of metabolites, we have considerably improved the representation of phenolic compounds, highly relevant for nutrition and human health. In addition, by means of metabolomic studies, we were able to validate the capacity of AGREDA to predict compounds derived from bacterial metabolism in the gut microbiota. The second part of this doctoral thesis involves the development of computational methods able to explore the metabolism of the human gut microbiota. Firstly, we developed q2-metnet, a QIIME2 plugin that allows inferring functional metabolic features of a microbial community by integrating 16S gene sequencing data and metabolic reconstructions. We observed that the metabolic predictions of q2-metnet can effectively classify a cohort of patients with various clinical conditions in comparison with competing algorithms. Finally, we present the BN-BacArena algorithm, which enable the dynamic modeling of bacterial communities considering microbe-microbe and nutrient-microbe regulatory interactions, empirically extracted using Bayesian Networks. We demonstrate with experimental data that BN-BacArena improves the ability of other methods in the literature to predict relative abundance of species.