Grant: Discovery Grants Program - Individual

Host-adapted microbial communities (or microbiomes) are assemblies of commensal, symbiotic and pathogenic microorganisms present at specific body sites. These communities are often represented as vectors of relative abundances at defined taxonomic levels, such as genera or species. Globally, certain configurations of these microbial communities appear more frequently than others. Multiple factors, ranging from environmental pressures to intrinsic microbe-microbe interactions, could explain these prevalent configurations. In humans, significant efforts have been dedicated to defining microbial configurations (MCs) within various microbiomes, including those of the vagina, skin, oral cavity, and gut. It is noteworthy that MCs of the gut, for example, can correlate with factors like the host's diet, health status, and specific metabolic or immunological traits. Historically, many of these definitions have leaned on linear, unsupervised models or employed clustering of microbiome compositional data. Several studies advocate for viewing MCs as continuous gradients, though observations of exclusionary microbial interactions suggest the presence of at least quasi-discrete states. Current models of MCs still tend to be overly simplistic, frequently defined by the dominance of particular taxa and neglecting the functional interactions that may define specific MCs. There is a pressing need to re-evaluate and refine our understanding of MCs, particularly in terms of their functional roles and the dynamics ensuring their stability and resilience. This research program seeks to advance this understanding, proposing a novel conceptual framework for studying host-adapted MCs across a diverse global population. This will be achieved by combining advanced computational methodologies with cutting-edge microbial single-cell profiling techniques. Objective 1 aims to develop deep learning methods to model non-linear trends across features, such as taxa or gene/pathway abundances, capturing the plasticity of MCs across global populations through a low-dimensional latent space termed 'deep-MC'. Objective 2 aims to develop a single-cell RNA profiling platform, termed sc-MC, for refined functional characterization of complex microbiomes. Objective 3 focuses on validating and evaluating the abilities of the deep-MC models and sc-MC profiles to predict the response of microbial communities to interventions (exercise, probiotics). As our understanding of MCs evolves, so will the potential applications to modulate microbiomes for a diverse range of biomedical, environmental, food sciences and synthetic biology applications. Regardless, the deep-MC models and sc-MC assay represent important tools that will capture the global properties and (single-cell) functions of human gut microbiomes, while also having applicability to other microbiomes.