Gut Microbiome Insights for Pediatric Health Using Machine Learning Analyses

The human gut microbiome, especially in infants and children, is influenced by various factors such as delivery mode, antibiotic exposure, and breastfeeding. This thesis investigates the application of machine learning methods to analyze complex and multidimensional data from microbiome studies in pediatric cohorts. The aim is to elucidate how early-life factors impact gut microbiota and their potential implications for health outcomes. The key findings from this research indicate that infants born via Cesarean section or those exposed to perinatal antibiotics exhibited altered gut microbiota profiles, particularly a significant reduction in Bacteroides and related metabolic pathways. Using machine learning models, the research demonstrated that the microbiota changes associated with cesarean delivery were generalizable across different infant cohorts from various countries. The relative abundance of Bacteroides was identified as a crucial factor in classifying the mode of delivery. Furthermore, the study found weak associations between early-life gut mycobiome composition and the development of atopic dermatitis and overweight in children. Machine learning methods proved effective in managing the complexity of microbiota data, allowing for robust analysis and classification in pediatric microbiome research. This thesis highlights the potential of machine learning in advancing our understanding of pediatric gut microbiome by facilitating the validation and generalization of microbial patterns across diverse clinical cohorts.