David Lund, Applied Mathematics and Statistics

Antibiotic resistance is increasing among pathogens, representing a serious threat to public health. Bacteria often become resistant by acquiring mobile antibiotic resistance genes (ARGs), which are disseminated via horizontal gene transfer. To anticipate the emergence of new ARGs and limit their spread, we must increase our knowledge about resistance genes that exist in different environments and about their horizontal dissemination among bacteria. The six papers presented in this thesis aim to provide an extensive characterization of the resistome and an analysis of horizontal ARG dissemination. In Paper I, a previously unseen diversity of genes giving resistance to aminoglycoside antibiotics was identified, including 50 previously unknown mobile ARGs carried by human pathogens. In Paper II, the abundance of ARGs, both well-studied and computationally predicted, was estimated in different microbiomes, revealing a widespread presence of previously unknown ARGs across all analyzed environments. In Paper III, a detailed characterization of the resistomes of the human gut and wastewater microbiomes was performed, highlighting the relationship between ARG prevalence in these microbial communities and potential implications for human health. Papers IV and V present a phylogenetic method to identify horizontal ARG transfer between evolutionarily divergent bacteria, which was used to analyze inter-phyla ARG transfers, and combined with machine learning to quantify the impact of different factors on horizontal ARG dissemination. Finally, in Paper VI, the potential use of machine learning to predict the dissemination of emerging ARGs was evaluated. The resulting models showed promise but need further refinement to inform clinical decision-making. Together, the findings presented in this thesis increase our understanding of how ARGs transfer between bacterial species and communities, highlighting the presence in anthropogenic microbiomes and genetic compatibility as key factors associated with successful ARG dissemination. Moreover, the results demonstrate the utility provided by data-driven methods for improving surveillance and diagnostics of antibiotic resistance.