Debugging lung diseases: applying mathematical techniques for precision medicine on the pulmonary microbiome through modelling, data integration and machine learning

Abstract

Background: The pulmonary microbiome plays a crucial role in chronic respi- ratory diseases. However, the translation of mathematical approaches for clinical value remains limited. This thesis aims to bridge the gap between mathematical techniques and clinical sciences to explore the potential of the pulmonary micro- biome for precision medicine. Results: Integrative-microbiomics (https://integrative- microbiomics.ntu.edu.sg), a novel approach integrating multiple microbiomes, en- hances patient stratification in bronchiectasis. Microbial association networks (in- teractome) and its network-based metrics outperform microbial abundance alone in associating with clinical outcomes, and identifies a dysregulated ‘gut-lung” axis in high-risk bronchiectasis. A novel application of Compositional Data Analysis (CoDA) to the pulmonary microbiome in COPD reveals time-dependent effects of antibiotics not captured by traditional microbiome analysis. Conclusion: This thesis highlights the essential role of innovative mathematical techniques in pul- monary microbiome analysis. It contributes to data-integration, network science, compositional data analysis, patient stratification, and clinical interventions for chronic respiratory diseases.