Advancing multi-omics statistical methods for precision medicine

Abstract

Recent developments in biotechnology have empowered scientists to investigate the molecular profile of biological samples, at an unprecedented scale and resolution. This carries vast potential for precision medicine where we can better understand the etiology of diseases and discover biomarkers. However, this requires statistical methods, tailored to the data and context. There remains a lack of tools and workflows that exploit different data modalities for clinical applications. This thesis proposes novel workflows, methods and software, to take advantage of multi-omics data, bringing progress towards precision medicine.

We developed a cross-species, lipidome-wide and genome-wide analytical workflow to compare between humans and chimpanzees. This elucidates evolutionary causes for coronary artery disease resilience, shedding light on the heterogeneity in disease susceptibility among humans. This workflow provides a novel approach to draw connections between genomics, intermediary omics, and a clinical phenotype, allowing us to uncover underlying causes of cohort heterogeneity.

We built a framework (MultiP) to construct multi-platform clinical pathways in an automated, data-driven way. Our framework also provides a range of tools to interpret, visualise and compare the pathways, enabling their evaluation on accuracy and cost. This framework facilitates the translatability of multi-omics technologies into a clinical application, helping to realise the potential of these technologies for precision medicine.

Finally, we show that our MultiP framework can be used to assess the fairness of clinical prediction algorithms. We are able to optimise confidence thresholds to ensure fair predictions, even if individual models are biased. Furthermore, we develop a simulation framework to model population-level data in a multi-platform setting. This provides a strategy to use multi-omics technologies to improve fairness, allowing an equitable implementation of precision medicine.