Statistical and deep learning approaches to phenotype discovery for precision oncology

Abstract

This thesis demonstrates the potential of advanced deep learning methods, specifically deep autoencoders, to extract features and reveal potential disease phenotypes in colorectal cancer in an unsupervised manner. Through the creation and validation of these tools, this research lays foundations for more tailored treatment regimens based on patient-specific disease profiles. This work supports the feasibility of applying deep learning to complex healthcare data, addressing key obstacles of data quality, dimensionality, and interpretability, which are crucial for clinical translation.

Thesis is embargoed until 31st July 2028.

Date of Award	Jul 2025
Original language	English
Awarding Institution	Queen's University Belfast
Sponsors	LifeArc, Engineering and Physical Sciences Research Council & Health Data Research UK
Supervisor	Paul Miller (Supervisor), Helen Coleman (Supervisor) & Ian Overton (Supervisor)