Al modelling of real-world patients: applications in prostate cancer

This thesis investigates the integration and analysis of OMICs data to enhance in silico modelling of real-world patients, with a special focus on prostate cancer. By leveraging publicly available transcriptomics data, this research addresses the challenges of integrating diverse datasets, which is crucial for improving the precision and reliability of disease models. The study is motivated by the rapid advancements in OMICs technologies, which have generated extensive datasets with the potential to deepen disease understanding and improve patient care. However, the effective integration of this data to yield meaningful insights remains a significant challenge. To address this, the research focuses on developing in silico models that integrate Real-World Data (RWD) and OMICs, with a particular emphasis on transcriptomics (specifically microarrays). A novel post-processing technique, MicroArray Cross-plAtfoRm pOst-prOcessiNg (MACAROON), is introduced, which enhances data precision by converting random errors into systematic biases across a large-scale repository of patient data. This repository, which combines gene expression data with clinical and demographic information, enables the creation of customised patient datasets, thereby improving statistical power and the ability to derive reliable biological insights across different patient groups. The practical application of this repository is demonstrated through the identification of novel biomarker candidates for high-grade prostate cancer. By employing Machine Learning (ML) techniques on patient-derived datasets, this research proposes a list of new candidates that demonstrate a strong ability to reduce false positives, while maintaining aspects of performance comparable to existing biomarkers. These biomarkers have the potential to significantly refine prostate cancer diagnostics and inform more effective patient management strategies. In conclusion, this thesis contributes to the field of precision medicine by advancing the integration of RWD and OMICs data into clinically relevant models. The findings presented offer new pathways for improving prostate cancer diagnosis and treatment, underscoring the importance of data-driven approaches in modern healthcare.