Abstract
Oesophageal adenocarcinoma (OAC) incidence in Western countries has increased in the last decades. In the UK, 70-80% of patients are diagnosed with OAC at an advanced stage with metastatic disease. The late diagnosis impacts the success of OAC treatment resulting in a five-year survival rate of ~15%. This poor patient survival is further compounded by a high level of genetic inter-tumour heterogeneity. Despite international consortia extensively investigating the driver landscape of OAC, a significant proportion of OACs are partially explained by too few cancer driver genes.Consequently, the aim of this thesis is to complete the driver repertoire of a cohort of 675 OACs by employing a machine learning-based algorithm, sysSVM2, and to investigate the drivers’ role in OAC development. sysSVM2 learns about the molecular and systems-level properties (SLPs) of well-known cancer drivers to score and predict new ones in individual patients. Molecular properties define the driver alterations of individual OACs. SLPs describe the genes’ evolutionary origin and central role within the cell.
Given that sysSVM2 prioritises drivers in single patients using a scoring system, we used a functional approach to compare two models predicting varying number of drivers per tumour. We estimated that five drivers per sample fully explained the development of OAC. Therefore, we compiled a comprehensive list of cancer drivers for all samples and investigated their frequency and functionality. Newly predicted drivers, some of which were targetable by oncological drugs, were preferentially sample-specific and hit immune- and DNA repair-related pathways. Well-known drivers were recurrent across samples and preferentially perturbed proliferative and invasion pathways.
Finally, we found that OAC clinical stratification is sustained by different pathways that are active in a stage-specific fashion throughout the development of OAC.
| Date of Award | 1 Sept 2022 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Jesper Lagergren (Supervisor) & Francesca Ciccarelli (Supervisor) |