Back to search

The University of Edinburgh

Integrated analysis of patient gene expression data for precision medicine applications in breast and pancreatic cancer

Abstract

dc:description.abstract

Precision medicine in cancer treatment is an approach aimed to tailor medical care to the individual characteristics of each patient’s disease, considering genetic, environmental, and lifestyle factors to develop targeted and effective therapies. Advanced statistical methods and machine learning algorithms can be leveraged to analyse vast amounts of data, uncover previously hidden patterns and provide new insights in cancer, shedding light on markers of the disease. In this thesis, I focused on two diseases: 1) breast cancer, where biomarkers associated with response to neoadjuvant treatment (NAT) were identified and used as input for the training of supervised NAT response prediction models and the derivation of NAT response subtypes with modern unsupervised methods, and 2) pancreatic cancer, where biomarkers that distinguish normal samples from tumour samples were identified and further investigated for their association with survival, miRNA interactions, enriched biological networks and clinically useful pharmacogenomic interactions. In my first project, I focused on the development of a signature which identifies breast cancer patients who could potentially benefit from NAT. I pooled gene expression data from 10 published studies to identify pre-treatment and on-treatment biomarkers of response to NAT (1861 samples, 1020 patients). Differential gene expression analysis was used to derive a NAT response signature (NRS) consisting of genes with significantly different expression patterns between responders and non-responders. Monte Carlo Consensus Clustering was then used to identify NAT response subtypes based on the NRS markers. The NRS was then used along with additional variables to fit a series of supervised models (logistic regression, decision trees and support vector machines) on a training set using 10x10 repeated cross-validation. Models were subsequently evaluated on a validation set. The model with the highest recorded Area Under the Curve (AUC) on the validation set was selected for evaluation on a held-out test set. I used an additional 4 studies (418 samples, 258 patients) to externally evaluate the performance of our classifier. I identified 166 differentially expressed genes between responders and non-responders, which are involved in cell cycle and DNA repair pathways. I derived two molecular subtypes which were associated with NAT response and other clinical predictors such as tumour subtypes and risk scores. The NAT response classifier achieved an AUC of 0.79 on a held-out test set (134 samples). In one external validation study, AUC increased from 0.64 to 0.82 when focusing on the estrogen receptor positive (ER₊) samples only. In my second project I focused on pancreatic ductal adenocarcinoma (PDAC), which is usually diagnosed at late stages and associated with poor survival. My aim was to identify diagnostic biomarkers from both tissue and blood samples. Publicly available gene expression data from 10 studies with tumour tissue (448 samples) and/or blood samples (128 samples) from PDAC patients were pooled together and analysed for the identification of stage-specific and global differentially expressed diagnostic markers. The list of statistically significantly differentially expressed genes (DEGs) was used to carry out enrichment analysis via active subnetworks and miRNA enrichment analysis. Results from these analyses were then used to identify the most significant genes and pathways and map these to marketed drugs’ pharmacological targets. The same process was replicated for studies with blood samples and results were compared to those from the tissue analysis. I derive, and validate in external cohorts, a Pancreatic Tumour Signature (PTS) consisting of consistently deregulated genes in both tissue and blood samples. Tumour. Notable gene expression deregulation was found in all tumour stages with significant overlap. I identified 820 consistently deregulated genes, which comprised the PTS, in tissue samples of all stages and blood samples. Active subnetwork analysis revealed enriched ribosome, proteasome, adherens junction and cell cycle pathways across all stages and blood samples. My findings suggest that miRNA contribution to PDAC pathology plays a significant role and is likely mediated by distinct miRNAs across stages of PDAC. Stage-specific enriched miRNAs with diagnostic potential included miR-21, miR-29, miR-124, miR-30, for stages 1-4 respectively. By investigating pharmacogenetic interactions of the identified targets with clinically approved drugs, I outlined potential paths for personalized interventions. Importantly, the PTS showed a significant association with survival in TCGA data. By leveraging advanced biomarker analysis and gene expression studies, this thesis has identified markers associated with treatment response in breast cancer and possible early diagnosis of PDAC. These findings suggest targets that could further our understanding of the molecular underpinnings of these cancers that may lead to more effective treatment and outcomes.

Degree

thesis:*
Grantor dc:publisher
The University of Edinburgh
Year dc:date.issued
2024

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Sionakidis, Aristeidis
Advisors dc:contributor.advisor
  • Figueroa, Jonine
  • Cannings, Timothy

Subjects

dc:subject × 9

Rights

Language dc:language.iso
en

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
University of Edinburgh
Base URL
era.ed.ac.uk/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Sionakidis, Aristeidis. Integrated analysis of patient gene expression data for precision medicine applications in breast and pancreatic cancer. The University of Edinburgh, 2024. https://hdl.handle.net/1842/41810