Decoding Cancer Epigenetics: Bioinformatics Thesis Projects

Epigenetic processes are essential for regulating gene expression in specific tissues and conditions. When these processes malfunction, they can contribute to diseases such as cancer. Our research is at the forefront of developing and evaluating DNA methylation-based classification tools, and our previous work has led to the implementation of new diagnostics in the clinic. In addition, we have established advanced model systems to study brain tumour cancer stem cells, both in vitro and in vivo.

We now offer several master thesis projects that provide an excellent opportunity to gain hands-on experience in innovative cancer research. If you're eager to contribute to cutting-edge diagnostics and cancer biology, we invite you to visit our homepage for more information https://www.gu.se/en/research/helena-caren. The projects are most suitable as 60 hp projects but could potentially be adjusted to 30 hp.

Project 1: Diagnosing Cancer of Unknown Primary (CUP) Using DNA Methylation Classifiers

Cancer of Unknown Primary (CUP) represents a challenging clinical entity where metastatic disease is present, but the tissue of origin cannot be identified using standard diagnostic workups. Correctly determining the tumour’s primary site is crucial for selecting the most effective treatment. DNA methylation profiling has emerged as a powerful approach for tumour classification and tissue-of-origin prediction, offering new opportunities to improve CUP diagnostics.

In this master thesis project, you will analyse DNA methylation data from a local cohort of CUP samples and evaluate the performance of multiple available tumour-origin classifiers, including our in-house method. You will benchmark and compare classifiers across relevant clinical and technical factors, assess prediction confidence and failure modes, and explore how methylation-based predictions align with available pathology and clinical information. The goal is to identify strengths and limitations of current approaches and contribute to a more robust workflow for CUP diagnosis.

This project is an excellent opportunity for students interested in computational biology, translational cancer research, and clinical applications of machine learning. Basic knowledge in R (or general programming skills, or strong motivation to learn) is recommended. Experience with methylation array/NGS data, classification methods, or reproducible analysis workflows is a plus, but not required.

Project 2: Multiomics Analysis of Paediatric Glioma Stem Cells

High-grade gliomas remain incurable, underscoring the urgent need for new treatments. To develop effective therapies, it's essential to understand the molecular drivers that propel these aggressive tumours. We have generated unique paediatric glioma stem cell cultures, with neural stem cells serving as a comparison, to explore these critical pathways. In this master thesis project, you will analyse our generated multiomics datasets to identify tumour drivers and decipher the complex regulatory networks underlying paediatric gliomas. This project is an excellent opportunity for those passionate about bioinformatics and cancer research to contribute to the discovery of novel therapeutic targets. A basic knowledge in R (or general programming skills or alternatively a strong motivation to learn), is recommended for the project.

Project 3: Leveraging a Large Genomic Language Model for DNA Variant Interpretation and Epigenomic Analysis in Paediatric Brain Cancer

Paediatric brain tumours, including medulloblastomas and gliomas, frequently harbour genetic mutations that alter chromatin states and transcriptional programs. Understanding how DNA variants impact DNA methylation and RNA expression is critical for identifying potential biomarkers and therapeutic targets. However, the mechanistic links between genetic variants, epigenomic alterations, and transcriptomic changes remain poorly understood. Traditional computational methods often require extensive task-specific fine-tuning and large annotated datasets, limiting their ability to generalize across different cancer types. Evo 2, the largest-scale fully open language model to date, presents a unique opportunity to overcome these limitations. In this study, we propose to leverage Evo 2 to analyse a large cohort of paediatric brain tumours with corresponding clinical data. By integrating multiomics data, we aim to improve the interpretation of somatic and germline mutations, identify novel regulatory elements, and provide deeper mechanistic insights into tumour biology.

This project is an excellent opportunity for students interested in bioinformatics, machine learning, and translational cancer genomics. Basic knowledge in R or Python (or general programming skills) is recommended. Familiarity with genomics/epigenomics data (e.g., VCFs, methylation arrays/NGS, RNA-seq), reproducible analysis workflows, and/or machine learning methods is a plus, but not required.