Glioblastoma multiforme (GBM) is the most prevalent and lethal form of brain cancer in adults. While it remains incurable, various treatment options exist for select patients. Methylation of the O6-methylguanine-DNA methyltransferase (MGMT) promoter serves as a critical biomarker in GBM, correlating with patient responses to temozolomide, the standard chemotherapy for this condition. Currently, MGMT methylation is assessed through targeted laboratory assays post-surgery, with results typically available one to two weeks later. The Vilhelm Magnus Laboratory has recently pioneered ultra-rapid diagnostic tools for brain tumors utilizing methylation data derived from low-coverage nanopore sequencing. Nanopore sequencing can read DNA methylation directly, but at low genome-wide coverage, the MGMT promoter itself is often missing. This creates an interesting challenge: can MGMT methylation status be inferred indirectly from broader genome-wide methylation patterns?
This student project will explore how machine learning methods might be applied to this problem. A set of 200 deeply sequenced whole-genome methylation profiles from GBM samples with known MGMT status will be used as a training resource. From these, sparse datasets can be simulated and different strategies for feature construction can be investigated, for example, summarizing methylation in large genomic bins or exploring global methylation signatures. The student will then evaluate a range of models, from simple statistical classifiers to more advanced machine learning approaches such as neural networks, to see which best captures patterns that are predictive of MGMT status under sparse conditions.
The goal is to develop a proof-of-concept framework for MGMT methylation prediction from sparse nanopore methylation data. The student will gain hands-on experience in computational biology from a real-world, directly clinically applicable project.
Supervisors: Dr. Skarphedinn Halldorsson (external main supervisor) and Prof. Torbjørn Rognes (internal co-supervisor)
Location: Vilhelm Magnus Laboratory, Institute for Surgical Research, Oslo University Hospital (Rikshospitalet)