Machine learning prediction of multiple myeloma outcomes from complex genome rearrangements

Multiple myeloma is the second most common blood cancer, with highly variable patient outcomes that are only partly explained by commonly used risk-stratifying cytogenetic abnormalities (e.g. 1q gain, IGH translocations and TP53 alterations). Whole-genome sequencing has revealed that myeloma genomes are often highly complex. Many tumours contain intricate rearrangements in which multiple chromosomes break and reassemble in unexpected ways, patterns more typical of solid tumours. We recently discovered that these rearrangements often involve constitutive heterochromatin, genomic regions only fully resolved four years ago by the Telomere-to-Telomere consortium. Our preliminary analyses suggest that rearrangements involving these regions occur in more than 60% of cases and may be associated with faster disease progression.

This PhD project aims to develop a biologically informed prognostic framework for myeloma by integrating heterochromatin rearrangements with clinical outcome data using advanced statistical and machine learning approaches. The student will analyse the international CoMMpass dataset, comprising approximately 1,000 patient genomes, and integrate genomic, transcriptomic and clinical data to develop predictive models of patient risk.

The student will be based at the Institute of Genetics and Cancer, working within a multidisciplinary supervisory team combining expertise in myeloma genomics (Dr Aneta Mikulasova), statistical modelling and machine learning (Dr Catalina Vallejos), and complex cancer genome analysis (Prof Colin Semple). The project will provide training in large-scale genomic data analysis, structural variant detection and interpretation, machine learning for survival prediction, and reproducible computational research using high-performance computing. The project is also linked to the Edinburgh Myeloma Genome Initiative, enabling biological follow-up of key rearrangements using NHS Lothian patient samples and providing exposure to long-read genome sequencing and epigenomic data.

This project is ideal for students interested in data science, genomics and precision medicine, who are motivated to apply advanced computational approaches to understand cancer biology and improve patient stratification.

• Dr Aneta Mikulasova (Primary supervisor, Theme: Biomedical Genomics)
• Dr Catalina Vallejos (Co-supervisor, Theme: Epidemiology & Clinical Trials)
• Prof Colin Semple (Co-supervisor, Theme: Biomedical Genomics)