Uncovering hidden genomic rearrangements to improve risk prediction in multiple myeloma

Background

Multiple myeloma (MM) is a malignancy of bone marrow plasma cells and the second most common blood cancer. In the UK, there are 16 new diagnoses and eight deaths daily. Despite therapeutic advances, MM remains incurable. The five-year survival rate is 55% but in high-risk cases life expectancy can be as short as three years. A central challenge is the unpredictable clinical trajectory, largely driven by heterogeneous genomic instability. Classical risk stratification is based on primary events (IGH translocations or hyperdiploidy). However, more recent evidence highlights that secondary genomic events (e.g. 1q gain) are stronger predictors of patient outcomes.¹

Uniquely among blood cancers, MM shows a high prevalence of complex rearrangements (>60%)², where multiple genomic loci recombine in a way that is more typical of solid tumours. Whole-genome sequencing revealed these events but their impact on disease pathogenesis and prognosis remains unclear due to detection and interpretation challenges. Indeed, most complexity in MM genomes is still uncharacterised.² One personalised prognostic model has recently been published.³ It incorporates chromothripsis (a complex event where chromosomes shatter and are reassembled in highly rearranged order) yet overlooks much of the 8% of the genome newly resolved by the Telomere-to-Telomere consortium. Our preliminary analyses show that rearrangements involving these newly characterised heterochromatin regions occur in ~60% of MM cases and associate with faster disease progression.⁴

This interdisciplinary project will fill a critical gap by integrating analysis of heterochromatin rearrangements with machine learning modelling, delivering a biologically informed prognostic framework to improve MM risk stratification.

Aims

This project aims to develop a biologically informed prognostic framework for MM by integrating genomic complexity, including heterochromatin rearrangements, with clinical outcome data (overall/progression-free survival) using advanced machine learning approaches.

Aim 1. Curate and integrate genomic and clinical data from the CoMMpass study: Extract and harmonise whole-genome, transcriptome and outcome data, identify complex and heterochromatin associated rearrangements, engineer features to generate a feature matrix suitable for machine learning model development.

Aim 2. Develop and validate machine learning prognostic models for MM. Train statistical/machine learning models using nested stratified k-fold cross-validation with strict leakage control, incorporating conventional genomic features, complex abnormalities and novel heterochromatin rearrangements. Assess model performance across patient subsets.

Aim 3. Prioritise and biologically explore key rearrangements identified by the models as important predictors of risk. Pinpoint genomic rearrangements most strongly associated with poor prognosis. Investigate their epigenomic consequences using locally collected patient samples from the Edinburgh Myeloma Genome Initiative and collaborative Patient Derived Xenografts (PDXs, University of Miami). Generate proof-of-concept biological insights that may inform future therapeutic strategies.

Training Outcomes

The student will gain interdisciplinary expertise at the interface of computational biology, precision medicine and quantitative sciences:

Cancer genomics and data curation: detection, annotation and interpretation of structural variants, with emphasis on heterochromatin instability

Computational analysis: building computational pipelines, HPC usage, version control (GitHub), programming (R, Bash, Python)

Advanced technologies: long-read whole-genome sequencing and epigenomic data analysis (DNA and histone modifications), variant phasing, multiomic integration

Quantitative and statistical methods: survival analysis, statistical modelling, big data handling, machine learning for risk prediction, reproducibility/transparent reporting (e.g. TRIPOD guidelines)

Sensitive data governance: handling controlled-access human genomic data, complying with ethical frameworks (NHS/NIH), adhering to FAIR principles

Precision medicine perspective: linking genomic features to patient outcomes, building clinically relevant prognostic models

Patient/public engagement: engaging with patients and the public through collaborations with DKMS UK, PPE events such as Pint of Science

Academic/professional skills: developing transferable skills including scientific writing/communication, oral/poster presentation, mentorind junior students/contribution to teaching

References

1. Avet-Loiseau H et al. International Myeloma Society/International Myeloma Working Group Consensus Recommendations on the Definition of High-Risk Multiple Myeloma. J Clin Oncol. 2025 Aug 20;43(24):2739-2751. doi: 10.1200/JCO-24-01893.
2. Rustad EH et al. Revealing the impact of structural variants in multiple myeloma. Blood Cancer Discov. 2020 Nov;1(3):258-273. doi: 10.1158/2643-3230.BCD-20-0132.
3. Maura F et al. Genomic Classification and Individualized Prognosis in Multiple Myeloma. J Clin Oncol. 2024 Apr 10;42(11):1229-1240. doi: 10.1200/JCO.23.01277.
4. Mikulasova A et al. Multiomics Data Integration in the Complete Myeloma Genome Reveals Frequent Centromeric Rearrangements and Their Epigenomic Consequences. Blood 2024; 144 (Supplement 1): 768. doi: 10.1182/blood-2024-200043.