Multiscale modelling and statistical inference of tumour initiation and evolution in neuroblastoma

Details

Neuroblastoma is the most common solid tumour of early childhood and, in its aggressive forms, remains hard to treat. These tumours arise when embryonic cells - trunk neural crest cells and their derivatives - fail to differentiate correctly. This is driven by chromosomal copy number alterations (such as gains of chromosome arms 17q and 1q) acting together with overactivation of the MYCN oncogene (Saldana-Guerrero et al., 2024, doi:10.1038/s41467-024-47945-7). A major open question is how these genetic lesions shape the striking cell-to-cell heterogeneity of tumours, how that heterogeneity fuels resistance to chemotherapy, and how tumours evolve over time. Answering it requires quantitative, predictive models rather than description alone.

This PhD project will develop and analyse computational models to address this. Working within a larger Cancer Research UK-funded programme, you will build a multiscale, off-lattice agent-based model of tumour initiation and clonal evolution using Chaste, an open-source simulation framework for cell populations co-developed in Sheffield (Cooper et al., 2020, doi:10.21105/joss.01848). The model will go beyond existing stochastic descriptions by explicitly coupling the spatio-temporal dynamics of key signalling pathways to individual-cell rates of proliferation, differentiation, death and movement. To do so, it will build on reaction-diffusion modelling capabilities within the Chaste family (Johnson et al., 2022, doi:10.1093/gigascience/giac051). You will use the model to test competing hypotheses for how neuroblastoma evolves, such as collateral phylogenetic branching versus convergent evolution. Each hypothesis can be simulated under different mechanistic assumptions, and its predicted tumour structure compared against experimental data using phylogenetic tree measures and other summary statistics (Scott et al., 2020, doi:10.1093/sysbio/syz070).

A central theme is confronting models with data. Collaborators in the programme will generate single-cell RNA-sequencing and single-cell lineage-tracing datasets from an experimental human stem-cell model of neuroblastoma and from mouse xenografts, with and without chemotherapy. You will calibrate and validate your simulations against these data, and reconstruct cell-state landscapes to infer plasticity and transition routes between tumour subpopulations. This will employ techniques such as maximum probability flow trees and transition path analysis (Roux et al., 2023, doi:10.1101/2023.12.07.570359). There is considerable scope to develop the statistical inference side of the project, including likelihood-free and Bayesian approaches for fitting stochastic simulators to high-dimensional single-cell data.

The successful applicant will hold or expect a strong degree in mathematics, statistics, physics, computer science, engineering, or a closely related quantitative discipline. Prior biological knowledge is not required. You will gain training in dynamical systems, stochastic modelling, statistical inference, scientific software development and the analysis of large single-cell datasets.

The student will be based in the School of Mathematical and Physical Sciences at the University of Sheffield, supervised by Professor Alexander Fletcher, and co-supervised by Dr Anestis Tsakiridis in the School of Biosciences. You will join a vibrant interdisciplinary community spanning mathematical biology and stem-cell and cancer research, and contribute to widely used open-source software with impact well beyond this single project.