Choi, SungWoo; (2025) Investigating muscular dystrophy using advanced tissue modelling and machine learning. Doctoral thesis (Ph.D), UCL (University College London).

Text SungWooChoi_PhD_Thesis_corrections_approved.pdf - Accepted Version Access restricted to UCL open access staff until 1 August 2026. Download (45MB)

Abstract

Muscular dystrophies are a heterogenous group of muscle wasting disorders and severe forms are associated with early loss of ambulation and premature death. Although currently incurable, several clinical and pre-clinical therapies are being developed. To facilitate regeneration of degenerating skeletal muscle tissues, transplantation of myogenic cells has been attempted, albeit with limited efficacy. One key factor contributing to lack of clinical success in experimental therapies for muscular dystrophies is the limited predictive capacity of existing animal models. Hence, this PhD project aimed to: 1) overcome key limitations of muscle cell therapy by developing a humanised, in vitro model for studying myogenic cell migration, engraftment and functional improvement. 2) Utilise the potential of machine learning to identify cellular phenotypes relevant for muscular dystrophies in an unbiased manner. Human induced pluripotent stem cell-derived myogenic progenitors (hiMPs) were treated with cell fate modulators DLL4 & PDGF-BB. Acutely injured 3D bioengineered muscles, combined with live-imaging and single-cell tracking revealed increased migration of hiMPs following DLL4 & PDGF-BB treatment. The functional relevance of myogenic cell transplantation was evaluated using a novel approach based on machine learning to analyse the contractile properties of bioengineered muscles, demonstrating differentiation and functional integration of donor cells in muscle constructs. The potential for further utilising deep neural networks for identifying fine-grained cellular phenotypes was explored in the context of a severe form of muscular dystrophy caused by dysfunctional nuclear envelope. A cross attention-based U-Net was used to perform instance segmentation of overlapping and abnormally shaped nuclei and a model-based similarity score was explored to evaluate unsupervised feature extraction methods of single nuclei images. The combination of advanced, in vitro tissue modelling and in silico machine learning techniques could provide future avenues for studying complex dystrophic phenotypes and accelerate therapeutic development.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as JSONCSVXML

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item