Back to search

University of Edinburgh

Invisible hand of ageing: computational modelling of epigenetic and transcriptomic dynamics

Abstract

dc:description.abstract

Ageing is the primary risk factor for major chronic diseases and is characterised by complex molecular changes at the methylation and transcriptional levels. Elucidating the mechanisms driving these changes is critical for differentiating between healthy and pathological ageing and for identifying targets for therapeutic intervention. This thesis addresses this challenge by developing and applying novel computational frameworks to analyse large-scale cohort data, dissecting the epigenetic and transcriptomic dynamics of the ageing process. First, I addressed limitations in the existing DNA methylation age predictor, the epigenetic clocks. While these clocks can accurately predict chronological age, existing models are often statistically driven rather than biologically grounded, making them susceptible to confounding factors such as smoking and alcohol consumption. Using bulk methylation data from Generation Scotland (∼16,000 participants), I developed ProbAge, a novel statistical model that characterises methylation dynamics by modelling underlying cellular processes. I introduced two distinct parameters: acceleration (the speed of methylation transitions) and bias (global changes in methylation levels). By decoupling these processes, ProbAge demonstrated improved associations with age-related physiological traits and diseases. Additionally, a genome-wide association study identified seven genomic loci associated with the acceleration parameter, reinforcing the model’s biological relevance. Notably, one of these regions highlights MYCN, a gene crucial during embryonic development. Suggesting a potential link between the developmental ‘pathways and ageing. Second, I extended predictive modelling to single-cell RNA-sequencing (scRNA-seq) data, using scRNA-seq data from 981 individuals to measure donor-level gene expression in six homogenous peripheral mononuclear cell subtypes. Using a variety of statistical and machine learning approaches, I constructed a series of cell subtype-specific age predictors and combined them into a stacked ensemble model, which demonstrated greater robustness in predicting chronological age. To further characterize how gene expression trajectories evolve with age, I developed TimeClustering, a novel time-series clustering method. This method improves upon existing techniques by increasing computational efficiency and allowing for a data-driven determination of cluster counts. The downstream analysis revealed five distinct clusters of genes with unique ageing trajectories. Crucially, this analysis uncovered a universal structural constraint: genes that are downregulated during ageing exhibit significantly longer gene body lengths, suggesting that transcriptional length imposes a vulnerability to age-associated decay. Collectively, this thesis provides both new computational tools to measure molecular ageing and specific biological insights into its cell-type-specific nature, identifying as potential targets for future intervention.

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Yang, Emma Joy
Advisors dc:contributor.advisor
  • Vallejos Meneses, Catalina
  • Marioni, Riccardo

Subjects

dc:subject × 8

Rights

Language dc:language.iso
en

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
University of Edinburgh
Base URL
era.ed.ac.uk/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Yang, Emma Joy. Invisible hand of ageing: computational modelling of epigenetic and transcriptomic dynamics. 2026. https://hdl.handle.net/1842/45231