Background
What if we could resolve the epigenome of individual cell types from complex human tissues without ever having to physically separate the cells?
The majority of disease-associated genetic variants exert their effects by modulating gene expression, driving considerable interest in profiling epigenomic variation at scale. Epigenomic patterns, chemical modifications to DNA and chromatin that controls which genes are active, are highly cell-type specific. When studying epigenetics, the tissue choice strongly influences the findings. A critical limitation of existing studies is that bulk tissues (such as whole blood) comprise a mixture of cell types, potentially masking true biological signals.
Historically, resolving this has required the physical isolation of individual cell populations, costly, low-throughput and impractical for population-level studies. Long-read sequencing changes this paradigm by detecting epigenetic modifications alongside genetic sequencing, and by capturing continuous stretches of a single DNA molecule so that each read provides a snapshot of its cell type of origin. We have developed a framework that classifies cell type for individual reads by comparing methylation patterns to reference profiles for different cell types. This enables the generation of cell type-specific epigenetic profiles directly from long-read sequencing data, bypassing the need for experimental cell separation. The project will exploit ~50,000 Oxford Nanopore long-read genomes available in UK Biobank, providing a unique opportunity to resolve cell-type-specific epigenetic variation at population scale.
Objectives
- Evaluate the predictive accuracy and optimal conditions for computational cell-type-of-origin prediction from long-read sequencing data.
- Curate a comprehensive atlas of reference cell-type methylomes to generalise the approach across multiple tissues.
- Apply the framework to large population biobank datasets to generate cellular composition estimates and cell-type-specific epigenetic profiles at scale.
- Leverage these cell-type-resolved profiles to conduct epigenome-wide association studies in large cohorts, identifying epigenetic effects linked to disease in specific cell types.
Methodology
The student will apply our suite of computational tools to blood-based long-read sequencing datasets, benchmarking cell-type-of-origin predictions against reference purified datasets to determine optimal conditions for the approach. This phase will include systematic fine-tuning of the prediction framework. The validated tools will then be applied to large biobank cohorts to generate cell composition predictions and cell-type-specific epigenetic profiles at population scale. Finally, these profiles will be used to perform a cell-type-specific epigenome-wide association study, identifying methylation changes associated with phenotypes of health and disease such as smoking and age, that are resolved to individual cell types.
Interdisciplinarity, Facilities and Resources
The supervisory team work at the intersection of functional genomics, mathematics and computer science, with expertise in large complex genomic datasets, method development, and advanced data science and AI methodologies. The student will work across two world-leading research groups, the Complex Disease Epigenetics Group at Exeter and the Birney Group at EBI. They will need strong mathematical and programming skills, a genuine interest in cutting-edge biotechnologies such as long-read sequencing, and willingness to engage with cell biology to interpret their findings. While predominantly a data-science project, long-read sequencing facilities are available to generate reference datasets where appropriate, and computational work will be supported by local (ISCA) and regional (ISAMBARD 3) HPC clusters. The student will benefit from training through Exeter's Coding for Reproducible Research programme and EBI's suite of resources.
Funded by: MRC Medical Research Council
