Medical and Regulatory Genomics

The genes embedded in your genome have complex patterns of activity and particular constellations of genes must be active in particular cells and at particular times for biological processes, such as embryonic development, to conclude successfully. Our group is interested in the fundamental biology of gene regulation: when, how and why genes are turned on and off. We advance understand of the mechanisms underlying gene regulation, using computational analyses of datasets measuring the activities of many thousands of genes. This can provide new insights into human evolution itself, and also helps us to interpret disease processes with disrupted regulation, such as cancers and developmental disorders. Human genomes vary at millions of DNA sites between individuals, but often we do not know which variants matter most to our biology. Ultimately we want to develop predictive models, based upon our knowledge of gene regulation, that help us to forecast the effects of variants in health and disease.

We use computational approaches to reveal the interdependencies between transcription, chromatin and the underlying genomic sequence - and to investigate transcriptional control mechanisms and their disruption in disease. We have an established track record in studying the roles of chromatin structure in mutational spectra, and in how tumour genomes evolve during cancer progression. We aim to test hypotheses addressing three broad aims.
Our first aim is to reveal new functional interactions between chromatin structure and transcriptional activity. We still lack a convincing and quantitative account of the relationships between nuclear organisation, chromatin structure and expression dynamics over time. Developing such models will have important consequences for our understanding of noncoding variation in disease.
Our second aim is to understand the roles of chromatin structure and mutational bias in the evolution of regulatory sites, within the catastrophe strewn genomes of tumours, and during the evolution of isolated human populations. Using a novel approach we have demonstrated unexpectedly high mutational loads at active regulatory sites in most cancers, relative to matched control sites. We are extending this approach to allow us to test the hypothesis that certain site classes accumulate lower levels of mutation than expected, suggesting purifying selection in tumours. Classes showing unexpected increases or decreases of mutational load, suggesting selection in specific tumour types, may suggest new candidate targets for therapeutics. In parallel we will extend our regulatory mutational load metrics to comparisons between isolated human populations (eg from the Shetland archipelago) and cosmopolitan populations, to detect regulatory site classes with unexpected loads, correlated to phenotypic traits. We aim to gain novel insights into the impact of rare regulatory variants on quantitative traits relevant to disease predisposition, ultimately aiding the development of ‘personalised’ medicine based upon genome sequencing.
Finally, we aim to explore the frequencies and roles of regulatory domain lesions in developmental disorders and cancers. Structural variants (SVs) are known to play critical roles in tumourigenesis, but there are currently no methods to reliably disentangle mutational bias from selection operating on SVs in cancers. Such methods are essential to establish which SVs host potential new therapeutic targets, and which are simply well tolerated by tumours. We are developing new models based upon experimental measurements of the mutational spectra in relevant cell types, to estimate the expected spectra in tumours, and rigorously infer candidate breakpoints that may be under selection.