Chris Wallace – firstname.lastname@example.org
I use statistics to make inference about the causes of and potential treatment targets of immune-mediated diseases using generally high-dimensional genetics and omics data. Given the shared aetiology for many immune-mediated diseases, I firmly believe that we should ideally study the whole collection of diseases, which will afford us greater power to identify shared factors as well as reveal disease-specific effects. I have a particular interest in doing this using horizontal integration. Please see our group page for more information about our group and our research.
I value PhD students as an important part of my research group. As well as the camaraderie and support that comes with being part of the PhD cohort in the BSU, they can expect my support to develop a PhD along their own interests and to develop and practice the skills they need to communicate their research. Supervision includes regular one-to-one meetings and group meetings where students can get feedback on their work and learn by feeding-back on others too. My supervision was recognised by the Cambridge University Student Union when I was shortlisted for best PhD supervisor twice (in 2017, 2018) in their Student Led Teaching Awards. You can also learn what two of my students, Stephen Coleman and Anna Hutchinson, thought about their experiences studying for a PhD at the BSU.
Potential PhD Projects
I am open to developing a PhD project around any of the areas or diseases I work on, but propose two specific projects below (one jointly supervised with Brian Tom).
Methodological extensions to colocalisation
Colocalisation aims to determine shared genetic causes of human traits by horizontal integration of genetic association data, and coloc (software and references) is a popular colocalisation method that has widespread adoption in academic research and industry. It takes a Bayesian formulation which allows us to leverage the wealth of knowledge of human genetics now accumulated in setting prior probabilities for key events. However, the coloc method in its current form makes several simplifying assumptions: that each genetic variant has an a priori equal chance of being causal for a trait and that the prior probability any pair of traits share a causal variant is fixed and known. In both cases, we now have data which we could leverage – for the first, expression quantitative trait loci (eQTL) datasets tell us that genetic variants near a gene are more likely to alter its expression, and so our priors should depend on distance. For the second, we now have data from sufficient pairs of traits that we could jointly analyse them in a hierarchical framework. This project would explore these extensions, and how other covariate information could also be included in improved prior formulations. It would require the student to have an interest in Bayesian modelling, and ability, or interest in learning, to code in R and Rcpp. Depending on the student’s interest, this project has the potential to be more theoretical or applied, but is likely to have direct impact on how millions of colocalisation analyses are conducted.
The molecular relationships between child and adult arthritis
This project is jointly supervised with Brian Tom.
Arthritis is an immune-mediated disease characterised by inflammation of the synovial tissues around joints that leads to chronic pain and progressive disability. Childhood arthritis is the most common chronic rheumatic disease of unknown aetiology in childhood and considered a collection of seven subtypes, with several considered paralogs of better studied adult arthritis types. For example, genetic data suggest rheumatoid factor positive childhood arthritis is most like adult rheumatoid arthritis (Hinks et al). However, the division into subtypes has been controversial. One important but unresolved question is whether adult rheumatoid arthritis and (some subtypes of) childhood arthritis are the same disease at different ages?
Chris Wallace and Brian Tom have led data analysis workstreams for stratified medicine consortia in childhood (CLUSTER) and adult (RA-MAP) arthritis consortia, which have collected a wealth of molecular data characterising the immune system in hundreds of child and adult patients early in their disease course. We propose that a systematic comparison of these high-dimensional molecular data can be used to address this central question. Further important questions include:
- If the diseases are not the same, what are the similarities/differences in terms of immune profile?
- If they are “close enough”, then what information from the larger adult studies can we use to better understand the childhood disease (and vice versa)?
- Can the molecular data be used to better align childhood arthritis subtypes with adult rheumatoid arthritis?
- Given the known age-dependent differences in the immune system, what are the differences in how the same treatment given to adults and children affects immune gene activity?
There are interesting statistical challenges in this project, not least in dealing with age (because the immune system changes with age, even in healthy individuals), inevitable “batch effects” between adult and child samples which were collected and processed differently, and the dimensionality of the covariate space. We also need to consider statistical definitions of what “the same” means, and what would it look like in this kind of data. There are always multiple possible approaches, but we envisage that the use of clustering and latent variable modelling techniques will be relevant, as may other dimension reduction techniques.
How to apply
For details of the MRC BSU application process please see How to apply
To be considered for funding applications need to be submitted to the University of Cambridge application system by 23:59 (GMT) on January 5th 2023