Research / Interests
Overview
Although my background and particular expertise are both in Bayesian statistical inference, I have a broad range of research interests, cutting across mathematics, statistics, probability theory, modelling, computing science, and molecular and systems biology.
Most of my current research interests involve applications of Bayesian statistics to a variety of challenging problems in Molecular biology, Bioinformatics and Systems Biology. I am especially interested in parameter inference for dynamic biological models, and the use of approximate models and (stochastic) model emulators for rendering computationally prohibitive algorithms for "big data" more tractable. I'm also interested in the way that HPC and Cloud computing technology can help here too. A recent invited talk I gave at the LMS/EPSRC Durham Symposium on the Mathematics of Data Assimilation is available on-line, and this gives an overview of some of my current interests, as do recent entries on my research blog.
Keywords
Bayesian analysis, scalable Bayesian computation, and Bayesian software development; Dynamic and graphical models, and Markov processes; Statistical bioinformatics and stochastic models in computational systems biology; Cloud computing and e-Science technology for computational statistics and big data; Efficient Bayesian algorithms and Bayes linear methods.
Heterogeneity
I recently held a BBSRC Research Development Fellowship to investigate Integrative modelling of stochasticity, noise, heterogeneity and measurement error in the study of model biological systems, and this topic remains the main focus of my research. The project involved a combination of statistical modelling and experimental lab work exploiting model organisms Bacillus subtilis (a gram-positive bacterium) and Saccharomyces cerevisiae (budding yeast). I am still working closely with the Bacillus lab of Leendert Hamoen and the yeast lab of David Lydall.
Budding yeast
In the Lydall lab
we are using robotic genetic screens to identify mechanisms involved in
DNA maintenance and the cellular response to telomere
uncapping.
A typical experiment can generate up to 40,000 high resolution images - this is "big data". Each image contains information on 384 different yeast mutants. Our image analysis pipeline therefore leads to around 15 million colony growth measurements, corresponding to around 400,000 time series. Analysis of this data poses considerable computational, as well as statistical, challenges. We have developed a computational analysis and modelling pipeline that we call Quantitative Fitness Analysis (QFA), allowing us to discover thousands of previously unknown genetic interactions.
The slides of a recent talk I gave on the statistical analysis of the data generated by this experimental technology are available.
Above are some photos relating to the yeast robotic
genetics, and some
more are available on flickr.
Bacillus subtilis
In the Hamoen lab
we use live cell imaging and flow cytometry to study control and
regulation of key phenotypic decision processes such as competence
development, sporulation and cell division. We are currently focusing
experimental work on the regulation of cellular motility.
Above are
some images showing heterogeneity in gene expression by using a GFP
reporter, and some
more are available on flickr.
The rationale for undertaking this project can be found in my recent Nature Reviews Genetics article: Stochastic modelling for quantitative description of heterogeneous biological systems.
