Data assimilation, also known as model data fusion, involves combining data and mathematical models to produce optimal predictions. Historically, developed within the context of weather prediction, data assimilation techniques are now being used in health applications, such as real time prediction of optimal treatment strategies.
We have expertise in various techniques for characterising the complexity of signals including the use of entropy methods and novel variants of Lempel-Ziv complexity. We are developing new methods to analyse physiological data based on the method of delays and Takens embedding theory.
Deployment and development of computational approaches for the analysis, integration and interpretation of data for furthering understanding of biological systems at the molecular level.
Examples include the handling and processing of –omics data, inference of molecular interaction networks, and identifying functional modules associated with a given status/response of a system.
Machine learning describes a collection of statistical and computational techniques for identifying patterns within data. More generally, other artificial intelligence techniques such as evolutionary algorithms, neural networks, and deep learning play an important role in identifying large scale networks or bioinformatics problems with small data.
- Advanced machine learning
- Attractor reconstruction methods for physiological data combined with machine learning
- Bayesian machine learning
- Ensemble learning and drop-out learning
- Evolutionary multi-objective machine learning
- Evolutionary optimisation
- Neural networks and deep learning
- Secure machine learning
- The identification of classifiers/predictors of clinical outcomes
Mathematical models can be used to describe core biological mechanisms in order to help develop insight into the way that systems behave and predict behaviour. Ultimately, this may inform policy decisions, for example, when and who to vaccinate for infectious diseases, drug design or optimal light patterns for healthy circadian rhythmicity.
The kind of models that occur depend strongly on the nature of the research question and the particular system. Members of our centre have expertise in developing models in a variety of forms including ordinary differential equations, partial differential equations and agent-based models. Depending on the application, models may be deterministic or stochastic.
Models are analysed using a broad range of techniques including geometric singular perturbation methods, multiple spatial and/or time scale analysis, asymptotics and nonlinear dynamical systems techniques including bifurcation theory.
Members of our centre are also actively involved in developing novel numerical methods, for example for model data fusion (data assimilation).
- Elasticity theory and mechanics of cell growth
- Evolutionary biology
- Open quantum systems
- Population models
- Sleep and circadian rhythms
- Stem cell differentiation modelling
- Systems pharmacology (PKPD)
- Attractor reconstruction methods for physiological data
- Bayesian statistics
- Computerised medical record systems
- Data analysis
- Descriptive statistics
- Design of experiments, including design robustness against observation loss and fractional and fractional factorial experiments
- Gaussian models
- Latent linear models
- Markov models
- Medical statistics
- Monte Carlo methods
- The use of SQL/R-studio