IMA Seminar Series 2008 / 2009

The Affinity Propagation algorithm is a novel clustering method proposed by Frey and Dueck in 2007 [1], which combines advantages of both affinity-based clustering (like Hierarchical algorithms) and model-based clustering (like Expectation-Maximisation). The method takes as input measures of similarity between pairs of data points and real-valued messages are exchanged between data points until a set of centres (called exemplars) and corresponding clusters gradually emerge. The Affinity Propagation algorithm also provides a procedure to determine the number of clusters to be considered. In this talk, the approach proposed by Frey and Dueck will be presented and results coming from the application of the technique to breast cancer data sets will be reported and compared with the original ones. Results from Affinity Propagation are consistent with the results already obtained having the advantage of providing an indication about the number of clusters to consider in the analysis. Moreover, results also provide novel insights, with respect to conventional approaches. Seminar by Jamie Twycross: Executable Modelling for Plant Systems Biology My research explores the application of executable models to systems biology. In collaboration with biologists, I am developing a number of multiscale executable models investigating the transport of plant hormones, and the effects of these hormones on root phenotypes. Executable models differ from mathematical models in that an executable model defines a procedure which is applied to transform the modelled system from one state to the next. The model is simulated, rather than solved, to determine the time evolution of the modelled system. I am using a stochastic P system framework to describe my models. Models can contain many hundreds or thousands of compartments, molecular species, and reactions, corresponding to complex multicellular tissues, signalling and transportation pathways. Specification of such models can be eased though the use of modularity, which allows commonly repeated sets of rules (modules) to be specified only once. Using P systems, executable models of biological systems can be constructed without the need to write down complex systems of differential equations, permitting preliminary studies to be performed in silico and reducing time spent in the laboratory. I will present a general introduction to executable models and P systems for plant systems biology, overview the software I have developed, and outline some of my recent work in developing executable models of plant growth. Tutorial by Peer-Olaf Siebers: Simulation with AnyLogic The focus of my tutorial will be on AnyLogic, a multi-paradigm …