Biological systems' dynamics represent a prime example of complex behaviour, providing a powerful motivation and application domain for computational modelling. Of central importance is the learning problem, i.e. how can we reconstruct information about the underlying state of the system from noisy measurements; this is often an essential first step to further modelling and analysis, and to attempts to predict and control system behaviour. The complex behaviour observed in biological dynamics is often the result of a non-trivial network of interactions between individual components of the biological system: this is true at a variety of scales and level of abstractions, from physiology, to neuroscience, to molecular interactions within a single cell.
Modelling of biological networks is often divided into two sub-problems: given the structure of the interaction network (i.e., which components interact with each other), one may be interested in reconstructing the dynamics of the individual nodes from partial observations of (some of) the nodes' states, using techniques from system identification to estimate states and parameters of the system. In this case, the known structure allows researchers to utilize plausible models of system dynamics, often involving a continuous time representation and/ or nonlinearities. On the other hand, much research in the last decade has focused on learning the structure of the underlying interaction network, assuming a greatly simplified model for the dynamics of the nodes (typically an autoregressive linear Gaussian model) and focussing computational power on the combinatorial problem of learning network structure. In most cases, the structure of the network is assumed to be fixed, but recently time-varying networks have been proposed, which enable richer dynamics and may better reflect underlying biological changes in processes such as development or adaptation.
In this workshop, we aim to bring together experts in dynamical systems and structure learning in order to promote cross fertilisation between the two communities, thus generating novel collaborations that will lead to a more integrated and holistic modelling of biological networks. The workshop will be include four keynote talks by leading international experts, as well as shorter contributions from the workshop participants. Plenty of time will be allowed for discussion and definition of the key problems ahead for the community.
Dirk Husmeier (Glasgow)
Heinz Koeppl (ETH Zurich)
Olga Troyanskaya (Princeton)
Eric Xing (CMU)
The workshop registration will start at 11:30am on Tuesday, 25 June.
Attendance of the workshop is free. At this point we invite contributed talks and posters. Please register and submit your abstract using the following website: Workshop Registration
Guido Sanguinetti (Edinburgh)
Edoardo Airoldi (Harvard Statistics)
Matthias Hennig (Edinburgh)
BBSRC International Partnering Award between Edinburgh and Harvard (to G.S.)