Next: Spatial models Up: Continuous models Previous: Neural network models
Hybrid models
In the last decade, a number of models have been developed that take a hybrid approach to modelling gene regulatory networks. In these models, biochemical processes that are characterised by sharp thresholds are represented by Boolean elements, while genes whose activations vary more continuously with time, or for which intermediate levels of activation are significant, are modelled continuously. Early work in this direction was carried out by McAdams and Shapiro, who characterised the phage-
circuit in terms of an electrical circuit, incorporating discrete and continuous elements, time delays and feedback dynamics [82].
A similar approach has been developed by Eric Davidson and colleagues, who have taken a strongly integrative approach to modelling the regulatory networks responsible for development [28,29]. This work has ranged from detailed characterisation of the logic underlying individual regulatory interactions [145] through to a network level view of regulatory dynamics [30]. One of the novel conceptual distinctions drawn in this approach is between the ``view from the genome'' and the ``view from the nucleus'' [10]. The former represents all possible regulatory interactions that genome encodes, while the latter restricts itself to those that are active in a particular cell at a particular time.
A key feature of this approach is the use of both continuous and Boolean functions, which result in a model lying somewhere between a continuous kinetic model and a Boolean model. The primary advantages of this level of abstraction is the clarity with which complex circuits may be represented, computationally simulated and empirically validated. The main cost associated with this approach is the loss of many of the analytical techniques that can be applied to more ``pure'' continuous or logical models. As the motivation for this work lies more in the direction of integrating and guiding experimental data, with less focus on abstract theoretical results, this trade-off is considered to be acceptable. An emphasis has been placed on the role of models as ``the developmental biologists essential organizer for getting causal relationships between genes straight'' [19].
Next: Spatial models Up: Continuous models Previous: Neural network models Nic Geard 2004-05-06