22 November 2010
P S Thiagarajan
Department of Computer Science
National University of Singapore
A system of ordinary differential equations (ODEs) is often used to model the dynamics of a biochemical network. Such systems are difficult to analyse. To get around this, we construct a discrete probabilistic approximation of the ODE dynamics. We do so by discretizing the value and time domain and assuming a distribution of initial states with respect to the discretised value space. We then sample a representative set of initial states and generate a set of trajectories through numerical simulations. Finally, by exploiting the network structure, we encode this set of trajectories compactly as a dynamic Bayesian network.
Consequently, pathway trajectories can be analysed using standard Bayesian inference techniques instead of resorting to a large number of ODE simulations. We have tested our method on a number of pathway models. We have also carried out a combined computational and experimental study of the complement system under inflammation conditions. The results we obtain are quite promising in terms of both accuracy and efficiency.