Next: Stochastic modelling approaches Up: Stochastic models Previous: Stochastic models
Noise from within and without
As described in Section 2, gene activation is controlled by molecular signals, some of which are proteins producted by other transcription events. In general, genes are activated when the concentration of signal molecules crosses a threshold. Although many mathematical models make the simplifying assumption that proteins are produced at a continuous rate, evidence suggests they are actually produced stochastically in short ``bursts'' [81]. Therefore, the time taken for a concentration to reach it's critical threshold will vary stochastically. This variability time delay length can result in significant differences in the timing of similar events across an otherwise homogeneous population of cells. Due to the complex nature of interactions between regulatory elements, it is possible that individual cells may take different branches of regulatory pathway. Another potential source of stochasticity in the timing of events arises from the fact that, if a particular signal is represented by only a very small number of molecules, random molecular fluctuations may affect the time taken for a signal to be transferred [93].
Stochastic events in gene expression have several implications [97]. First, identical systems provided with similar inputs may produce different outputs as a result of stochastic elements of their regulatory mechanism. Second, it is likely that the evolution of gene regulatory networks has been driven in part by the requirement to produce deterministic outputs from a system constructed from noisy components operating in a noisy environment. While it is the case that in many situations regulatory systems are able to produce ordered results from chaotic starting points, in other instances, noise is exploited to the benefit of the system. The final implication is that deterministic modelling techniques may be insufficient to capture some of the dynamics of inherently noisy systems [75,97].
Mechanisms by which the effects of noise may be diminished include negative and integral feedback (intensifying intermediate frequencies and dampening high and low frequencies), redundancy mechanisms and regulatory ``checkpoints''. In some systems, noise is amplified and used to generate heterogeneity in a population and hence increase diversity. Simulations have also found that complex systems involving many interacting feedback loops may be stabilised by noise [97]. In several studies [14,138,66], systems have been found whose robustness to noise appears to be a systemic product of network structure, rather than any explicit combinations of parameter settings or attenuating mechanisms.
Next: Stochastic modelling approaches Up: Stochastic models Previous: Stochastic models Nic Geard 2004-05-06