The probability of binding of the transcription regulator to a given promoter is determined by its affinity for the promoter, which is analogous to a binding constant and is often referred to as a promoter strength, and the number of molecules of the regulator. With a low number of regulators molecules, i.e., a low local concentration, the probability of transcription event occurrence is very low and, under a certain threshold, does not occur at all. Transcription starts when the local concentration of the regulator is sufficient, and the rate of transcription grows proportionally to the regulator concentration until a certain level. At this level, the promoter is saturated, and the transcription rate is at its maximum; a further increase in the amount of the regulator does not increase the rate of transcription. The relationship between regulator and gene transcript concentrations has therefore a sigmoidal character. Level of influence, the affinity for binding of the regulator to DNA, can be expressed as a weight, specific for a given promoter and a regulator. This simple analysis leads to a formulation of a model where the rate of expression of a given gene transcript is proportional to the regulator concentration and its weight,Citiolone transferred by a sigmoidal function, and is reduced by degradation. Under this assumption, using an analogy with recurrent neural networks, a simple model of gene expression was derived and extended further in the works, using genome-wide location data and previously reported findings, identified a transcriptional regulatory network for cyclins. The reasons for the choice of the cyclins network were that the network was identified using genome-wide location analysis; the network was relatively small, Chlorothiazidecomprising only 22 genes, and closed; i.e. most of the interactions occurred within the network. The influence of unknown factors from outside the network is thus minimized. There was also a previous experiment with microarrays available that measured expression by sampling relatively densely throughout the yeast cell cycle; this experiment was performed in triplicate allowing for a basic determination of the confidence limits of the measurement. In this paper, we used the yeast cyclins genetic network as a representative case of a gene regulatory network. Together with the microarray kinetic data and ChIP-on-chip measurements, we were able to create a numerical model of this network and analyze its dynamic properties using virtual gene deletion. The cyclins network analyzed here was reconstructed from the experimental data as described in the Methods. Constraints for the creation of the networks used in the analysis were as follows: 1. interaction between regulators and promoter of the regulator gene had to be confirmed experimentally by ChIP-on-chip experiments, here, we used data published by Simon et al. ; 2. the gene expression profile reconstructed using the model had to fall within the 5% confidence interval of the experimentally measured gene expression profiles ; and 3. although the inherent experimental and biological variation does not allow for the creation of a single ‘‘best’’ network, for the purpose of this paper, we had to chose a single network. Therefore, when constructing the network, we considered only those connections that were previously documented in literature. The resulting network is shown in Figure 3. Panel A shows the wiring diagram; panel B shows the same diagram redrawn to demonstrate the causal connection between the genes of the network.