As panel B shows, there exists a first level of genes that are not controlled within the network, and the bottom-most layer represents the terminal nodes of the network, which do not control any other genes in the network. The reconstruction using the model allowed us to compute the parameters that best fit the experimental time series for each of the connections. Therefore, the network was fully characterized by the parameters, and by knowing the expression profiles of the first layer, it is possible to directly compute expression profiles of all of the remaining genes. This example is only one illustration of the influence of weighted regulator concentration on the expression level of a target gene. In this case, none of the virtual mutations led to the complete repression of the target gene. If we consider Boolean relationships, then deletion of FKH1 would cause repression of CLB4. For deletion of FKH2, SWI5 would be repressed and CLB4 over-expressed. For MCM1, SWI6, and CLN3, we cannot make a prediction because all of them result in a change of SWI5 control, which cannot be estimated from Boolean rules unless we know a Boolean function for multiple regulators. Even more striking is the effect of CLN3, which indirectly controls all genes of the bottom layer. However, its deletion has an effect only on CLB4; in all other cases, its deletion is compensated within the pathway. Although the ChIP-on-chip data indicate possible binding of CLN3 to the promoter of MBP1, Ceftazidime the w computed for this connection is also very low. Deletion of CLN3 therefore does not influence other genes. Literature search indeed indicated posttranslation control instead of the transcrip- tional control. MBP1 was reported as regulated by CLN3. As the presented model does not include posttranscriptional events, such connection cannot be discovered and the low value of wCLN3-MBP1 is quite justified. In contrast, the low value of wMBP1-other genes has low influence upon deletion, but overexpression of MBP1 has pronounced and divergent effect on the genes lower in cascade. Therefore this connection is, in comparison with CLN3-MBP1 interaction, meaningful. Another similar example is the influence of deletion of SWI6 on the expression of CLN2 and CLB6. Both genes are controlled through SWI4, and SWI4 is the dominant regulator of the genes. For SWI4, the most important control effect is its self-induction; therefore, deletion of SWI6 has almost no effect on its expression, resulting in the loss of the deletion effect of SWI6 on the expression levels of CLN2 and CLB6. For CLN1,Cefoperazone which is also controlled by SWI4 and SWI6, the effect of SWI4 is low, and deletion of SWI6 causes the complete repression of CLN1 by strong FKH1. The same effect is observed for GIN4, where the repressor is MCM1 instead of FKH1. A conclusion from these observations is that the effect of regulator concentration and weight determine the expression level of its target genes at the bottom of the regulatory cascade in a highly unpredictable manner, which does not follow simple logic. In certain cases, the deletion of a gene on top of the cascade can have a striking effect, whereas in other cases, this effect completely vanishes in the regulatory cascade, as in the case of genes controlled indirectly by CLN3. An important consequence of the quantitative network behavior is that existing causal relationships between regulators and regulated genes in a cascade may not be discovered by gene deletion. Even direct connections, as in the case of SWI6-CLN2, may not be discovered if the influence of the regulator is not sufficiently pronounced. This statement has a profound consequence on the interpretation of mutagenesis experiments.