Figure 2.

Network condensation - a gallery of various scenarios. For each gene, its (expression) values are represented by color. For each link, its score is represented by color and thickness. The coloring scheme is red, pale red, white, pale green and green, in order of decreasing values (see text). Links connecting genes with measurement values changing in an inconsistent way are marked by wavy lines. As in Fig. 1, if the interacting genes are linked by a stimulation S → T , the stimulation is assumed to start up, if for both genes, the values go up from E₁ to E₂ ((a), E₁ value: left side of circle, E₂: right side of circle), and it is assumed to be shut down if both values go down (b). An inhibition I ⊣ T is assumed to start up, if the inhibitor value goes up, but the target value goes down (c); it is assumed to shut down if the inhibitor value goes down and the target value goes up (d). In cases (e) and (f) the startup of the stimulation as presented is still a justified hypothesis, even though the target does not go up. For example, in (e) and (f), the stimulation by the source (the stimulator) goes up but it may be counteracted by other inhibiting effects (dashed T-Bar arrow) on the target, as the target does not change (e) or even goes down slightly (f) (source principle, see text). In cases where the amount of the stimulator is constant (g) or goes down slightly (h), the startup of a stimulation is still a justified hypothesis based on the target value. Strictly speaking, we hypothesize the startup of the stimulatory effect on the target gene. For example, in (g), the startup is not concluded from the change in the value of the stimulator, but it may be due to stimulator accumulation in time, and/or due to cooperation of the stimulator with other stimulations of the target which go up at the same time; startup of the stimulating effect is concluded from the behaviour of the target (target principle, see text). Scenario (i) is another example of the source principle: it is a justified hypothesis that the inhibition starts up because the amount of inhibitor increases, even though counteracting stimulations drive the amount of the target. Scenario (k) is another example of the target principle: it is a justified hypothesis that the stimulatory effect goes up, observing the target and assuming other cooperating effects on it. Lastly, if values do not change at all, or if they change in a completely inconsistent way, the amount of change is zero or near to zero, as in (l) and (m) (also see Fig. 1 (c) and (d)). Note that cases (e)-(m) all result in reduced link scores. Hence, inconsistent links tend to be removed from the network, as the link score threshold is made more stringent.