Hybridization isotherms of the mouse data set: The isotherms in panel a were calculated by plotting the probe intensities as a function of their expression value which is directly related to the concentration of specific transcripts, LS∝ [ S ]. The data were subsequently smoothed over a moving window of 1000 probe intensities. The isotherms were calculated using either all PM- or MM-probe data of the chip or, alternatively, the sub-ensemble of probes containing the (GGG)1-motif, i.e. a run of three consecutive guanines starting at the first sequence position. The horizontal bar in the upper part of the figure assigns the hybridization ranges (N, mix and S) which are described in the text. The arrows indicate the regions which are dominated either by specific or non-specific hybridization. Panel b shows theoretical isotherms which were calculated using Eq. (2) assuming three scenarios: (A) the reference situation describing the behavior of all probes; (B) stronger non-specific binding compared with A and; (C) stronger non-specific and specific binding compared with A (see also the text for details). Note that the intensity level in the N-range is directly related to KN, the mean binding constant of non-specific hybridization, whereas the position of the inflection point halfway between the N- and asymptotic saturation levels is inversely related to 1/KS, the mean binding constant of specific hybridization as indicated in the figure. The experimental data are compatible with stronger non-specific binding and invariant specific binding to (GGG)1-motifs compared with the respective main level of the binding strength of the array (scenario B). The MM-probes virtually behave like weak-affine PM probes with respect to specific binding.