Figure 15

Systematic bias of 'Langmuir'-expression estimates (Eq. (12)) with respect to the estimates which consider washing (Eq. (11)): Part a and b: Correlation plot between both estimates and their logged difference, . The graphs were calculated assuming K^P,h = const for two non-specific background levels (red and black curves) and for PM and MM probes (dotted and solid curves) assuming the survival fractions w^PM,S(t) = 0.95, w^MM,S(t) = 0.50 and w^P,N(t) = 0.1. Neglecting washing underestimates the expression degree especially at small and large expression values. Part c: The survival fraction of bound probes depends on the intensity (or, equivalently, probe occupancy) before (t = 0) and after washing (t > 0). The graph for t = 0 was re-plotted from Figure 5a using Eq. (17) with w_max = 0.9, w_min = 0.06, γ = 1.6 and a' = 0.1. The graph for t = 6 refers to the standard number of washing cycles. It is obtained from the t = 0 graph by making the substitution logI(t) = logI(0)+log(w(t)) in the argument (see Eq. (8)). Part d shows the bias of the Langmuir approximation assuming a constant transcript concentration and variable K^P,S and thus a variable survival fraction w(Θ) which has been taken from part c of the figure for t = 6. The dashed curves labelled with '(+)' and '(-)' in panels c and d refer to 50%-deviations of the washing function, log w(Θ)^+/- = log w(Θ)·(1.5)^(+/-)1, to estimate the effect of the scattering of the probe level data from the mean (compare with Figure 5a). The bias of the Langmuir-approximation strongly resembles that shown in part b. Note that the bias applies to PM and MM probes as well in this case.