Fig. 1.

Logistic regressions of IHC-obtained receptor status versus gene expression. For each receptor we obtain one curve from the very receptor gene (solid curve) and a second one from the co-gene (dashed), both shown in the same colour (see legend). Left panel: probabilities (y-axis) of positive receptor status, given a GCRMA-normalized expression value (x-axis). For values of regression parameters and quality of regression see Table 8. Right panel: corresponding receiver operator characteristics (ROC)-curves. For quantitative estimates of regression quality, see Table 8