Balancing the tumor and stromal HNSCC signature components results in a more robust and accurate predictive profile. (a) Tumor cell specific and tumor stroma specific HNSCC signature genes can be dissected into four compartments: stroma N+, tumor N+, stroma N0 and tumor N0. Light grey indicates N0 association, and dark grey indicates N+ association. (b) Model for the relative contribution of the four components shown in (a) to the initial HNSCC signature. Combining the four components into one predictive outcome (indicated by arrows) results in tumor percentage signature bias. Low tumor percentage samples (left-hand side) show a more N+ biased profile (dark grey), whereas samples with a very high tumor percentage (right-hand side) exhibit a bias towards an N0 profile (light grey). (c) As (b), but for a corrected signature composition that does not exhibit a strong bias in the predictive outcome of low and high tumor percentage samples. (d) Selection of a set of 119 HNSCC signature genes that are equally distributed across the four different components, plotted similarly as in (a). (e) Predictive outcomes based on the corrected signature that consists of the 119 genes shown in (d). The corrected signature shows a strong reduction in predictive bias for samples with a low or very high tumor percentage; colors are as in Figure 3b. (f) Odds ratios for the signature outcome for prediction of metastasis based on the original signature, the balanced signature and through weighted correction based on the tumor cell percentage of samples.
Roepman et al. Genome Biology 2006 7:R117 doi:10.1186/gb-2006-7-12-r117