Figure 2

model validation. Sensitivities, Figure 2a, and specificities, Figure 2b, (with standard deviations as error bars) estimated by linear SVM classifier and 100 independent two-fold cross-validations using two disease specific and two control specific components. Components were extracted from the linear mixture models based on control reference (c.r.) sample, model (2a), and disease reference (d.r.) sample, model (2b), where each sample was comprised of ten orthogonal components containing K= 15000 features. One component contained in prevailing concentration disease specific features, one control specific features and eight components contained features equally expressed in control and disease labelled samples. Relative concentration (expressed through a mixing angle) across the sample population has been: for disease specific features in the range of 50⁰ to 89.99⁰; for differentially not expressed features in the range of 25⁰ to 65⁰; and for control specific features in the range of 0.01⁰ to 40⁰. Assumed overall number of components has been M= 2 (red bars), M = 3 (green bars), M= 4 (blue bars) and M = 5 (magenta bars).