Fig. 3

Assessment of information content of IMF concerning disease progression. a Differential infrared spectra, showing the mean difference per wavenumber between measured data of lung cancer patients of different stages and age- and gender-matched healthy control individuals. The shaded area corresponds to the standard deviation per wavenumber in the control group. b Empirical ROC curves (and the corresponding AUC values) for the binary classification between lung cancer patients (of different stages) and matched healthy control individuals. The classification was performed using logistic regression within 10-fold cross-validation. c Modeled covariate effect of the predicted lung tumor stage on the ROC curve using a generalized linear model (ROC-GLM). See the Methods section for more details. d Learning curves for mean AUC, for the binary classification between lung cancer patients and matched healthy control individuals (experimental data depicted as blue dots; Stage I, II, III - orange fit line; Stage IV - dark red fit line). e Mean effect size indicating the overall average differences between infrared fingerprints of cases (of different stages) and controls