Peaks selected for models when quantile normalization was used before averaging and feature selection.

List of the classifier-selected peaks for Asterand-based classifier (A), RNTech-based classifier (R) and Mixed set-based classifier (M). Models studied data obtained when quantile normalization was used before averaging and feature selection. The influence that each peak has on the prediction of the linear kernel SVM-based model was evaluated by entering an zero values vector and comparing this result to that of a vector containing zeros in all but one peak which has the maximum possible value. Directions were corrected so that signs would mean the same in all three models (minus: cancer; plus: control). Possible identification: Similar m/z values from the study by Villanueva et al.[11] and EPO-KB (searchable knowledge base of biomarker to protein links) [34] are presented.