Fig. 2.

Incorporation of Protein Abundance in Peptide Detection Classifier Improves Performance. A, Scheme of the neural network model used for classification of peptide detection resulting in 16 features including 15 physicochemical properties and protein abundance, resulting in the Peptide Prediction with Abundance (PPA) score. The model was trained with 10-fold cross-validation. B, Distribution of PPA scores for detected versus not detected peptides of the 120 singly analyzed protein digests. C, The Receiver Operating Characteristic (ROC) was calculated for three models and compared with the previously published ESP predictor (15): red: 15 neural network model physicochemical features only; blue: PPA including protein abundance (PA) as FLEXII-peptide intensity; orange: PPA including protein abundance (PA) as sequence coverage; green: ESP predictor on the 10-fold cross-validated data set.