Serum peptide signatures for pancreatic cancer based on mass spectrometry: a comparison to CA19-9 levels and routine imaging techniques

Abstract

Purpose: The detection of pancreatic tumors lacks a sensitive and specific diagnostic tool. Mass spectrometry (MS)-based profiling of serum proteins is a promising approach for discovery of new clinical biomarkers or biomarker signatures.

Methods: Serum samples from pancreatic cancer (PC) patients and control individuals were collected and processed using a standardized protocol. Samples were divided in a calibration set (n = 49 PC and 110 controls) and a validation set (n = 39 PC and 75 controls). Peptide profiles were obtained using a combination of automated solid-phase extraction with reversed-phase C18 paramagnetic beads and matrix-assisted laser desorption ionization time-of-flight MS.

Results: Linear discriminant analysis with double cross-validation resulted in a discriminating peptide signature for PC in the calibration set with a sensitivity of 78 % and a specificity of 91 % [area under the curve (AUC) of 92 %]. Classification was validated with a sensitivity of 93 % and a specificity of 100 % (AUC of 98 %), and the results were compared with carbohydrate antigen 19-9 levels and currently available clinical imaging techniques. The ten most discriminating peptide peaks were identified as fragments of proteins involved in the clotting cascade, acute phase response and immunologic response.

Conclusions: In this study, it is shown that MS-based serum peptide profiles can discriminate between PC and control samples. The approach has great potential for high-throughput analysis in surveillance programs and appears to be most promising for patients with an inherited risk for PC, who benefit from more frequent screening.

Fig. 1

Example of an RPC18 profile including multiple overlays of the most discriminating peaks. On the x-axis, the m/z-values are depicted, and on the y-axis, the intensities (in arbitrary units). Peaks with m/z-values of 1,206.7, 1,518.9, 1,536.9, 1,865.2, 2,271.1, 2,602.5, 3,190.6 and 3,261.7 were statistically evaluated as discriminating cases from controls. Profiles of cases are plotted in blue, profiles of controls in green

Fig. 2

ROC curves of the calibration set and validation set after peak selection with standardization. The sensitivities are plotted in function of the 1-specificities for different cut-off points of the classification threshold. The area under the ROC curve is a measure of the between-group separation (diseased/normal)

Fig. 3

Comparison between results from RPC18-based peptide signatures and CA19-9 levels

Fig. 4

Stem-and-leave plot of the validation data based on MEANPOSEXPR–MEANNEGEXPR. The blue circles correspond to the cases, the red crosses with the controls. Items with the same expression were plotted vertically. The controls are more located on the left of the horizontal axis; the cases are more located on the right