Identification of a Collagen Marker in Urine Improves the Detection of Colorectal Liver Metastases

Previously, we reported a combination of an urine collagen alpha-1­(I) natural occurring peptide (NOP) AGPP­(−OH)­GEAGKP­(−OH)­GEQGVP­(−OH)­GDLGAP­(−OH)­GP (AGP) and serum carcinoembryonic antigen (CEA) to have the potential to detect colorectal liver metastasis (CRLM). The combined method requires further adaption for better sensitivity and specificity prior to clinical implementation. This mass spectrometry study aimed to identify additional collagen NOPs in urine and determine the most discriminating NOP panel. We improved the combined method on the basis of analysis of urine samples from 100 healthy controls and 100 CRLM patients. Two additional NOPs were identified: GPPGEAGK­(−OH)­P­(−OH)­GEQGVP­(−OH)­GDLGAP­(−OH)­GP (GPP), collagen alpha-1­(I), and GNDGARGSDGQPGPP­(−OH)­GP­(−OH)­P­(−OH)­GTAGFP­(−OH)­GSP­(−OH)­GAK­(−OH)­GEVGP (GND), collagen alpha-1­(III). A molecular model combining NOPs (AGP, GPP, and GND) and CEA was generated. Molecules that did not contribute significantly were removed, resulting in a model consisting of GND and CEA. With this model, 88% sensitivity and 88% specificity were reached in the discovery set and 75% sensitivity and 100% specificity in the validation set (control, n = 12; CRLM, n = 10). The AUC of the ROC curve is significantly higher than the current model based on AGP and CEA (p = 3.3 × 10–4). The new model performs better than the currently used techniques in the clinic that have a 57–70% sensitivity and a 90–96% specificity.