Background: Diabetes is a risk factor associated with pancreatic ductal adenocarcinoma (PDAC), and new adult-onset diabetes can be an early sign of pancreatic malignancy. Development of blood-based biomarkers to identify diabetic patients who warrant imaging tests for cancer detection may represent a realistic approach to facilitate earlier diagnosis of PDAC in a risk population. Methods: A spectral library-based proteomic platform was applied to interrogate biomarker candidates in plasma samples from clinically well-defined diabetic cohorts with and without PDAC. Random forest algorithm was used for prediction model building and receiver operating characteristic (ROC) curve analysis was applied to evaluate the prediction probability of potential biomarker panels. Results: Several biomarker panels were cross-validated in the context of detection of PDAC within a diabetic background. In combination with carbohydrate antigen 19-9 (CA19-9), the panel, which consisted of apolipoprotein A-IV (APOA4), monocyte differentiation antigen CD14 (CD14), tetranectin (CLEC3B), gelsolin (GSN), histidine-rich glycoprotein (HRG), inter-alpha-trypsin inhibitor heavy chain H3 (ITIH3), plasma kallikrein (KLKB1), leucine-rich alpha-2-glycoprotein (LRG1), pigment epithelium-derived factor (SERPINF1), plasma protease C1 inhibitor (SERPING1), and metalloproteinase inhibitor 1 (TIMP1), demonstrated an area under curve (AUC) of 0.85 and a two-fold increase in detection accuracy compared to CA19-9 alone. The study further evaluated the correlations of protein candidates and their influences on the performance of biomarker panels. Conclusions: Proteomics-based multiplex biomarker panels improved the detection accuracy for diagnosis of early stage PDAC in diabetic patients.

Although the gut microbiome has been associated with dietary patterns linked to health, microbial metabolism is not well characterized. This ancillary study was a proof of principle analysis for a novel application of metaproteomics to study microbial protein expression in a controlled dietary intervention. We measured the response of the microbiome to diet in a randomized crossover dietary intervention of a whole-grain, low glycemic load diet (WG) and a refined-grain, high glycemic load diet (RG). Total proteins in stools from 9 participants at the end of each diet period (n = 18) were analyzed by LC MS/MS and proteins were identified using the Human Microbiome Project (HMP) human gut microbiome database and UniProt human protein databases. T-tests, controlling for false discovery rate (FDR) <10%, were used to compare the Gene Ontology (GO) biological processes and bacterial enzymes between the two interventions. Using shotgun proteomics, more than 53,000 unique peptides were identified including microbial (89%) and human peptides (11%). Forty-eight bacterial enzymes were statistically different between the diets, including those implicated in SCFA production and degradation of fatty acids. Enzymes associated with degradation of human mucin were significantly enriched in the RG diet. These results illustrate that the metaproteomic approach is a valuable tool to study the microbial metabolism of diets that may influence host health.