Abstract B5: Identification of serum biomarker signatures associated with pancreatic cancer.

Abstract

Motivated by the unmet clinical need for accurate markers for pancreatic cancer, this study was conducted with the aim to identify and pre-validate disease specific serum protein signatures, using an in-house, state-of-the-art, recombinant antibody microarray platform. Directly biotinylated sera from 148 individuals with pancreatic cancer (PC), chronic pancreatitis (CP), autoimmune pancreatits (AIP), and normal, healthy controls (N), were profiled on arrays consisting of 121 antibodies mainly targeting immunoregulatory proteins. Bound analytes were visualized using confocal fluorescence scanning. Using a support vector machine classifier with leave-one-out cross validation, the screening revealed serum protein portraits distinguishing PC from N (Area Under ROC-Curve (AUC) 0.95), CP (AUC 0.86), and AIP (AUC 0.99), respectively. In order to further condense the candidate biomarker profiles, we applied an iterative backward elimination strategy, with which two distinct protein signatures were derived in training cohorts. The first, consisting of 18 proteins, predicted PC from N with AUC 0.95 in an independent sample cohort. The second, consisting of 25 proteins, discriminated PC from the combined group of N, CP and AIP with AUC 0.88, thus representing a high diagnostic potential. Notably, the classification performance was not improved by adding CA19-9, an a priori known, commonly applied serum biomarker, to the signatures. In conclusion, we present the first pre-validated, multiplexed serum biomarkers signatures of high predictive power for PC, which might prove useful for improved diagnosis and non-invasive screening of high-risk individuals. This data demonstrated the unique opportunities for biomarker discovery and disease diagnosis provided by affinity proteomics approaches, here represented by recombinant antibody microarrays. Citation Format: Anna Sandstrom, Christer Wingren, Ralf Segersvard, Anders Carlsson, Roland Andersson, Mattias Lohr, Carl A.K. Borrebaeck. Identification of serum biomarker signatures associated with pancreatic cancer. [abstract]. In: Proceedings of the AACR Special Conference on Pancreatic Cancer: Progress and Challenges; Jun 18-21, 2012; Lake Tahoe, NV. Philadelphia (PA): AACR; Cancer Res 2012;72(12 Suppl):Abstract nr B5.