Detection of serum markers for pancreatic cancer has been elusive. Although CA 19-9 is most commonly used, its sensitivity and specificity are modest. We used large-scale proteomics to identify potential serum markers for pancreatic cancer. Samples were analyzed using high-resolution two-dimensional gel electrophoresis to identify differentially expressed proteins in 32 normal and 30 pancreatic cancer patients. Up to 1,744 protein spots were resolved for each serum sample. Candidate proteins were identified using mass spectrometry. ANOVA was used to identify proteins that could discriminate cancer from normal sera. Serum fibrinogen level was also measured using enzymatic assay. Immunohistochemistry was used to detect fibrinogen in resected pancreatic cancers. One hundred fifty-four proteins were commonly overexpressed in all pancreatic cancers. Nine protein spots (four with identifications by mass spectrometry) could effectively separate cancer from normal controls using cross-validation. These proteins successfully discriminated all pancreatic cancer samples (30 of 30) and 94% of normal (30 of 32) samples. Prominent among these candidates was fibrinogen gamma, which was subsequently confirmed to be overexpressed in pancreatic cancer sera by enzymatic analysis (54.1 +/- 64.1 versus 0.0 +/- 0.0 mg/dL, P < 0.05) and tissue by immunohistochemistry (67% versus 29%, P < 0.05) relative to normal pancreas. Proteomic analysis combining two-dimensional gel electrophoresis and mass spectrometry successfully identified 154 potential serum markers for pancreatic cancer. Of these, fibrinogen gamma, a protein associated with the hypercoagulable state of pancreatic cancer, discriminated cancer from normal sera. Fibrinogen is a potential tumor marker in pancreatic cancer.