Back to 2008 Program and Abstracts
Applying Proteomics Based Biomarker Tools for the Accurate Diagnosis of Pancreatic Cancer
John D. Christein*1, Kyoko Kojima2, Senait G. Asmellash3, Christopher a. Klug2, James a. Mobley1
1Surgery, University of Alabama at Birmingham, Birmingham, AL; 2Microbiology, University of Alabama at Birmingham, Birmingham, AL; 3Biochemistry and Molecular Genetics, University of Alabama at Birmingham, Birmingham, AL
Background: Clinical proteomics is the study of proteins and the physiologic state of an individual. The human proteome or biologic signature contains biomarkers associated with varying diseases. There is no biomarker set specific to pancreatic cancer and only a few proteomic studies have attempted to differentiate patients with pancreatic cancer from those without disease. Aims: To evaluate the biologic fluid proteome of patients with and without pancreatic disease. For the first time, to accurately differentiate patients with pancreatic cancer based on proteome analysis by mass spectrometry.
Methods: From 2003-2007, 245 patients underwent prospective serum, plasma, and urine collection. Endoscopic ultrasound and/or surgical pathology were used to distinguish normal (N) patients from cancer (CA) or chronic pancreatitis (CP). A reproducible high throughput (HTP) method using a high affinity solid core lipophilic extraction resin enriched the low molecular weight protein fraction of a sample. Proteome analysis coupled the extraction with a high speed 200Hz matrix-assisted laser desorption/ionization-time of flight (MALDI-ToF/ToF) mass spectrometer (Bruker Ultraflex III). Poor quality samples and outliers were excluded from the sample set. Samples from N, CA, and CP groups underwent software processing with FlexAnalysis, Clinprot, MatLab, and Statistica to align and normalize spectra. Statistical non-parametric pairwise analysis, multidimensional scaling (MDS), hiearcical analysis, and leave one out cross validation (LOOCV) using a k-means based approach completed the comparison.
Results: Fifty usable serum samples (15 N 24 CA, 11 CP) underwent proteomic based analysis. Sensitivity (sn) and specificity (sp) of group comparisons were determined. Using 6 serum features, we accurately differentiated CA from N (sn 88.9%, sp 93.8%), CA from CP (sn 88.9%, sp 61.1%), and N from both CA and CP when combined (sn 88.9%, sp 76.4%). When combined with 14 features from urine spectra, CA was differentiated from N and CP with a sensitivity approaching 94%. Interestingly, the plasma samples (considered by the Human Proteome Organization to be the preferred biological fluid) did not show significant differences between patient groups.
Conclusion: Human biologic fluids can successfully be used to differentiate patients with pancreatic cancer from those without disease or chronic pancreatitis. Proteomic analysis of human serum and urine can provide a high level of predictability for diagnosing pancreatic cancer. The proteomic analysis of biologic fluids may be able to be used to screen individuals for pancreatic cancer or differentiate benign from malignant disease.
