The aim of the study was to analyze the gene expression profile of pancreatic cancer to derive novel molecular markers of this malignancy. The snap-frozen or RNA-later preserved samples of 18 pancreatic adenocarcinomas, 5 chronic pancreatitis cases and 6 specimens of grossly normal pancreas were used for microarray analysis by HG-U133 Plus 2.0 oligonucleotide Affymetrix arrays. Validation was carried out by real-time quantitative PCR (Q-PCR) in the set of 66 samples: 31 of pancreatic cancer, 14 of chronic pancreatitis and 21 of macroscopically unchanged pancreas. By Principal Component Analysis of the microarray data we found a very consistent expression pattern of normal samples and a less homogenous one in chronic pancreatitis. By supervised comparison (corrected p-value 0.001) we observed 11094 probesets differentiating between cancer and normal samples, while only seventy six probesets were significant for difference between cancer and chronic pancreatitis. The only gene occurring within the best 10 genes in both comparisons was S100 calcium binding protein P (S100P), already indicated for its utility as pancreatic cancer marker by earlier microarray-based studies. For validation we selected two genes which appeared as valuable candidates for molecular markers of pancreatic cancer: neuroblastoma, suppression of tumorigenicity 1 (NBL1) and anillin (ANLN). By Q-PCR, we confirmed statistically significant differences in these genes with a 9.5 fold-change difference between NBL1 expression in cancer/normal comparison and a relatively modest difference between cancer and pancreatitis. For ANLN even more distinct differences were observed (cancer/normal 19.8-fold, cancer/pancreatitis 4.0-fold). NBL1 and anillin are promising markers for pancreatic carcinoma molecular diagnostics.