Pancreatic Ductal Adenocarcinoma (PDAC) is a disease with a great heterogeneity in the response to treatments. To improve the responsiveness to treatments there are two different approaches, the first one consist to develop new and more efficient drugs that intent to cure all patients and the second one is to use already-approved drugs, alone or in combination, but selecting beforehand the most sensitive patients. In this work we explored the efficiency of the second possibility. We developed a collection of 17 PDAC samples collected by Endoscopic Ultrasound-Guided Fine-Needle Aspiration (EUS-FNA) or surgery and preserved as xenografts and as primary cultures. This collection was characterized at molecular level by a transcriptomic analysis using an Affymetrix approach. In this paper we present data demonstrating that a subgroup of PDAC responds to low doses of 5-aza-dC. These tumors show a specific RNA expression profile that could serve as a marker, but there is no correlation with Dnmt1, Dnmt3A or Dnmt3B expression. Responder tumors corresponded to well-differentiated samples and longer survival patients. In conclusion, we present data obtained with the well-known drug 5-aza-dC as a proof of concept that a drug that seems to be inefficient in solid tumors in general could be applicable to a particular subgroup of patients with PDAC.