Dengue fever is caused by four distinct serotypes of the dengue virus (DENV1-4), and is estimated to affect over 500 million people every year. Presently, there are no vaccines or antiviral treatments for this disease. Among the possible targets to fight dengue fever is the viral NS3 protease (NS3PRO), which is in part responsible for viral processing and replication. It is now widely recognized that virtual screening campaigns should consider the flexibility of target protein by using multiple active conformational states. The flexibility of the DENV NS3PRO could explain the relatively low success of previous virtual screening studies. In this first work, we explore the DENV NS3PRO conformational states obtained from molecular dynamics (MD) simulations to take into account protease flexibility during the virtual screening/docking process. To do so, we built a full NS3PRO model by multiple template homology modeling. The model comprised the NS2B cofactor (essential to the NS3PRO activation), a glycine flexible link and the proteolytic domain. MD simulations had the purpose to sample, as closely as possible, the ligand binding site conformational landscape prior to inhibitor binding. The obtained conformational MD sample was clustered into four families that, together with principal component analysis of the trajectory, demonstrated protein flexibility. These results allowed the description of multiple binding modes for the Bz-Nle-Lys-Arg-Arg-H inhibitor, as verified by binding plots and pair interaction analysis. This study allowed us to tackle protein flexibility in our virtual screening campaign against the dengue virus NS3 protease.