As an important transcription factor of the Ral family, nuclear factor-kappa B (NF-kappaB) is involved in numerous cellular processes, such as the responses to cellular stress and to inflammation. For better elucidating the quantitative structure-activity relationship of NF-kappaB inhibitors and determining possible ligand-protein interaction, a pharmacophore model, Hypo1, was built based on 35 training molecules by Catalyst/HypoGen algorithm. The five pharmacophore features of Hypo1, including three hydrophobic groups, one hydrogen-bond acceptor, and one hydrophobic aromatic group, were correctly mapped onto NF-kappaB surface. This model has strong capability to identify NF-kappaB inhibitors and to predict the activities of structurally diverse molecules, thus to provide a valuable tool in the design of new leads with desired biological activity by virtual screening.