Beta-secretase (BACE) inhibitors have potential as anti-Alzheimer's disease treatments prompting us to explore the pharmacophoric space of 129 known BACE inhibitors. QSAR analysis was employed to select optimal combination of pharmacophoric models and 2D physicochemical descriptors capable of explaining bioactivity variation (r(2)=0.88, F=60.48, r(LOO)(2)=0.85, r(PRESS)(2) against 25 external test inhibitors=0.71). We were obliged to use ligand efficiency as the response variable because the logarithmic transformation of bioactivities failed to access self-consistent QSAR models. Three pharmacophoric models emerged in the successful QSAR equation suggesting at least three binding modes accessible to ligands within BACE binding pocket. QSAR equation and pharmacophoric models were validated through ROC curves and were employed to guide synthesis of novel pyridinium-based BACE inhibitors. The best inhibitor illustrated an IC(50) value of 1.0 microM against BACE.