The utilization of the membrane transport protein PEPT1 as a drug delivery system is a promising strategy to enhance the oral bioavailability of drugs. Since very little is known about the substrate binding site of PEPT1, computational methods are a meaningful tool to gain a more detailed insight into the structural requirements for substrates. Three-dimensional quantitative structure-activity relationship (3D-QSAR) studies using the comparative molecular similarity indices analysis (CoMSIA) method were performed on a training set of 98 compounds. Affinity constants of beta-lactam antibiotics and tripeptides were determined at Caco-2 cells. A statistically reliable model of high predictive power was obtained (q(2) = 0.828, r(2) = 0.937). The results derived from CoMSIA were graphically interpreted using different field contribution maps. We identified those regions which are crucial for the interaction between peptidomimetics and PEPT1. The new 3D-QSAR model was used to design a new druglike compound mimicking a dipeptide. The predicted K(i) value was confirmed experimentally.