Bulky, flexible molecules such as peptides and peptidomimetics are often used as lead compoundsduring the drug discovery process. Pathophysiological events, e.g., the formation of amyloid fibrils inAlzheimer's disease, the conformational changes of prion proteins, or
![](/images/gifchars/beta2.gif)
BORDER=0 ALIGN="middle">-secretase activity, may besuccessfully hindered by the use of rationally designed peptide sequences. A key step in the molecularengineering of such potent lead compounds is the prediction of the energetics of their binding to themacromolecular targets. Although sophisticated experimental and in silico methods are available to helpthis issue, the structure-based calculation of the binding free energies of large, flexible ligands to proteinsis problematic. In this study, a fast and accurate calculation strategy is presented, following modification ofthe scoring function of the popular docking program package AutoDock and the involvement of ligand-based two-dimensional descriptors. Quantitative structure-activity relationships with good predictive powerwere developed. Thorough cross-validation tests and verifications were performed on the basis ofexperimental binding data of biologically important systems. The capabilities and limitations of the ligand-based descriptors were analyzed. Application of these results in the early phase of lead design will contributeto precise predictions, correct selections, and consequently a higher success rate of rational drug discovery.