HIV-1 aspartic protease (PR) is a promising target for acquired immunodeficiency syndrome (AIDS) therapy, and the development of PR inhibitors can be accelerated by computer-aided design methods. We describe an approach for the design of new inhibitors, based on the modification of a known reference inhibitor, and the calculation of relative binding energies, taking into account contributions from all species in the binding equilibrium (inhibitor, PR and inhibitor/PR complex), as well as their solvation. This allows for a rational selection of new structures that are likely to have increased inhibition potency. We have analyzed reduced amide bond hexapeptides (Ac-P3-P2-P1-phi[CH2-NH]-P1'-P2-P3'-NH2), based on the structure of the known inhibitor MVT-101. A maximum gain in binding energy (approximately -55 kcal/mol) is observed when Phe or Tyr are present in positions P1 and P1', Glu in position P2' and aromatic residues (Phe, Trp or Tyr) in positions P3 and P3', while, in general, the presence of positively charged residues is destabilizing. This specificity is explained in terms of the interaction of individual inhibitor residues with proximal and distal PR residues. The validity of this computational approach has been confirmed by solid-phase synthesis of several of the designed pseudopeptides, followed by in vitro enzyme inhibition assaying. The best candidate structures show IC50 values in the low nanomolar range.