A theoretically rigorous and computationally tractable methodology for the prediction of the free energies of binding of protein-ligand complexes is presented. The method formulated involves developing molecular dynamics trajectories of the enzyme, the inhibitor, and the complex, followed by a free energy component analysis that conveys information on the physicochemical forces driving the protein-ligand complex formation and enables an elucidation of drug design principles for a given receptor from a thermodynamic perspective. The complexes of HIV-1 protease with two peptidomimetic inhibitors were taken as illustrative cases. Four-nanosecond-level all-atom molecular dynamics simulations using explicit solvent without any restraints were carried out on the protease-inhibitor complexes and the free proteases, and the trajectories were analyzed via a thermodynamic cycle to calculate the binding free energies. The computed free energies were seen to be in good accord with the reported data. It was noted that the net van der Waals and hydrophobic contributions were favorable to binding while the net electrostatics, entropies, and adaptation expense were unfavorable in these protease-inhibitor complexes. The hydrogen bond between the CH2OH group of the inhibitor at the scissile position and the catalytic aspartate was found to be favorable to binding. Various implicit solvent models were also considered and their shortcomings discussed. In addition, some plausible modifications to the inhibitor residues were attempted, which led to better binding affinities. The generality of the method and the transferability of the protocol with essentially no changes to any other protein-ligand system are emphasized.