4D-QSAR analysis incorporates pharmacophore, conformational and alignment freedom into the development of 3D-QSAR models for training sets of structure-activity data by performing ensemble averaging, the fourth ™di-mension∫. The data required to perform 4D-QSAR analysis includes a training set of compounds, usually analogs, and their measured biological activities in a common screen/assay. The 4D-QSAR approach can be applied to both receptor-dependent (RD) and receptorindependent (RI) problems. In the first scheme, the geometry of the receptor (molecular target, usually an enzyme) is available. In contrast, in the second scheme the geometry of the receptor is not part of the data available to perform the analysis. The descriptors in 4D-QSAR analysis are lattice grid cell (spatial) occupancy measures of atoms composing each molecule in the training set realized from the sampling of conformational and alignment spaces. These grid cell occupancy descriptors (GCODs) are generated for a number of different atom types, the interaction pharmacophoric elements (IPEs). Non-GCOD descriptors can also be included with the set of GCODs in building the trial descriptor pool for model development. The idea underlying 4D-QSAR analysis is that the differences in activity among a set of ligands are related to differences in their Boltzmann average spatial distribution of molecular shape with respect to the IPEs. The 3D-QSAR models are generated and evaluated by a scheme that combines a genetic algorithm (GA) optimization with partial least-squares (PLS) regression. A single ™active∫ conformation is postulated for each compound in the training set, which, when combined with the optimal alignment, can be used in additional molecular design applications, including other 3D-QSAR methods. The 4D-QSAR models can also be used as virtual screens in the processing of real and/or virtual ligand libraries. In this paper the 4D-QSAR paradigm is given in detail. Moreover, we report the application of the (RI) 4D-QSAR formalism to a set of novel nonpeptidic HIV protease inhibitors. The 4D-QSAR models generated are robust and provide insight regarding the probable mechanism of action of the analogs, as well as hints concerning new synthetic routes. Furthermore, these models can be used as a starting point for future receptor-dependent anti-HIV drug design.