Given a specific receptor, the problem is to find or design a binding small ligand molecule. If the receptor site geometry is known, the problem is to find a molecule that satisfies some geometric constraints and is also a good chemical match. After finding good candidates according to those criteria, a docking step with energy minimization can be used to predict binding strength.
If the site geometry is not known, as is often the case, the designer must base the design on other ligand molecules that bind well to the site. If those other molecules are rigid, the problem becomes one of identifying the substructures or active groups that contribute to the fit. By joining the groups with alternative molecular scaffolding, one can build molecules with good or better affinity for the site. The joining process requires solutions of kinematics and distance geometry problems. If the receptor site geometry is unknown and the ligands are flexible, then the designer must first posit the configurations of the ligands in their bound state by assuming that their active groups are in similar places. New drugs are enumerated which can be folded into a similar appearance.