Improvements in correlating biological activity with its most important determinant, molecular shape, should improve the efficiency of drug discovery. In Phase I, a particularly promising method of shape correlation (3D-QSAR), known as DYLOMMS, has been implemented and enhanced. In Phase II, this method would be compared with the two other leading approaches to 3D-QSAR, a representative """"""""shape-difference"""""""" method, and the distance geometry approach, by comparing their correlative and predictive power when applied to five series of compounds. At the same time, all three methods would be embodied in a software product for shape correlation, tentatively named MOLYL. Study of the published results of Phase II would help a user of this software product to select the method best suited to his own SAR problem. A second objective of Phase II is to develop approaches for mapping molecular shape representations back into molecular structures, either by generating structures de novo, or by searching a molecular data base for compounds fit by a molecular shape description.
