Efficient conformational analysis of small molecules

In computational methodologies that aid the drug discovery process, increasingly large datasets often make the conformational analysis of the molecules the limiting step in the processing of the data. In the case of more flexible molecules, current conformational analysis techniques provide results of limited utility. The contributions of this dissertation address both of these problems. All the techniques described assume that the primary flexibility in molecules results from rotation about the bonds, i.e. bond lengths and bond angles are fixed.; The first part of this work encompasses methods that allow the robust analysis of large numbers of molecules. The focus is on the combination of numerous techniques that together form a practical conformational analysis program. Its implementation is part of a drug discovery software suite that is in use in many sites around the world.; Because of their additional geometric constraints, cyclic structures are a particular challenge to conformational analysis programs. This work describes a geometric algorithm that allows the efficient analysis of macrocyclic systems. The implementation is comparable to current methods and lays the foundation for a more powerful approach to these systems.; Flexible molecules have an extremely large number of low-energy conformations, and current approaches use some heuristic to select a possibly very small fraction of these conformations. In general, the more flexible the molecule is, the more low-energy conformations are excluded by these approaches in the final conformational model. This work explores the novel application of multi-valued decision diagrams (MDDs) to the representation of sets of conformations. Because of the implicit storage of the conformations, the data structure allows the storage of the complete conformational space of even flexible molecules. The approach assumes a discretization of the angle of rotation about the bonds. Combined with a fragment-based algorithm for conformational analysis, MDDs enable the efficient generation of all low-energy conformations of these molecules.