A Monte Carlo-Based Torsion Construction Algorithm for Ligand Design

Citation

Kekenes-Huskey, Peter Michael (2009) A Monte Carlo-Based Torsion Construction Algorithm for Ligand Design. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/A1MQ-3116. https://resolver.caltech.edu/CaltechETD:etd-05282009-131419

Abstract

A wealth of computational strategies [1,2,3,4,] is available for predicting the binding site and affinities of a putative ligand inside a target receptor. Although numerous techniques focus on the orientation of ligands or fragments thereof, few methods have delved into improving the accuracy of generating reliable ligand conformations within predicted binding modes. In an effort to comprehensively sample the torsion space available to a flexible ligand and focus on low-energy conformations, a recursive, Metropolis Monte Carlo (MC)-based rotamer design protocol has been developed. This approach recursively samples adjacent rotatable bonds from a defined anchor and directs the search along low-energy pathways, such that high-affinity conformations of the ligand can be identified. Furthermore, this program applies spatial constraints within the search that restrict the solutions to structurally dissimilar conformations, thus encouraging a diverse solution set. The performance of moleculeGL has been evaluated for a set of 55 co-crystals, for which the number of rotatable bonds ranged from 2 to 32. Approximately 80 percent of the structures are predicted within 2.0 A2 root mean square deviations (RMSD) with respect to the crystal structure, starting from an arbitrary ligand conformation. This level of accuracy suggests the program's applicability to the design of pharmacaphore substituents, for which the position of a chemically active pharmacaphore is well-known.

Thesis Files