Standard techniques for generating ligand conformations frequently fail to correctly predict the native structure of large macrocycles, and therefore structure-based drug design approaches cannot be reliably implemented. Without a high-quality description of a near-native backbone or ring configuration of macrocycles, correct prediction of the binding orientation is unlikely to occur. We address this challenge by developing new methods based on kinematic loop sampling. Emphasis was placed on generalizability and mathematical rigor. A total of 50 diverse, pharmaceutically-relevant macrocycles with structural data were sampled with these methods. Accuracy and efficiency were assessed through comparison to experimental data as well as best-in-class computational approaches. Sampling was initialized from random conformations with a heavy atom RMSD greater than 2.5 � to the experimental structure in order to avoid bias during modeling. Significant results include the accurate prediction of low-energy conformations within 0.09-1.0 � of the experimental structure. Using these low-energy conformations predicted by our sampling algorithm, we established a docking protocol against peptidylprolylisomerases using a set of known binders and decoys. Our results illustrate the necessity of allowing both protein and macrocycle flexibility in order to successfully identify the experimental binding interactions. Importantly, separation in relative binding rank was seen for known binders and decoys. We applied our docking protocol to a series of piperazic-acid macrocycles predicted to bind peptidylprolylisomerases. Several were identified as high-affinity binders based on the docking results, and these hits were confirmed experimentally.

Exploring the bioactive states of macrocyclic natural products: Conformational analysis and target identification

248th National Meeting (2014)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.