Accurate modeling of 3D structure for natural products is especially challenging, due to complexities involving numerous chiral centers, ring conformations with inherent strain, flexibility in macrocycles, and the novelty in structures that limit template-based algorithms. ForceGen is a template-free, non-stochastic approach to 2D to 3D structure generation and conformational elaboration. The introductory study for the method demonstrated macrocyclic ligand search on the Foloppe/Chen 30-compound macrocycle set showed comparable accuracy, but with 10--100x speed improvements over competing molecular-dynamics approaches. Here, we focus on a more extensive set of validation experiments, primarily on large, public, macrocycle data set benchmarks. We will present detailed results on three challenging data sets: the ForceGen set of 182 macrocycles, the MacroModel benchmark set of 150 macrocycles, and the Prime-MCS benchmark of 208 macrocycles. On non-macrocyclic molecules, standard ForceGen calculation modes yield conformer pools as quickly as the fastest methods while maintaining superiority in terms of performance across a wide range of RMSD cutoffs. On macrocyclic molecules, the standard ForceGen search mode yields conformer pools much more quickly than alternative methods and produces pose pools that are as accurate. In the thorough search mode, ForceGen meets or exceeds the performance of all competing methods on macrocyclic ligands.

Template-free 3D structure generation and conformer search: Complex natural products and macrocycles

256th National Meeting (2018)

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