Conformational search using molecular mechanics methods is a key part of both reaction exploration and calculation of NMR spectra. For most organic molecules, many more conformers are found than can be considered at higher levels of theory. To deal with this, energy cut-offs, RMSD pruning and manual conformer selection are all used, but each have their limitations. Combination of conformer internal coordinates with modern clustering algorithms allowed us to develop an unsupervised method for intelligent grouping of conformers into chemically meaningful clusters. This method allows reliable coverage of the conformational space in NMR calculations and faster knowledge discovery in reaction prediction.<br/>We have tested this approach on both single molecules and multi-molecular complexes. The methods handle even large, flexible molecules well, and is applicable to reactive complexes in asymmetric organocatalysis. This is very useful in computational organic chemistry, where it is often challenging even for experts to quickly identify conformer families, especially in large 150-300 atom systems. The HDBSCAN clustering algorithm does this reliably and quickly and makes the calculations and the resulting predictions more reliable and more responsive. This has helped in the computational study of a TRIP-catalyzed oxy-Michael reaction by identifying key conformational families in otherwise very complex conformational landscape. HDBSCAN has now been integrated in PyReact reaction computation workflow and is crucial to the full automation of ground-state calculations.<br/>Adequate conformer selection in the calculation of NMR spectra is just as important. While in reaction calculation it is only the lowest conformer that matters, for accurate NMR prediction, the whole ensemble of conformations must be reproduced. With the new clustering and conformer selection procedure we have shown that previous approaches often missed important conformations, and we are able to achieve greater accuracy and greater reliability with only modest increase in computational cost.<br/>Methods for better conformational space visualization have also been developed. This allows chemists to visually check the conformational space coverage and to focus on understanding the key representative conformations. All of these advancements result in more intelligent, more responsive and more reliable spectra calculation and reaction exploration.

Chemically relevant conformation clustering: Applications to reaction exploration and calculation of spectra

256th National Meeting (2018)

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