Brief overview of recently reported misassigned natural products and their in silico revisions enabled by DU8ML, a machine learning-augmented DFT computational NMR method

Abstract

A personal selection of recent misassigned structures of natural products and their revision with the aid of DU8ML, a machine learning-augmented DFT computational method for fast and accurate calculations of NMR chemical shifts and coupling constants.

Simple, accurate, adjustable-parameter-free prediction of NMR shifts for molecules in solution

“…8,9,22,23 These approaches have proven remarkably successful, and computational modelling is now widely used in determining the structures and absolute stereochemistries of bioactive molecules isolated from natural products, and correcting misassignments that are prevalent in the literature. [24][25][26] However, all of these statistical models share a common limitation; they must be fully re-parameterised to account for changes in either the underlying quantum chemical method or the experimental conditions, i.e. they are not generally transferable.…”

“…8,9,22,23 These approaches have proven remarkably successful, and computational modelling is now widely used in determining the structures and absolute stereochemistries of bioactive molecules isolated from natural products, and correcting misassignments that are prevalent in the literature. 24–26…”

Machine learning-assisted structure annotation of natural products based on MS and NMR data

“…141 Subsequently, the team used DU8ML to correct the structures of a large number of misidentified NPs, which covered sesquiterpenoids, diterpenoids, steroids and alkaloids in terms of structural types. 141,142…”

“…141 Subsequently, the team used DU8ML to correct the structures of a large number of misidentied NPs, which covered sesquiterpenoids, diterpenoids, steroids and alkaloids in terms of structural types. 141,142 In instances where DFT calculations are conducted for two or more potential structures, the optimal structures can be determined using probability statistical models, such as CP3 and DP4. However, if only one candidate structure and its corresponding experimental NMR spectra exist, statistical data sorting is not applicable for judgment.…”

ent-Clavilactone J and Its Quinone Derivative, Meroterpenoids from the Fungus Resupinatus sp.

“…in 2022 as clavilactone J . Very recently, Novitskiy and Kutateladze revised the configuration of clavilactone J using a DFT computational NMR method . In our study, we back up this structural revision based on ROESY data.…”