Assigning the stereochemistry of natural products by machine learning

Our recent paper Assigning the stereochemistry of natural products by machine learning is now published in the Journal of Cheminformatics!

Abstract

Nature has settled for L-chirality for proteinogenic amino acids and D-chirality for the carbohydrate backbone of nucleotides. Further stereochemical patterns exist among natural products produced by common biosynthetic pathways. Here we asked the question whether these regularities might be sufficiently prevalent among natural products (NPs) such that their stereochemistry could be machine learned and assigned automatically. Indeed, we report that a language model can be trained to assign the stereochemistry of NPs using the open access NP database COCONUT. In detail, our language model, called NPstereo, translates an NP structure written as absolute SMILES into the corresponding isomeric SMILES notation containing stereochemical information, with 80.2% per-stereocenter accuracy for full assignments and 85.9% per-stereocenter accuracy for partial assignments, across various NP classes including secondary metabolites such as alkaloids, polyketides, lipids and terpenes. NPstereo might be useful to assign or correct the stereochemistry of newly discovered NPs.

Author(s) Markus Orsi, and Jean-Louis Reymond