Predicting glycosylation stereoselectivity using machine learning.

Area of Science:

• Organic Chemistry
• Computational Chemistry
• Chemical Engineering

Background:

• Predicting stereochemical outcomes in complex chemical reactions, particularly glycosylations, remains a significant challenge.
• Glycosylation stereoselectivity is governed by numerous factors, including reagents, solvents, and temperature.
• Mechanistically ambiguous transformations necessitate advanced predictive tools.

Purpose of the Study:

• To develop a highly accurate predictive model for glycosylation stereoselectivity.
• To quantify the influence of various factors on glycosylation outcomes.
• To identify novel methods for controlling glycosylation stereoselectivity.

Main Methods:

• A random forest algorithm was employed, trained on a reproducible dataset.
• Steric and electronic contributions of reactants and solvents were calculated using quantum mechanics.
• Model predictions were validated experimentally using a microreactor platform.

Main Results:

• The developed model accurately predicts glycosylation stereoselectivities across diverse conditions (RMSE 6.8%).
• The model successfully predicted outcomes for previously unseen chemical participants and solvents.
• Environmental factors were found to have a greater influence on stereoselectivity than coupling partners.

Conclusions:

• A robust computational model can accurately predict glycosylation stereoselectivity.
• This approach offers insights into controlling glycosylation reactions and discovering new stereocontrol methods.
• Quantifying variable influences deepens the fundamental understanding of glycosylation mechanisms.