Machine learning-guided computational screening of new bio-orthogonal click reactions

TL;DR

This paper introduces a machine learning-based computational workflow that efficiently screens over 10 million potential bio-orthogonal click reactions, predicting their activation energies to identify promising candidates for experimental validation.

Contribution

The study presents a novel machine learning approach to rapidly predict activation energies in a vast chemical space, enabling efficient discovery of new bio-orthogonal click reactions.

Findings

Successfully sampled 0.05% of the search space

Predicted activation energies within ~2-3 kcal/mol accuracy

Identified diverse candidate reactions for experimental testing

Abstract

Bio-orthogonal click chemistry has become an indispensable part of the biochemist's toolbox. Despite the wide variety of applications that have been developed in recent years, only a limited number of bio-orthogonal click reactions have been discovered so far, most of them based on (substituted) azides. In this work, we present a computational workflow to discover new candidate bio-orthogonal click reactions. Sampling only around 0.05\% of an overall search space of over 10,000,000 dipolar cycloadditions, we develop a machine learning model able to predict DFT-computed activation and reaction energies within ~2-3 kcal/mol across the entire space. Applying this model to screen the full search space through iterative rounds of learning, we identify a broad pool of candidate reactions with rich structural diversity, which can be used as a starting point or source of inspiration for future experimental development of both azide-based and non-azide-based bio-orthogonal click reactions.