Machine learning meets mechanistic modelling for accurate prediction of experimental activation energies

Abstract

Accurate prediction of chemical reactions in solution is challenging for current state-of-the-art approaches based on transition state modelling with density functional theory. Models based on machine learning have emerged as...

Selectivity in organocatalysis—From qualitative to quantitative predictive models

“…In contrast, for larger data sets, RFs and NNs can be powerful approaches as they can identify patterns in complex data. In this context, Jorner et al have identified separate data regimes for the accurate prediction of S N Ar reaction barriers using a hybrid mechanistic/ML model 67 . They showed that for databases smaller than 50 samples traditional DFT modeling is the more accurate choice, while for data sets between 50 and 150 samples hybrid approaches are suitable.…”

Transformers for future medicinal chemists

“…The interested reader is referred to the TMAP (tree MAP 8 ) webpage 9 to try out the interactive reaction map visualizing on a sample dataset. It is early days, but as an example use Jorner et al 10 created a reaction map of nucleophilic aromatic substitution reactions and were able to cluster reactions with the same nucleophilic atoms and leaving atoms. Developing such a tool 9 is a laudable initiative since job descriptions for chemists trained in organic synthesis and in computational methods are merging into what will be the future definition of a medicinal chemist.…”

Navigating through the Maze of Homogeneous Catalyst Design with Machine Learning

“…24 More recently, a hybrid approach using both computational transition state modelling combined with machine learning has been shown to yield good accuracy for reproducing experimental Gibbs free energies of activation for nucleophilic aromatic substitution reactions, and this workflow is, in principle, also applicable to catalytic reactions. 25 Alternatively, in a series of recent papers, inspired by the classic Sabatier principle established in heterogeneous catalysis, 26,27 volcano plots were introduced as an effective tool to perform highthroughput virtual screening in homogeneous catalysis, and perform insightful result analysis at the same time, to find optimal catalysts. [28][29][30][31] These approaches, summarized in Figure 1, outline the status quo of computer-aided homogeneous catalyst design and show advances for individual pieces of the overall design workflow puzzle.…”