Skip to content
PaperFren

Computational chemistry

ML plus transition states predict SNAr barriers

Jorner K, Brinck T, Norrby PO, et al. · Chemical science · 2021

Open access · cc by · source: Europe PMC

A Gaussian-process model that includes SNAr transition-state features predicts experimental activation energies with 0.77 kcal mol−1 MAE, beating raw DFT.

Study at a glance

Design
Computational / modelling — DFT/physical-organic descriptors with GPR for SNAr barriers and selectivity
N
ML on reaction barriers/selectivity; learning curves emphasize 50–150 training samples — no single cohort N
Population
Nucleophilic aromatic substitution reactions (including patent-selectivity cases)
Outcome
Predicted activation free energies and regio-/chemoselectivity accuracy

Structured fields used in claim comparison tables when every cited study has a complete layer.

Key findings

GPR (Matern 3/2) reaches R2 = 0.93 and MAE = 0.77 kcal mol−1 on a held-out set, below 1 kcal mol−1 chemical accuracy. Raw DFT MAE is 2.93. Selectivity top-1 accuracy is 86%. Hybrid models help most with 50–150 samples.

Methodology

Authors combined physical-organic and DFT transition-state descriptors for nucleophilic aromatic substitutions, trained several regressors, and tested regio-/chemoselectivity on patent reactions plus learning curves versus sample size.

Limitations

Acid/base catalysis and other reaction classes are not yet in the model; TS geometry still needs reliable computation, which the authors call the bottleneck.

How this study connects

Role on claims

Each row is a claim on a concept or method page where this paper supports, challenges, or qualifies the statement. Roles are hand-checked — not a model guess.

  • This library holds 12 empirical chemistry papers on computational chemistry with isolated findings, rates or spectra rather than reviews.

    Evidence for the claim as stated.

  • A Gaussian-process model that includes SNAr transition-state features predicts experimental activation energies with 0.77 kcal mol−1 MAE, beating raw DFT.

    Evidence for the claim as stated.

  • ML trained on DFT descriptors can beat raw DFT on a held-out SNAr set and still depend on being able to compute the transition state. A Gaussian-process model reached R² = 0.93 and MAE = 0.77 kcal mol⁻¹ versus raw DFT MAE 2.93; selectivity top-1 accuracy was 86%. Acid/base catalysis and other reaction classes were not in the model; TS geometry remains the bottleneck.

    Evidence for the claim as stated.

  • These papers do not agree on what DFT is for. One trains a statistical model to correct DFT barriers; one trains a net to emulate DFT spin states; others use a single-point or periodic calculation to rationalise a crystal, a gas-phase ion, or a ligand redox event. Quoting a kcal mol⁻¹ figure without saying whether it is raw DFT, a GP correction, or an NNP barrier mixes those jobs.

    Evidence for the claim as stated.

Open questions

Tensions this paper is part of

From concept pages' “where studies disagree.” Disagreement means the same question; scope means different assays, populations, or outcomes.

Related papers in this topic

Same topic cluster — not a recommendation engine.