A Dataset of Plausible Proton Transfer Steps for Arrow-Pushing Mechanisms

Area of Science:

• Chemical kinetics
• Computational chemistry
• Reaction mechanisms

Background:

• Proton transfers are fundamental to polar reaction mechanisms.
• Accurate prediction of reaction pathways is essential in chemistry.

Purpose of the Study:

• To create a comprehensive dataset of proton transfer steps.
• To facilitate the development of machine learning models for predicting reaction mechanisms.

Main Methods:

• Generated over 51 million proton transfer steps between heteroatoms and carbon acids.
• Estimated rate constants using simplified Eigen and Eigen-Bernasconi equations.
• Encoded data in SMIRKS format for machine learning compatibility.

Main Results:

• Compiled a dataset of kinetically plausible proton transfer steps with estimated rate constants (≥10³ M⁻¹ s⁻¹).
• Included transfers between heteroatoms and from carbon acids to heteroatom bases.
• Prioritized structural diversity for broader applicability in aqueous environments.

Conclusions:

• The generated dataset is valuable for training machine learning models.
• This resource can improve the prediction of stepwise reaction mechanisms.
• The approach provides a foundation for computational studies in chemical kinetics.