ANI neural network potentials for small molecule pK_a prediction.

Area of Science:

• Computational Chemistry
• Physical Chemistry
• Chemical Physics

Background:

• The acid dissociation constant (pKa) is crucial in diverse chemical fields like pharmacology and environmental science.
• Existing experimental (NMR, titration) and computational (DFT) methods for pKa determination are often time-consuming and resource-intensive.

Purpose of the Study:

• To develop a rapid and computationally efficient method for calculating the pKa values of small molecules.
• To demonstrate the utility of neural network potentials in predicting thermodynamic cycles for pKa calculations.

Main Methods:

• Utilized a combination of neural network potentials, low-energy conformer searches, and thermodynamic cycles.
• Trained neural network potentials on various phase and charge states to predict molecular energy cycles.
• Focused on imidazolium-derived carbene species for method development.

Main Results:

• Successfully developed a rapid computational approach for pKa prediction.
• Demonstrated that neural network potentials trained on different states can accurately predict thermodynamic energy cycles.
• The method shows promise for efficient pKa determination of small molecules.

Conclusions:

• The novel method provides a significantly faster and less computationally expensive alternative for pKa determination.
• The approach is adaptable and can be extended to other chemical species, such as amines, with further model training.
• This work advances computational chemistry by offering a more accessible tool for predicting a fundamental molecular property.