Dense Neural Network for Calculating Solvation Free Energies from Electronegativity-Equalization Atomic Charges

Area of Science:

• Computational Chemistry
• Physical Chemistry
• Machine Learning in Chemistry

Background:

• Accurate prediction of solvation free energies is crucial for understanding chemical processes.
• Existing methods, particularly density functional theory (DFT)-based approaches, can be computationally intensive.
• There is a need for efficient and accurate models for evaluating solvation free energies across various solvent types.

Purpose of the Study:

• To introduce a novel deep neural network model, ESE-EE-DNN, for the evaluation of solvation free energies.
• To assess the accuracy and efficiency of ESE-EE-DNN compared to established computational methods.
• To demonstrate the model's applicability to both neutral molecules and ionic species in diverse solvent environments.

Main Methods:

• Development of a dense Neural Network (NN) model named Easy Solvation Energy with Electronegativity Equalization charges and Dense Neural Network (ESE-EE-DNN).
• Utilizing Conductor-like Screening Model (COSMO) electrostatic energy, atomic cavity surface areas, total cavity volume, and induced surface charges as input features.
• Employing electronegativity-equalization atomic charges for the COSMO calculations.

Main Results:

• ESE-EE-DNN achieved high accuracy, with root-mean-square errors (RMSE) of 1.25 kcal/mol (water), 1.36 kcal/mol (polar protic), 0.70 kcal/mol (polar aprotic), and 0.71 kcal/mol (nonpolar) for neutral solutes.
• The model demonstrated particular strength for ionic solutes, yielding RMSEs of 2.82 kcal/mol (aqueous) and 1.42 kcal/mol (nonaqueous).
• ESE-EE-DNN showed comparable or superior accuracy to mainstream DFT-based methods and offered significant computational efficiency.

Conclusions:

• ESE-EE-DNN provides a highly accurate and efficient approach for predicting solvation free energies.
• The model's performance across different solute types (neutral and ionic) and solvent environments highlights its versatility.
• The computational efficiency, stemming from fast electronegativity-equalization charge evaluation, makes ESE-EE-DNN a valuable tool for computational chemistry research.