Evaluating Classical Force Fields against Experimental Cross-Solvation Free Energies

Area of Science:

• Computational Chemistry
• Molecular Modeling
• Physical Chemistry

Background:

• Experimental solvation free energies are crucial for validating condensed-phase force fields.
• Current validation methods are often unsystematic, using limited solutes and solvents.

Purpose of the Study:

• To systematically assess the accuracy of popular molecular force fields.
• To introduce and utilize a comprehensive cross-solvation free energy matrix for force field evaluation.

Main Methods:

• Constructed a matrix of 625 cross-solvation free energies (ΔG_A:B⊖) for 25 diverse molecules.
• Evaluated four force fields: GROMOS-2016H66, OPLS-AA, AMBER-GAFF, and CHARMM-CGenFF.
• Compared force field predictions against curated experimental data.

Main Results:

• All four force fields demonstrated comparable performance in predicting solvation free energies.
• Root-mean-square errors ranged from 2.9 to 4.0 kJ·mol⁻¹.
• Average errors were between -0.8 and +1.0 kJ·mol⁻¹, with AMBER and CHARMM showing the lowest magnitudes.

Conclusions:

• The evaluated force fields exhibit similar accuracy for the tested molecular systems.
• Differences in performance are statistically significant but not substantial.
• Potential inaccuracies in experimental data may influence outlier results.