Statistical efficiency of methods for computing free energy of hydration

Area of Science:

• Computational Chemistry
• Physical Chemistry
• Biophysical Chemistry

Background:

• Hydration free energy (HFE) is crucial for understanding molecular behavior in water.
• Accurate HFE calculation is vital for drug discovery and biochemical process modeling.
• Various computational methods exist, broadly categorized as equilibrium or non-equilibrium.

Purpose of the Study:

• To compare the accuracy and efficiency of different computational methods for HFE calculation.
• To evaluate one-sided non-equilibrium (Jarzynski methods), two-sided non-equilibrium (CGI, BAR), and equilibrium (MBAR) methods.
• To assess performance against experimental HFE data for 34 small organic molecules.

Main Methods:

• Jarzynski Forward (JF) and Backward (JB) methods (one-sided non-equilibrium).
• Jarzynski mean, Crooks Gaussian Intersection (CGI), and Bennett Acceptance Ratio (BAR) (two-sided non-equilibrium).
• Multistate Bennett Acceptance Ratio (MBAR) (two-sided equilibrium) as the reference, alongside experimental data.

Main Results:

• All tested methods produced HFE estimates consistent with the MBAR reference.
• Mean absolute errors were approximately 0.8 kcal/mol, and root mean square errors were around 1 kcal/mol.
• One-sided Jarzynski methods (JF, JB) exhibited systematic deviations, and MBAR demonstrated the highest statistical efficiency.

Conclusions:

• All evaluated computational methods are reliable for HFE prediction of small organic molecules.
• Equilibrium methods, particularly MBAR, offer superior computational efficiency compared to non-equilibrium methods.
• MBAR is recommended as a reference method due to its accuracy and efficiency.