SAMPL5: 3D-RISM partition coefficient calculations with partial molar volume corrections and solute conformational...

Area of Science:

• Computational chemistry
• Molecular modeling
• Physical chemistry

Background:

• Implicit solvent models are widely used for fast, physics-based hydration free energy calculations of macromolecules.
• The transferability of these implicit solvent models to solvents other than water is less explored.
• The Statistical Assessment of Modeling of Proteins and Ligands 5 (SAMPL5) dataset offers a benchmark for testing solvent model transferability.

Purpose of the Study:

• To evaluate the transferability of implicit solvent models to a non-aqueous solvent (cyclohexane).
• To assess the performance of the 3D reference interaction site model (3D-RISM) theory with a new cyclohexane model for partition coefficient calculations.
• To investigate the impact of solute conformational sampling and corrections for ionization/tautomers on prediction accuracy.

Main Methods:

• Utilized the 3D reference interaction site model (3D-RISM) statistical-mechanical solvation theory.
• Developed and employed a new united atom cyclohexane model alongside a validated water model.
• Calculated partition coefficients for the SAMPL5 dataset using single solute conformations and subsequently with molecular dynamics (MD) coupled with 3D-RISM (MD/3D-RISM).
• Applied corrections for solvation free energy, ionization, and tautomers.

Main Results:

• The cyclohexane model showed good performance on training and testing data for amino acid analogues but required a parameterized correction.
• Single conformation 3D-RISM calculations performed poorly on the SAMPL5 dataset, likely due to large solute sizes.
• MD/3D-RISM improved agreement with reference partition coefficients to [Formula: see text].
• A simple [Formula: see text] correction improved agreement with experimental partition coefficients from [Formula: see text] to [Formula: see text], despite some outliers.

Conclusions:

• Implicit solvent models exhibit limited transferability to non-aqueous solvents, necessitating model-specific parameterization or corrections.
• Solute conformational sampling via MD/3D-RISM enhances partition coefficient predictions, particularly for larger solutes.
• Accurate prediction of partition coefficients requires accounting for solute ionization states and tautomers, alongside improved solvation models.