How accurate are continuum solvation models for drug-like molecules?

Area of Science:

• Computational Chemistry
• Molecular Modeling
• Drug Discovery

Background:

• Accurate estimation of hydration free energy is crucial for drug design.
• Continuum solvation models are widely used but their accuracy for drug-like molecules is debated.
• Drug-like molecules present unique challenges due to their polarity.

Purpose of the Study:

• To evaluate the accuracy of 24 different continuum solvation methods for drug-like molecules.
• To compare the performance of polarized continuum, Langevin dipole, finite-difference Poisson, and generalized Born methods.
• To assess the impact of molecular polarity and charge on solvation energy prediction accuracy.

Main Methods:

• Applied four continuum solvent approaches: polarized continuum, Langevin dipole, finite-difference Poisson, and generalized Born.
• Tested 24 variants of these methods on 20 neutral drug-like molecules and three inhibitor series.
• Calibrated methods using experimentally known small organic molecules and ions.

Main Results:

• Methods showed good calibration (1-6 kJ/mol MAD) for small molecules but lower accuracy for polar drug-like molecules.
• Relative energy calculations within inhibitor series significantly improved accuracy (2-5 kJ/mol MAD).
• Generalized Born methods generally yielded better results than Poisson-Boltzmann methods.

Conclusions:

• Continuum solvation accuracy should be assessed relatively, especially for drug-like molecules.
• Relative free energy calculations are reliable for ranking inhibitor binding affinities.
• Generalized Born methods offer a promising approach for predicting relative solvation energies in drug discovery.