Validation of the Generalized Force Fields GAFF, CGenFF, OPLS-AA, and PRODRGFF by Testing Against Experimental Osmotic Coefficient Data for Small Drug-Like Molecules

Accurate force field parameters for proteins and drugs are of fundamental importance for predicting protein-drug binding poses and binding free energies in structure-based drug design. Osmotic pressure simulations have recently gained popularity as a means to validate common force fields for biomolecules such as proteins; up to now, however, little such effort has been made to use these methods to validate common force fields for small drug-like molecules. In this work, osmotic coefficients of 16 drug-like molecules were calculated from osmotic pressure molecular dynamics simulations using four common force fields for drug-like molecules: GAFF, CGenFF, OPLS-AA, and PRODRGFF. While GAFF, CGenFF, and OPLS-AA produced osmotic coefficients that are in good agreement with the corresponding experimental osmotic coefficients, PRODRGFF, which attempts to use GROMOS parameters, often produced osmotic coefficients that are in poor agreement. All four force fields poorly reproduced the experimental osmotic coefficients of purine-derived molecules including purine, 6-methylpurine, and caffeine, suggesting common issues in describing interactions of this particular molecule type. The poor overall performance of PRODRGFF can be mainly attributed to the poor results with two macrocyclic molecules HMT and TAT, which can be significantly improved by reparameterizing the van der Waals (vdW) parameters of alkyl carbons and using lower solute concentration. In addition, the recently developed GAFF2 with revamped vdW parameters was found to produce osmotic coefficients that are in slightly better agreement with experiments than GAFF. Overall, the four common force fields for drug-like molecules tested in this study performed reasonably well at reproducing experimental osmotic coefficients of drug-like molecules, although failures of certain force fields on certain molecules suggest that further force field reparameterizations, especially for vdW parameters, might be required to improve their accuracy.