Assessment of Halogen Off-Center Point-Charge Models Using Explicit Solvent Simulations

Compounds containing halogens can form halogen bonds (XBs) with biological targets such as proteins and membranes due to their anisotropic electrostatic potential. To accurately describe this anisotropy, off-center point-charge (EP) models are commonly used in force field methods, allowing the description of XBs at the molecular mechanics and molecular dynamics level. Various EP implementations have been documented in the literature, and despite being efficient in reproducing protein-ligand geometries and sampling of XBs, it is unclear how well these EP models predict experimental properties such as hydration free energies (Delta G(hyd)), which are often used to validate force field performance. In this work, we report the first assessment of three EP models using alchemical free energy calculations to predict Delta G(hyd) values. We show that describing the halogen anisotropy using some EP models can lead to a slight improvement in the prediction of the Delta G(hyd) when compared with the models without EP, especially for the chlorinated compounds; however, this improvement is not related to the establishment of XBs but is most likely due to the improvement of the sampling of hydrogen bonds. We also highlight the importance of the choice of the EP model, especially for the iodinated molecules, since a slight tendency to improve the prediction is observed for compounds with a larger sigma-hole but significantly worse results were obtained for compounds that are weaker XB donors.

Automated summary

This study assessed the impact of three off-center point-charge (EP) models on the prediction of hydration free energies. The results showed that using certain EP models led to a slight improvement in the prediction of hydration free energies, especially for chlorinated compounds. However, this improvement was not related to the formation of halogen bonds but rather to the improvement in hydrogen bond sampling. The choice of EP model was particularly important for iodinated molecules, with compounds possessing a larger sigma-hole showing some improvement in prediction, while weaker XB donors showed significantly worse results.