Search

Empirical force field methods typically rely on point charges to describe the electrostatic interactions, which is problematic when anisotropy needs to be considered, as in the case of the electrostatic potential of covalently bound halogens that possess a positive site, termed σ -hole, surrounded by a large negative belt. To address this, an off-center point charge (extra point, EP) is usually placed at a given distance from the halogen to emulate the σ -hole and commonly used implementations are based on the restrained electrostatic potential (RESP) procedure to fit atomic charges, being one of the most used charge models. In this context, no specific Merz–Singh–Kollman (MK) radius for iodine is available in the literature, which is an essential parameter in the RESP fitting procedure. In this work, we explored the impact of the iodine MK radius on the obtained RESP charges for a set of 12 iodinated molecules. We verified that the relative root mean square (RRMS) values obtained with and without an EP kept decreasing with increasing radii for most compounds, thus impairing optimization using such a procedure. Nevertheless, the use of an iodine MK radius lower than 2 Å is not advisable since the RRMS kept decreasing considerably until this value was reached. Moreover, the performance of three iodine MK radii was studied with the estimation of the free energy of hydration (Δ G hyd) values using alchemical free energy calculations, which are particularly sensitive to the charges used. Despite the usage of different radii not leading to remarkable differences, our results indicate that using a value of 2.70 Å leads to lower mean absolute errors (MAE) and root mean squared error (RMSE) values when comparing the calculated with the experimental Δ G hyd values. Restrained Electrostatic Potential (RESP) charges are commonly used in force fields and Merz-Singh-Kollman (MK) radii are needed in the fitting. Since iodine radius is not available, in this work, we explored the impact of changing the iodine MK radius on the RESP charges of a set of 12 iodinated molecules and their hydration free energy predictions. Our results indicate that, although the changes are small, a value of 2.70 Angstrom leads to a better agreement with experimental values. [ABSTRACT FROM AUTHOR]

Halogen bonds (XBs) have become increasingly popular over the past few years with numerous applications in catalysis, material design, anion recognition, and medicinal chemistry. To avoid a post factum rationalization of XB trends, descriptors can be tentatively employed to predict the interaction energy of potential halogen bonds. These typically comprise the electrostatic potential maximum at the tip of the halogen, VS,max, or properties based on the topological analysis of the electronic density. However, such descriptors either can only be used with confidence for specific families of halogen bonds or require intense computations and, therefore, are not particularly attractive for large datasets with diverse compounds or biochemical systems. Therefore, the development of a simple, widely applicable, and computationally cheap descriptor remains a challenge as it would facilitate the discovery of new XB applications while also improving the existing ones. Recently, the Intrinsic Bond Strength Index (IBSI) has been proposed as a new tool to evaluate any bond strength, however, it has not been extensively explored in the context of halogen bonding. In this work, we show that IBSI values linearly correlate with the interaction energy of diverse sets of closed-shell halogen-bonded complexes in the ground state, and therefore, can be used to quantitatively predict this property. Although the linear fit models that use quantum-mechanics-based electron density provided MAEs typically below 1 kcal mol−1, this type of calculation might still be computationally heavy in large sets or systems. Therefore, we also explored the exciting possibility of using a promolecular density approach (IBSIPRO), which only requires the geometry of the complex as an input, being computationally cheap. Surprisingly, the performance was comparable to the QM-based methods, thus opening the door for the usage of IBSIPRO as a fast, yet accurate, XB energy descriptor in large datasets but also in biomolecular systems such as protein–ligand complexes. We also show that the δgpair descriptor emerging from the Independent Gradient Model that leads to IBSI can be seen as a term proportional to the overlapping van der Waals volume of the atoms at a given interaction distance. Overall, ISBI can be thought of as a complementary descriptor to VS,max for situations where the geometry of the complex is available and QM calculations are not feasible whereas the latter still remains a hallmark of XB descriptors. [ABSTRACT FROM AUTHOR]