The HYDROPHOBE Challenge: A Joint Experimental and Computational Study on the Host-Guest Binding of Hydrocarbons to Cucurbiturils Allowing Explicit Evaluation of Guest Hydration Free Energy Contributions

摘要

The host-guest complexation of hydrocarbons (22 guest molecules) with cucurbit[7]uril (CB7) was investigated in aqueous solution. Association constants were determined by using the indicator displacement strategy, which allows binding constant determinations also for poorly water-soluble (hydrophobic) guests. The binding constants (10³–10⁹ M⁻¹) increased with the size of the hydrocarbon, pointing to the hydrophobic effect and dispersion interactions as driving forces. Besides potential applications for the sensing and separation of hydrocarbons, the measured affinities provide unique benchmark data for the binding of neutral guest molecules. Consequently, a computational blind challenge, the HYDROPHOBE challenge, was conducted in order to allow a comparison with state-of-the-art computational methods for predicting host-guest affinity constants. In total, 5 computational data sets were submitted, which allowed the comparison of experimental binding constants with those predicted by coupled-cluster theory (DLPNO-CCSD(T)), dispersion-corrected density functional theory (DFT), and explicit solvent molecular dynamics (MD) simulations parameterized with two different force field combinations from the AMBER simulation package. All submissions were capable of predicting the general binding trend, with a slightly better correlation for the MD compared to the quantum-chemical (QM) data sets (R²MD = 0.80 vs R²QM = 0.66, average values for the submitted data sets). On the other hand, QM calculations showed better predictions for the absolute values of the binding affinities as reflected by the mean signed errors (4.3 kcal mol⁻¹ for MD vs 1.8 kcal mol⁻¹ for QM). When searching for sources of uncertainty in predicting the host-guest affinities, the experimentally known hydration energies of the investigated hydrocarbons could be employed, which provided a distinct advantage of the HYDROPHOBE challenge. The comparison with the employed solvation models (explicit solvent for MD and COSMO-RS for QM) confirmed a good correlation for both methods, but revealed a rather constant offset of the COSMO data, by ca. +2 kcal mol⁻¹, which was traced back to a required reference-state correction in the QM submissions (2.38 kcal mol⁻¹). Introduction of the reference-state correction improved the predictive power of the QM methods, particularly for small hydrocarbons up to C5. The correlations of both QM and MD submissions also exposed specific outliers, which could be due to peculiarities of the investigated guests, for example, different degrees of conformational changes upon complexation, such as helical structures of the longer n-alkyl chains within the cavity. The latter was confirmed by 2D NMR experiments and both the MD as well as QM calculations.