RESUMO
Host-guest complex has attracted much attention because of their fantastic capability. Accurate prediction of their binding affinity and enthalpy is essential to the rational design of guest molecules. The attach-pull-release (APR) method proposed by Henriksen et al. (J. Chem. Theory Comput., 2015, 11:4377.) shows good prediction capability of binding affinity especially for host-guest system. In order to further evaluate the performance of APR method in practice, we have conducted the calculations on the macrocycle cucurbit [7]urils (CB7) encapsulated with four structurally similar alkaloids (berberine, coptisine, epiberberine and palmatine) with two force fields (GAFF and GAFF2) and three water models (TIP3P, SPC/E and OPC). Compared to the experimental data, the calculation by the combination of GAFF2 and SPC/E force field presents the best performance, of which the Pearson correlation coefficients (R2) is 0.95, and the root-mean-square-deviation is 3.04 kcal/mol. While the predictions from GAFF force field all overestimated the binding affinity, suggesting a systematic error may be involved. Comparison of calculation also indicates that the accuracy of prediction was susceptible to the combination of force field. Therefore, it would be necessary to repeat the simulation with different combination of force fields in practice.