In silico prediction of binding capacity and interaction forces of organic compounds with alpha- and beta-cyclodextrins

As commonly-used cyclic oligomers, cyclodextrins are usually composed of 6 and 7 glucopyranose units (alpha-cyclodextrin and beta-cyclodextrin). The binding constant is a significant parameter to estimate the association stability of inclusion complexes and evaluate the influence of cyclodextrins on guest compounds. In order to cover the shortage of experimental methods, we intended to predict the binding constants and interaction forces of cyclodextrin inclusion complexes by developing consensus models and calculating energy changes. 418 cyclodextrin inclusion complexes were collected with their experimental binding constants. By combining 6 description methods and 5 machine learning methods, we built 60 single models, which were selected to develop 104 consensus models. The coefficients of determination of the test sets in the best alpha-cyclodextrin consensus model and the best beta-cydodextrin consensus model were 0.84 and 0.89, respectively. The applicability domains of models were defined. After performing molecular docking and calculating energy changes, we could observe the binding poses, identify the interaction forces, and select the dominant forces. The consensus models and the calculation of energy changes can be applied to explore the binding capacity and interaction forces of inclusion complexes. (C) 2020 Elsevier B.V. All rights reserved.