Abstract
Inhibitors of the Sodium/Glucose co-transporter 2 (SGLT2) have been evolving into an important contribution to the treatment of diabetes mellitus. As the inhibition of SGLT2 is sensitive to the structural configuration at the sugar moiety of the inhibitors, it is of high interest to provide in silico-based methods for the prediction of the activity of potential SGLT2 inhibitors that take three-dimensional information into account. To attain this objective, a classification model based on the docking scores obtained from the best-performing docking-based virtual screening was created. Furthermore, the impact of ensemble docking using docking results from five SGLT2 structures and the incorporation of structural similarity information was assessed by creating classification models using these approaches. Taking a combined approach of docking score and structural similarity modelling led to the best performance with a Matthews Correlation Coefficient (MCC) of 0.64. Finally, to explore the ability of the used docking algorithms to correctly predict the influence of different three-dimensional information, a library of molecules with a negatively contributing configuration was created and docked, showing decreased docking scores for the molecule library with a disadvantaged configuration.