A Testing Bed for Computational Modelling of Host-guest Binding

Host-guest modelling remains rather challenging in computational chemistry. While a batch of ‘techniques’ could be found to handle this problem in current literatures, the practical predictive power remains unsatisfactory. Many blind prediction sets like previous SAMPL challenges provide good testing beds, but the coverage of chemical space and the sample size often seem insufficient. Here, a testing bed for host-guest binding is reported. The host targets are pillararene derivatives with varying cavity size (5-7 repeating units) and substitution groups (WPn with n=5-7 and SP6). For each host target, tens of guest molecules with diverse chemical compositions have experimentally measured binding data extracted from references. This dataset serves as a nice testing bed for benchmarking practical performance of molecular simulation techniques. I additionally report numerical data of a variety of docking and end-point screening techniques widely applied in this field, providing some basic picture about the behavior of common practices. Finally, I would express my sincere gratitude to Prof. Piero Procacci for valuable helps, discussions and insightful suggestions.