Figure 1.

Overview of the adaptive sampling protocol. Initially, multiple independent molecular dynamics simulations are run of an input system with multiple copies of the ligand in solution. All configurations of the molecular system observed in simulation trajectories are clustered, then used to update a statistical model. Clusters are selected for resampling in the next round by assigning a score to each cluster, and new molecular system configurations are built from selected clusters as starting points for another round of simulation.