“…In order to assess the quality of the obtained poses, empirical, forcefield-based, knowledge-based or consensus scoring functions can be used [46]. Whereas pose generation works reliably to sample possible ligand geometries, the performance of scoring functions heavily varies depending on the target protein, mainly due to the challenging task of predicting the entropic component of ligand binding [26]. The main attempts for docking in drug design are virtual screening, binding mode prediction and the estimation of ligand binding affinity.…”