Fig. 2From: Correcting the impact of docking pose generation error on binding affinity predictionBox plots of performance of the four models trained on crystal and docked poses and tested also on crystal and docked poses of the PDBbind v2007 benchmark. Model 1 is AutoDock Vina, model 2 is MLR::Vina, model 3 is RF::Vina, and model 4 is RF::VinaElem. In the x axis, trn:1 means the four models were trained in scheme 1, i.e. on crystal poses, and trn:2 means the four models were trained in scheme 2, i.e. on docked poses. Likewise, tst-1 and tst-2 mean the four models were tested on crystal and docked poses, respectively. Model 1 was executed out-of-the-box, so its results for training scheme 1 were repeated for training scheme 2Back to article page