Fig 1.
The reverse docking results of 5000 molecules.
The probability density curves of (a) DOCK score, (b) Glide score, (c) Vina score of 100 proteins (different colors represent different proteins). (d) The hit frequence, (e) the average rank of 100 proteins in 5000 reverse docking cases.
Fig 2.
The relationship between median docking score and hit frequence of proteins.
The scatter diagrams of (a) median Dock score, (b) median Glide score, (c) median Vina score of 100 proteins with their hit frequence.
Table 1.
The proteins with top-10 hit frequence in 5000 reverse docking cases.
Fig 3.
The property analysis by decision trees prediction.
The decision trees analysis of interference proteins and underrated proteins in the reverse docking by (a) DOCK, (b) Glide, (c) AutoDock Vina.
Fig 4.
The correlationship between median docking score and the properties of protein pockets.
The reverse docking by (a) DOCK; (b) Glide; (c) AutoDock Vina.
Fig 5.
The relationship between median docking score and the highly relevant protein pocket properties of three classes of proteins.
The scatter diagram of median DOCK score with (a) median contact area, (b) volume of protein pockets. The scatter diagram of median Glide score with (c) median contact area, (d) phobic of protein pockets. The scatter diagram of median Vina score with (e) median contact area, (f) size of protein pockets. The fitting lines of scatter points are shown in the diagram.
Fig 6.
The probability density curves for the highly relevant protein pocket properties of three classes of proteins.
The distribution of (a) median contact area, (b) volume of three classes of protein pockets in the reverse docking by DOCK. The distribution of (c) median contact area, (d) phobic of three classes of protein pockets in the reverse docking by Glide. The distribution of (e) median contact area, (f) size of three classes of protein pockets in the reverse docking by AutoDock Vina.
Fig 7.
Analysis of score normalization results for benchmark dataset.
(a) The average rank, (b) the hit frequence of 100 proteins in 5000 reverse docking cases.
Fig 8.
The probability density curves for the properties of three classes of protein pockets after score normalization.
The distribution of (a) median contact area, (b) volume of three classes of protein pockets after score normalization for DOCK. The distribution of (c) median contact area, (d) phobic of three classes of protein pockets after score normalization for Glide. The distribution of (e) median contact area, (f) size of three classes of protein pockets after score normalization for AutoDock Vina.
Table 2.
The target prediction results in the reverse docking by DOCK, Glide and AutoDock Vina.
Fig 9.
Analysis of the rank difference of cocrystallized proteins before and after the score normalization of the refined Astex Diverse Set.
The rank difference of cocrystallized proteins for (a) DOCK, (b) Glide, and (c) AutoDock Vina.
Fig 10.
Analysis of the average ranks of four classes of proteins before and after the score normalization of the refined Astex Diverse Set.
The boxplot of the average ranks for (a) DOCK, (b) Glide and (c) AutoDock Vina.