Fig. 1: Comparison of sampling and scoring ability.
From: Augmented BindingNet dataset for enhanced ligand binding pose predictions using deep learning
A The sampling success rate is defined as any ligand RMSD less than 2.0 Å. “Superimposition without minimization” and “Superimposition with minimization” represent the hierarchal template-based modeling approach without and with MM-GB/SA minimization. The numbers in brackets indicate the count of benchmarking systems within each MCS coverage interval. B The scoring success rate is defined as the top 1 ligand RMSD less than 2.0 Å. The descriptions before and after “/” denote the sampling and scoring algorithms. “Superimposition” refers to the hierarchal template-based modeling approach with MM-GB/SA minimization. The red dashed lines represent the sampling success rates of different sampling methods. C The plot illustrates the relationship among ligand RMSDs, hybrid scores, and MCS coverages. The labels represent the scoring success rates within different hybrid score intervals.
