Docking and scoring with ICM: the benchmarking results and strategies for improvement

Abstract

Flexible docking and scoring using the internal coordinate mechanics software (ICM) was benchmarked for ligand binding mode prediction against the 85 co-crystal structures in the modified Astex data set. The ICM virtual ligand screening was tested against the 40 DUD target benchmarks and 11-target WOMBAT sets. The self-docking accuracy was evaluated for the top 1 and top 3 scoring poses at each ligand binding site with near native conformations below 2 Å RMSD found in 91 and 95% of the predictions, respectively. The virtual ligand screening using single rigid pocket conformations provided the median area under the ROC curves equal to 69.4 with 22.0% true positives recovered at 2% false positive rate. Significant improvements up to ROC AUC = 82.2 and ROC((2%)) = 45.2 were achieved following our best practices for flexible pocket refinement and out-of-pocket binding rescore. The virtual screening can be further improved by considering multiple conformations of the target.

Fig. 1

RMSD values for ICM predictions on all ligand binding sites. Red bars indicate RMSD values for the top 1 scoring poses, whereas blue bars indicate the lowest RMSD among the top 3 scoring poses. PBD entries with binding mode predictions above 2 Å RMSD are labeled.

Fig. 2

Ligand binding mode predictions above 2 Å RMSD. Top 1 scoring poses are represented with cyan carbons and magenta oxygens whereas co-crystal ligands are represented with yellow carbons and red oxygens. Crystallographic residues, water molecules, metal ions and other small molecules found in the binding site are labeled and numbered as they appear in the PDB files. Water molecules were excluded for docking purposes. Hydrogen bonds are represented with spheres and colored according to the estimated energy (blue – strong interaction, red – weak interaction). Fig. G shows the predicted binding modes for two sites in PDB entry 1w1p. Fig. H displays an alternative co-crystal binding mode represented with orange carbons. Electron density maps at 1.0 and 2.5 sigma levels are represented with yellow and blue meshes, respectively, around the co-crystal ligands of PDB entries 1w1p and 1tz8.

Fig. 3

Virtual ligand screening benchmark of 40 DUD test sets docked against the original target coordinates using the default ICM scoring method (A), or docked against refined induced fit models followed by out-of-pocket rescoring (B). ROC AUC values were calculated for the true ligands (blue) and null hypothesis (gray). Early enrichments are reported as the fraction of true positives recovered at 0.1% (blue), 1% (red) and 2% (green) false positive rates. A list of abbreviations is provided as Supporting Information.

Fig. 4

Virtual ligand screening benchmark of 11 WOMBAT test sets docked against the original target coordinates using the default ICM scoring method (A), or docked against refined induced fit models followed by out-of-pocket rescoring (B). ROC AUC values obtained with the WOMBAT test sets (blue) are compared with the corresponding DUD results (gray). Early enrichments for the WOMBAT test sets are reported as the fraction of true positives recovered at 0.1% (blue), 1% (red) and 2% (green) false positive rates. A list of abbreviations is provided as Supporting Information.