Chemical space docking enables large-scale structure-based virtual screening to discover ROCK1 kinase inhibitors

Abstract

With the ever-increasing number of synthesis-on-demand compounds for drug lead discovery, there is a great need for efficient search technologies. We present the successful application of a virtual screening method that combines two advances: (1) it avoids full library enumeration (2) products are evaluated by molecular docking, leveraging protein structural information. Crucially, these advances enable a structure-based technique that can efficiently explore libraries with billions of molecules and beyond. We apply this method to identify inhibitors of ROCK1 from almost one billion commercially available compounds. Out of 69 purchased compounds, 27 (39%) have K_i values < 10 µM. X-ray structures of two leads confirm their docked poses. This approach to docking scales roughly with the number of reagents that span a chemical space and is therefore multiple orders of magnitude faster than traditional docking.

Fig. 1. Docked poses of selected starting fragment poses.

A representative set of docked poses (80 of the 500 initial fragments) are shown in the ROCK1 binding site. The binding pocket surface is shown in gray, and the hinge-binding pharmacophore is shown as two green spheres. Linkers are shown in light blue and serve as reaction vectors for product library enumeration (each fragment represents a library of ~10k molecules).

Fig. 2. Initial fragment hits.

Two-dimensional depictions of the initial fragment hits that led to one or more of the final 27 active molecule products. Key protein interactions with the ATP binding site are shown. Below each example is the list of active molecules in Supplementary Table 1 that were derived from the initial fragment hit.

Fig. 3. Docked poses for the most active molecules from each of the four chemotypes identified.

The hinge residues are at the left and the P loop (green) is at the top of each panel, while the original 2ETR ligand is shown in thin white sticks for reference. a Compound 1 (pyrazole) in orange. b Compound 16 (pyridone) in magenta. c Compound 22 (azaindole) in purple d. Compound 25 (indazole) in yellow. Hits identified by Chemical Space Docking all interact with the kinase P-loop, and two interact with the catalytic lysine (see a and d). Neither of these interactions is present in the PDB ligand.

Fig. 4. Comparison of the X-ray structure of compound 1 with its docked pose.

a X-ray structure: the refined 2Fo-Fc electron density is depicted at 1 sigma contour in the vicinity of the ligand. b Overlay of the docked pose (light green) and X-ray conformation (dark green).

Fig. 5. Comparison of the X-ray structure of compound 22 with its docked pose.

a X-ray structure: the refined 2Fo-Fc electron density is depicted at 1 sigma contour in the vicinity of the ligand. b Overlay of the docked pose (cyan) and X-ray conformation (bone).

Fig. 6. Computational requirements for large-scale docking campaigns.

Traditional full enumeration docking curves are calculated based on the time needed to dock each molecule: 10 s (red), 1 s (yellow), 0.1 s (green). Chemical Space Docking curve shown in light blue. Large-scale docking campaigns from the literature are shown as individual data points (including Space Docking results described here and similar campaigns)^,,,,.

Fig. 7. The ATP binding site of ROCK1 (PDB ID: 2ETR).

The pharmacophore constraint for hydrogen bond interaction with the kinase hinge residues is shown as green spheres. A different kinase ligand (PDB ID 3V8S) is shown to illustrate both possible hinge interactions that are consistent with the pharmacophore constraint.