Improving docking and virtual screening performance using AlphaFold2 multi-state modeling for kinases

Abstract

Structure-based virtual screening (SBVS) is a crucial computational approach in drug discovery, but its performance is sensitive to structural variations. Kinases, which are major drug targets, exemplify this challenge due to active site conformational changes caused by different inhibitor types. Most experimentally determined kinase structures have the DFGin state, potentially biasing SBVS towards type I inhibitors and limiting the discovery of diverse scaffolds. We introduce a multi-state modeling (MSM) protocol for AlphaFold2 (AF2) kinase structures using state-specific templates to address these challenges. Our comprehensive benchmarks evaluate predicted model qualities, binding pose prediction accuracy, and hit compound identification through ensemble SBVS. Results demonstrate that MSM models exhibit comparable or improved structural accuracy compared to standard AF2 models, enhancing pose prediction accuracy and effectively capturing kinase-ligand interactions. In virtual screening experiments, our MSM approach consistently outperforms standard AF2 and AF3 modeling, particularly in identifying diverse hit compounds. This study highlights the potential of MSM in broadening kinase inhibitor discovery by facilitating the identification of chemically diverse inhibitors, offering a promising solution to the structural bias problem in kinase-targeted drug discovery.

Keywords: AlphaFold2; Ensemble screening; Kinase; Multi-state modeling; Protein-ligand docking; Structure-based virtual screening.

Fig. 1

The structure of kinase domain and key structural elements. The active state (DFGin-BLAminus) of BRAF (PDB ID: 6UAN, A) and the inactive state (DFGout-BBAminus) of BRAF (PDB ID: 8C7X, B). The active site is focused in the red box on the right of each panel. The activation loop and DFG motif changes conformation depending on the state.

Fig. 2

Distribution of crystal structures and standard AF2 structures for each kinase conformation. The X-axis is conformational states annotated by KinCoRe and the y-axis is the percentage of each state. All human crystal kinase structures deposited in RCSB-PDB, DUD-E target protein crystal structures, predicted AF2 models with default parameters (standard AF2), and AF3 are colored blue, orange, green, and red, respectively.

Fig. 3

Workflow of multi-state modeling of kinases.

Fig. 4

Predicted cognate docking structures of AKT2. (A) Superimposed structures of crystal structure (PDB ID: 3D0E, grey), MSM model (DFGin-BLAminus, magenta; TM-Score: 0.98), and standard AF2 (cyan; TM-Score: 0.96). (B) Predicted docking poses of the cognate ligand on the AF2 with MSM. The RMSD of the predicted pose is 0.86 Å. (C) Predicted docking poses of the cognate ligand on the standard AF2. The RMSD of the predicted pose is 3.24 Å.

Fig. 5

Predicted binding poses of ChEMBL245377 and CSF1R structures. (A) The superimposed complex structures of ChEMBL245377 and CSF1R. The crystal structure of CSF1R (PDB ID:3KRJ), standard AF2 predicted structure, and MSM model (DFGout-BBAminus, with TT sets) are represented as ribbon diagrams colored as gold, pink, and orange, respectively, while the predicted docking poses of the compound are represented as sticks. (B, C). Focused binding sites and docking poses of ChEMBL245377. The compound is ranked 10^th by ensemble docking (B) and 167^th in docking for the standard AF2 predicted structure (C).