Optimal strategies for virtual screening of induced-fit and flexible target in the 2015 D3R Grand Challenge

Abstract

Induced fit or protein flexibility can make a given structure less useful for docking and/or scoring. The 2015 Drug Design Data Resource (D3R) Grand Challenge provided a unique opportunity to prospectively test optimal strategies for virtual screening in these type of targets: heat shock protein 90 (HSP90), a protein with multiple ligand-induced binding modes; and, mitogen-activated protein kinase kinase kinase kinase 4 (MAP4K4), a kinase with a large flexible pocket. Using previously known co-crystal structures, we tested predictions from methods that keep the receptor structure fixed and used (a) multiple receptor/ligand co-crystals as binding templates for minimization or docking (“close”), (b) methods that align or dock to a single receptor (“cross”), and (c) a hybrid approach that chose from multiple bound ligands as initial templates for minimization to a single receptor (“min-cross”). Pose prediction using our “close” models resulted in average ligand RMSDs of 0.32 Å and 1.6 Å for HSP90 and MAP4K4, respectively, the most accurate models of the community-wide challenge. On the other hand, affinity ranking using our “cross” methods performed well overall despite the fact that a fixed receptor cannot model ligand-induced structural changes,. In addition, “close” methods that leverage the co-crystals of the different binding modes of HSP90 also predicted the best affinity ranking. Our studies suggest that analysis of changes on the receptor structure upon ligand binding can help select an optimal virtual screening strategy.

Predicting protein targets for drug-like compounds using transcriptomics

“…Unlike most available ligand-based prediction methods [12][13][14][15][16][17], the accuracy of our approach does not rely on chemical similarity between compounds in the training/test sets. For instance, our screen against CHIP, a target with no known small molecule inhibitors, delivered four out of six binding compounds, whereas a parallel analysis using a state-of-the-art structure-based virtual screening [38,60] yielded only two weak-binding compounds. Moreover, the predicted mechanisms of actions of the more potent LINCS compounds suggest novel interactions that were not prioritized by the ligand-based screen ( Figure S5).…”

“…For instance, our screen against CHIP, a target with no known small molecule inhibitors, delivered four out six binding compounds, whereas a parallel analysis using solely a state-of-the-art structure-based virtual screening (Koes and Camacho, 2012;Ye et al, 2016) yielded only two weak-binding compounds. Moreover, the predicted mode of actions of the LINCS compounds suggest novel chemotypes that would have been difficult to prioritize in a ligand-based screen ( Figure S10).…”

“…We selected one or more representative crystal structures for each gene, optimizing for sequence coverage and structural resolution. We then docked hits to their top 100 potential targets and ranked using a prospectively validated pipeline Koes et al, 2013;Koes et al, 2015;Ye et al, 2016).…”

“…We next visually examined the docking models of top ranking/scoring hits to select those that show suitable mechanisms of action, and purchased six compounds for testing (Table S3). In parallel, we performed a pharmacophore-based virtual screen of the ZINC database (Irwin and Shoichet, 2005) using the ZincPharmer (Koes et al, 2015) server, followed by the same structural optimization Koes et al, 2013;Koes et al, 2015;Ye et al, 2016) performed on the LINCS compounds. We purchased seven of the resulting ZINC compounds for parallel testing.…”

“…potential targets and ranked using a prospectively validated pipeline[38][39][40][41].On average, crystal structures were available for 69 out of the top 100 potential targets for each compound, and structures of known targets were available for 53 of the 63 hits. In order to avoid redocking into cocrystals of our hits, we made sure to exclude from our analysis all crystal structures containing these 53 ligands, ensuring that our results would not depend on prior knowledge of interaction partners or binding modes.…”