ES-Screen: A Novel Electrostatics-Driven Method for Drug Discovery Virtual Screening

Abstract

Electrostatic interactions drive biomolecular interactions and associations. Computational modeling of electrostatics in biomolecular systems, such as protein-ligand, protein-protein, and protein-DNA, has provided atomistic insights into the binding process. In drug discovery, finding biologically plausible ligand-protein target interactions is challenging as current virtual screening and adjuvant techniques such as docking methods do not provide optimal treatment of electrostatic interactions. This study describes a novel electrostatics-driven virtual screening method called 'ES-Screen' that performs well across diverse protein target systems. ES-Screen provides a unique treatment of electrostatic interaction energies independent of total electrostatic free energy, typically employed by current software. Importantly, ES-Screen uses initial ligand pose input obtained from a receptor-based pharmacophore, thus independent of molecular docking. ES-Screen integrates individual polar and nonpolar replacement energies, which are the energy costs of replacing the cognate ligand for a target with a query ligand from the screening. This uniquely optimizes thermodynamic stability in electrostatic and nonpolar interactions relative to an experimentally determined stable binding state. ES-Screen also integrates chemometrics through shape and other physicochemical properties to prioritize query ligands with the greatest physicochemical similarities to the cognate ligand. The applicability of ES-Screen is demonstrated with in vitro experiments by identifying novel targets for many drugs. The present version includes a combination of many other descriptor components that, in a future version, will be purely based on electrostatics. Therefore, ES-Screen is a first-in-class unique electrostatics-driven virtual screening method with a unique implementation of replacement electrostatic interaction energies with broad applicability in drug discovery.

Keywords: drug discovery; electrostatic energy; electrostatic potential; electrostatics; free energy; hit-to-lead identification; virtual screening.

Figure 1

(A) Schematic of ES-Screen for predicting ligand-target interactions. (B) Process of determining the electrostatic energy of interaction. Using predicted ligand-protein binding poses after pharmacophore screening: (1) ligand-free electrostatic potentials (ESP, Φ) given by the protein electric field (E) are calculated using Delphi at positions occupied by ligand atoms, (2) ligand atoms, represented as partial charges, are then assembled within the pocket from infinity within solvent of high dielectric constant to ligand-occupied positions within the solute of low dielectric constant, and (3) the electrostatic energy of interaction is calculated using Equation (1).

Figure 2

Areas under the curve for the DUD-E dataset determined from ROC curves in Figure 1 include MM-GBSA/PBSA methods.

Figure 3

Virtual screening performance using DUD-E benchmarks. (A,B) Receiver operating characteristic (ROC) curves for (A) GLIDE-SP docking and (B) ES-Screen. (C,D) Relative enrichment factors of (C) the top 1% and (D) the top 5% for GLIDE-SP docking, MM-PBSA, MM-GBSA, and ES-Screen.

Figure 4

Improvement of virtual screening performance with a combination of parameters and weight adjustment in building ES-Screen. Average performance across N = 53 protein targets using relative enrichment factors EF1% and EF5% and areas under the curve (AUC) derived from ROC curves. Error bars correspond to standard errors of the mean.

Figure 5

ES-Screen performance compared to GLIDE-SP for virtual screenings using a chemical dataset of FDA-approved and experimental drugs. (A) Heatmap of relative enrichment factors (EF1% and EF5%) for individual protein targets with respect to each virtual screening method; colored boxes for EF1% improvement and EF5% improvement represent whether ES-Screen performs superiorly (green box) or inferiorly (red box) with respect to each method; (B) Performance measured using the mean area under the curve (AUC), obtained from ROC curves, within each protein family, error bars represent standard errors of the mean (abbreviations: GPCR, G protein-coupled receptor; NR, nuclear receptor; OxRed, oxidoreductase; p-value < 0.05 on one-tailed t-test with respect to ES-Screen). (C) Log-transformed p-values of one-tailed t-tests comparing the ES-Screen to GLIDE docking, red dashed line represents statistical significance level at p < 0.05. (D) Mean relative enrichment factors for the top 1% and 5% of the database screened across all protein targets (N = 53) are depicted; significance at p < 0.01 and p < 0.05 reported using a paired, one-tailed student’s t-test with respect to ES-Screen (black bar).

Figure 6

(A) Ratios of the areas under the curves (AUCs) obtained from ROC curves between ES-Screen and the indicated methods with respect to each protein target. Green arrows correspond to ES-Screen providing a >1.5-fold improvement, grey arrows correspond to ES-Screen providing <1.5-fold improvement, and red circles indicate instances where ES-Screen performed inferiorly to the respective method. (B) Bar graph of the mean AUCs across all protein targets with respect to each virtual screening method.

Figure 7

Trends in ES-Screen performance with respect to reference ligand and protein target physicochemical properties. Scatter plots of relative ES-Screen to GLIDE-SP docking performance for each protein target were studied against the following parameters: (A) reference ligand molecular weight, (B) reference ligand volume, (C) reference ligand solvent-accessible surface area (SASA), (D) reference ligand combined number of potential acceptor/donor hydrogen bonds, (E) protein dipole moment, (F) reference ligand dipole moment, (G) length of protein determined by a number of amino acids, and (H) presence of metal cations within the protein.

Figure 8

Mebendazole, Indomethacin, licofelone, and oxaprozin bind directly, in a dose-dependent manner, to immobilized human adipocyte fatty acid binding protein (FABP4) and aldose reductase (AKR1B1) in Surface Plasmon Resonance (SPR) assay. Assays were performed in triplicates for each drug with respect to each protein target.