Enhancing Virtual Screening with Experimental Electron Density Data

Virtual screening has become a vital tool in drug discovery, allowing researchers to sift through vast compound libraries to identify potential drug candidates. Molecular docking, the cornerstone of virtual screening, involves predicting the binding affinity of small molecules to a target protein’s binding site. However, the accuracy of molecular docking remains limited due to simplifications in computational models and the neglect of solvent effects, which influence ligand binding.

In recent years, researchers have sought innovative ways to improve the efficacy of virtual screening. This article explores the novel approach of leveraging experimental electron density (ED) data to enhance active compound enrichment during virtual screening. Experimental ED maps, obtained through X-ray crystallography, offer valuable insights into the time-averaged behavior of ligands and solvents within binding pockets, potentially addressing some of the limitations of traditional molecular docking.

The Promise of Experimental Electron Density Data

Traditional virtual screening often relies on static protein structures and simplified solvent models. To overcome these limitations, researchers are turning to experimental ED maps, which provide a dynamic view of the molecular interactions within binding pockets. These maps offer information about ligand-solvent interactions, complementing the static models used in virtual screening.

ED maps can be particularly informative for noncovalent interactions (NCIs) identification, molecule generation, and parameter refinement in quantum mechanics studies. By incorporating ED data, researchers can gain a more comprehensive understanding of the dynamics within the binding pocket, potentially improving the accuracy of virtual screening.

Introducing ExptGMS: Experimental ED-based Grid Matching Score

To harness the power of experimental ED data for virtual screening, researchers have developed a novel method known as ExptGMS, or Experimental ED-based Grid Matching Score. ExptGMS combines two key components: an experimental ED-based grid and a scoring function.

In this approach, 2Fo–Fc ED maps with above-zero contour levels are used for grid generation, ensuring that only relevant ED data is considered. Grid points are placed within and around the binding pocket, and each point is assigned a value reflecting the ED intensity at that position. The scoring function measures the degree of matching between a ligand’s conformation and the ED grid.

ExptGMS incorporates three principles into its scoring function:

(1) Rewarding ligand atoms occupying grid points with strong ED intensity.

(2) Penalizing ligand atoms in areas without grid points.

(3) Penalizing grid points with strong ED intensity not occupied by ligand atoms.

The construction of ExptGMS grids is resolution-dependent, allowing researchers to balance detail and generality when scoring ligand binding conformations.

Enhancing Active Compound Enrichment with ExptGMS

To assess the performance of ExptGMS, researchers conducted extensive testing using the Directory of Useful Decoys-Enhanced (DUD-E) dataset, which includes 85 protein targets and their corresponding compounds. ExptGMS was evaluated alongside traditional benchmark methods, including molecular docking and molecular similarity comparisons.

Results showed that ExptGMS significantly improved the active compound enrichment in the top-ranked molecules when compared to traditional methods. Notably, ExptGMS outperformed benchmark approaches in terms of both enrichment and diversity of the selected compounds. This suggests that ExptGMS can be a valuable addition to virtual screening workflows.

Multi-Resolution ExptGMS: A Versatile Approach

Researchers also explored the potential of multi-resolution ExptGMS, which allows the scoring of ligand binding conformations at different levels of detail. Varying resolutions in ExptGMS grids mimic the varying levels of detail present in experimental ED maps.

Multi-resolution ExptGMS demonstrated the potential to complement different docking scenarios. Lower-resolution grids provided scaffold-level information, while higher-resolution grids offered atomic-level insights. The combination of multiple resolutions in ExptGMS showed promise in enriching active compounds, suggesting that this approach can enhance the performance of virtual screening across a wide range of targets.

Machine Learning Integration for Improved Performance

To further enhance the capabilities of ExptGMS, researchers developed a machine-learning model based on Gradient Boosting Decision Tree (GBDT). This model integrated ExptGMS features and demonstrated improved active compound enrichment in virtual screening.

The GBDT model, trained on a subset of the DUD-E dataset, highlighted the effectiveness of combining ExptGMS with traditional molecular docking scores. This hybrid approach showed substantial improvements in the identification of active compounds, indicating the potential of machine learning to optimize ExptGMS-based virtual screening.

Real-World Application: Covid-19 3CLpro Inhibitor Screening

To validate the real-world applicability of ExptGMS, researchers conducted virtual screening for COVID-19 3CLpro inhibitors. This study involved the screening of a vast compound library using ExptGMS in combination with molecular docking.

The results demonstrated that ExptGMS significantly enhanced the identification of active compounds compared to traditional molecular docking alone. Several promising inhibitors were identified, including one with an impressive IC50 value of 1.9 µM. This highlights the potential of ExptGMS to contribute to drug discovery efforts, particularly in urgent situations like the Covid-19 pandemic.

Building an ExptGMS Database and Online Service

Recognizing the value of ExptGMS for the scientific community, researchers developed an online database containing multi-resolution ExptGMS grids for over 17,000 proteins. This database aims to facilitate the use of ExptGMS by academic users and offers a user-friendly web-based service for scoring ligand binding conformations.

Discussion: Future Directions and Challenges

While ExptGMS shows promise in enhancing virtual screening, there are still opportunities for improvement and challenges to overcome. Researchers suggest investigating the use of multiple crystal structures for targets and exploring the incorporation of additional information, such as noncovalent interaction data, to further refine ExptGMS.

Additionally, addressing limitations, such as the dependence on accurate binding poses and the availability of experimental ED data, remains a priority. Future research may focus on developing high-speed binding-pose-search engines and finding alternative approaches for situations where experimental ED data is lacking.

In conclusion, ExptGMS represents a groundbreaking approach in virtual screening, leveraging experimental electron density data to improve active compound enrichment. This method has demonstrated its efficacy across various targets, making it a valuable addition to the drug discovery toolkit. As researchers continue to refine and expand the application of ExptGMS, it holds the potential to significantly impact the field of drug discovery.

Study DOI: doi.org/10.1038/s42004-023-00984-5

Engr. Dex Marco Tiu Guibelondo, B.Sc. Pharm, R.Ph., B.Sc. CpE

Editor-in-Chief, PharmaFEATURES