Structure-Based Virtual Screening Using Glide

Tutorial Created with Software Release: 2026-1
Topics: Hit Discovery, Small Molecule Drug Discovery, Virtual Screening
Products Used: Glide

Tutorial files

10.1 MB
This tutorial is written for use with a 3-button mouse with a scroll wheel.
Words found in the Glossary of Terms are shown like this: Workspacethe 3D display area in the center of the main window, where molecular structures are displayed

 

Tip: You can hover over a glossary term to display its definition. You can click on an image to expand it in the page.
Abstract:

 

This tutorial demonstrates how to perform a structure-based virtual screen for potential inhibitors of FXa using the ligand docking application Glide. You will learn how to generate a protein receptor grid, validate the docking protocol by docking the cocrystal/cognate ligand, dock a set of ligands into the receptor grid, and analyze the docking results.

This tutorial does not cover strategies to address binding site flexibility, quantitatively predict ligand binding affinities, covalent docking, or accurate pose prediction for running FEP+ calculations.

Tutorial Content

  1. Introduction to Docking for Structure-Based Virtual Screens

  1. Prerequisites for Running a Virtual Screen

  1. Creating Projects and Importing Structures

  1. Setting up a Receptor Grid with with H-Bond Constraint

  1. Preparing the Ligands

  1. Docking the Screening Compounds with Constraints

  1. Analyzing the Results

  1. Conclusions and References

  1. Glossary of Terms

1. Introduction to Docking for Structure-Based Virtual Screens

Figure 1. Schematic representation of a virtual screening funnel.

In a virtual screen, the goal is to sift through large libraries of virtual candidate molecules in order to find those compounds which are likely to bind to a given receptor. A full virtual screening funnel is a multi-stage process, where many tools are used consecutively to filter out molecules which are unlikely to bind and therefore ‘enrich’ the remaining list with probable binders. The stages are sorted such that the calculations which incur larger computational costs per compound happen later, when the list has already been reduced by easier-to-calculate methods. For a comprehensive overview of the available tools and best practices for constructing a virtual screening cascade, see the Virtual Screening learning path.

If a structure of the target is available and the relevant binding pocket is known, a docking tool such as Glide is frequently used as part of a structure-based virtual screening funnel as a powerful discriminator between compounds likely to bind to the receptor and non-binders. In a nutshell, computational docking describes the process of attempting to fit together the shapes of the receptor and a ligand as well as possible and assigning a score to the resulting ligand pose which quantifies how well the ligand fits in the pocket. For a detailed description of how Glide finds ligand poses and scores them, consult the Glide User Manual.

In order to make docking calculations feasible in a virtual screen with hundreds of thousands of compounds, the receptor is kept rigid, and only the ligand’s geometry is adjusted to find a good pose. As receptor flexibility and dynamics are essential for a correct description of the interactions between ligand and receptor, the scoring function for Glide is not optimized for approximating binding affinities. Instead, Glide scores are very well-suited for distinguishing between binders and non-binders in libraries with diverse chemical matter, such as encountered in a virtual screen.

In this tutorial, you will set up a docking model for use in a virtual screen in order to find inhibitors of the human Factor Xa system, an established drug target for anticoagulants. We will validate the docking model using the cognate ligand and a set of known binders, and use the options provided by Glide to impose constraints on the ligand pose and analyze how these constraints impact the results of our screen.

2. Prerequisites for Running a Virtual Screen

In order to perform a structure-based virtual screen using Glide, you will need to have a structure of your target. As Glide will consider the receptor as rigid, we recommend that, if available, you use a ligand-bound structure of your target. For an investigation into the effects of using apo structures or structures predicted by ML models such as AlphaFold in virtual screens, please consult this publication. If you need to account for receptor flexibility (e.g. if you want to dock ligands significantly larger than the cognate liganda ligand that is bound to its protein target), consult the Understanding and Visualizing Target Flexibility or Approximating Protein Flexibility without Molecular Dynamics tutorials or consider using induced-fit docking methods such as IFD or IFD-MD.

Structure files obtained from the PDB, vendors, and other sources often lack necessary information for performing modeling-related tasks. Typically, these files are missing hydrogens, partial charges, side chains, and/or whole loop regions. In order to make these structures suitable for modeling tasks, we use the Protein Preparation Workflow to resolve issues. Similarly, ligand files can be sourced from numerous places, such as vendors or databases, often in the form of 1D or 2D structures with unstandardized chemistry. LigPrep can convert ligand files to 3D structures, with the chemistry properly standardized and extrapolated, ready for use in virtual screening.

In this tutorial, the protein, cognate liganda ligand that is bound to its protein target, and virtual screening ligands have already been prepared. However, these preparation steps are a necessary part of a virtual screen and must be done before docking. Please consult the Best Practices for Protein Preparation before starting out with your own structure.

The Introduction to Structure Preparation and Visualization tutorial can guide you through using the Protein Preparation Workflow and LigPrep to prepare the ligands and the receptor used in this tutorial.

3. Creating Projects and Importing Structures

The ligand-receptor complex provided here has been prepared using the Protein Preparation Workflow to fill in parts missing in the X-Ray structure and determine the protonation states of titratable groups. Additionally, all water molecules have been removed as Glide will consider any non-ligand atoms to be part of the rigid receptor.

 

If following along with your own structure, make sure it is fully prepared before proceeding to the grid generation step. The Introduction to Structure Preparation and Visualization tutorial can guide you through the process.

4. Setting up a Receptor Grid with an H-Bond Constraint

In this section, you will generate a receptor grid (identifying the binding site) with an H-bond constraint. In order to make the docking calculation fast and scalable to screen large libraries of compounds, Glide reduces the structure of the receptor to a representation where steric and electrostatic properties are calculated at fixed points in space along regularly spaced grid points. Generation of this receptor grid must be performed prior to running a virtual screen with Glide. The shape and properties of the receptor are represented in a grid by fields that become progressively more discriminating during the docking process.

4.1 Identify the binding site

The grid is calculated only in a given region around the binding pocket, which is usually specified using the cognate liganda ligand that is bound to its protein target from the protein structure. If you are using an apo structure, or are unsure of where the binding site is, you can specify the binding site in the “Site” tab of the Receptor Grid Generation panel by selecting residues or spatial coordinates as described here.

What if I don’t have a ligand in my structure?
If you don’t know where the binding site of your target is, you can use SiteMap to identify putative binding sites and define a receptor grid from SiteMap results. In the Site tab of the receptor grid generation panel, just choose “pick to identify the ligand Entry” and click one of the site points in the Workspacethe 3D display area in the center of the main window, where molecular structures are displayed.

If you do know where the binding site is, but don’t have a ligand in the structure you’re using, you can specify the binding site in the “Site” tab of the Receptor Grid Generation panel by selecting residues or spatial coordinates as described here.

4.2 Set a hydrogen bonding constraint

To add more information about how binders are expected to interact with the pocket to a receptor grid, different kinds of constraints can be specified during the grid generation stage. These constraints allow filtering out results which do not satisfy key interactions known to be required for e.g. potency, stability, or selectivity reasons. Additional information such as known binders or binding modes is essential for deciding which (if any) constraints to use for your screening campaign. For a comprehensive overview of the available constraint options, see the Glide User Manual.

According to Adler et al., the salt bridge formed between the inhibitor and ASP 189 observed in the crystal structure of the complex contributes to the potency of the ligand. For this tutorial, you will set the constraint for this specific hydrogen bond in the receptor grid. Please see the Introduction to Structure Preparation and Visualization tutorial for instructions on how to add residue labels and show H-bonds if you want to have a detailed look at the interactions in the binding pocket.

In this section, setting a hydrogen bond constraint in the receptor grid is shown, as a strong interaction to ASP 189 in the Factor Xa pocket is known to contribute to the potency of the ligand.

5. Prepare the Ligands

While we currently have the X-ray pose of the ligand in our Workspace, LigPrep will strip all the 3D pose information and the hydrogen atoms from the ligand and it will be prepared exactly as the ligands in our screening library have been. The cognate ligand should be prepared in the same way as the ligands in the screening library will be. As the screening library used in this tutorial has already been prepared, the instructions below show the process for the cognate ligand.

LigPrep outputs multiple structures for the cognate liganda ligand that is bound to its protein target. These differ in protonation and tautomeric states and cover the states which are reasonable in the pH range we specified. You can align the ligands in 3D or use the 2D viewer to compare the differences between them. You will dock them all in the next section.

6. Set up and run the docking calculation

In this section, you will re-dock the cognate liganda ligand that is bound to its protein target to validate the docking protocol including your choice of constraints. The information gained from this step can help with evaluating poses and favorable ligand-protein interactions, which is useful for hit finding. The validated protocol can then be used for virtual screening of a ligand library.

6.1 Dock the cognate ligand

In order to understand whether this re-docking result is good enough to proceed with the virtual screen, a good understanding of the target pocket is essential. There are no general heuristics in terms of e.g. pose RMSD that distinguish a good re-dock from a bad one. If additional known binders are available, they can be used to substantiate the validation, as shown in the next section.

 

In our case, the parts of the best-scoring ligand pose which are deep in the binding pocket, including the H-bond at ASP 189, overlap excellently with the co-crystal ligand pose. However, all of the docked poses show deviations from the co-crystal pose in the outer part of the pocket. This is not surprising given that the binding pocket of 1FJS is solvent exposed and that docking is performed in vacuum. In our case, this is good enough to proceed with docking the small ligand library which includes known binders we can use as additional validation.

 

Feel free to perform the re-docking exercise with the H-Bond constraint to ASP 189 deactivated in the Ligand Docking (Beta) panel’s Biases & Constraints tab and compare the results as an optional step.

6.2 Dock the ligands from the screening library

After having established that the original ligand docks sufficiently well, we can now dock the screening compounds into the receptor grid. The provided library of 50 ligands contains four known active compounds (some with multiple protonation/tautomer states to consider) which we can use to check whether our protocol is able to distinguish between binders and non-binders. For a comprehensive overview of our current best practice recommendations for designing quality ligand libraries, see the Library Design learning path.

Follow the same steps you performed for the cognate ligand in the previous section to dock the screening library.

7. Analyzing the Results

Multiple Glide docking results can be viewed in the Entriesa simplified view of the Project Table that allows you to perform basic operations such as selection and inclusion and be identified by the job name. Docked results will show the receptor in the first row and the docked ligand(s) in the subsequent row(s), where they are ordered by best to worst docking score, or Glide Gscore if Epik state penalties were not applied in LigPrep. The Glide Gscore is broken down by van der Waals electrostatic components and can be seen in the Project Tabledisplays the contents of a project and is also an interface for performing operations on selected entries, viewing properties, and organizing structures and data, using the Property Tree. You can read more about how docking scores/poses are generated here and here and what dependencies they have here and here.

6.1 Visualize the results using Pose Viewer

This visual inspection step can highlight issues in system preparation, receptor grid, docking protocol and scoring function. Look out for known actives which do not dock at all or are very poorly ranked, highly-strained conformations, or individual residues which extremely influence the docking score.

Out of the initial 83 ligands, 27 of them docked. The top-ranked compounds are all known actives, and for each of the known actives, a docked pose is recovered, which is an additional piece in validating that the docking protocol is reasonable.

 

From a first look at the docking scores, you can see that the known actives are scored in the range of -7 to -10, whereas most non-binders have scores between -3 and -7. For this system, a score in the range of approximately -7 or lower therefore corresponds to a compound that is predicted to bind strongly.

 

Note that the docking scores are not parametrized to correlate with binding affinities, and are not suited to rank-ordering very similar or congeneric compounds.

The primary properties output by Glide are the docking score, the glide emodel and the glide gscore. In brief, the docking score is parametrized to distinguish binders from non-binders in diverse libraries for virtual screening applications.

 

Please see Knowledge Base articles 348 for the differences between docking score and GlideScore and 1027 for more information on the differences between GlideScore and Emodel.

 

Note that none of these scores is parametrized to rank-order highly similar compounds or congeneric series. For this application, we strongly recommend using FEP+ as quantifying the binding affinity differences between congeneric series requires a much more accurate description of the dynamics of protein and ligand as well as the solvation of the binding pocket.

8. Conclusion and References

In this tutorial, you completed a workflow for virtual screening using Glide. You generated a receptor grid with a hydrogen bond constraint, which was used in cognate liganda ligand that is bound to its protein target docking as a positive control to set up a virtual screen of test ligands. Then, a series of screening compounds were docked and the results were viewed using the Pose Viewer, with known actives being found as the top hits. The information gained from this virtual screen can be used to pass the identified binders to the next stage in the screening funnel, for example Re-scoring Docked Ligands with MM-GBSA.

For further reading:

 

9. Glossary of Terms

cognate ligand - a ligand that is bound to its protein target

Entries - a simplified view of the Project Table that allows you to perform basic operations such as selection and inclusion

included - the entry is represented in the Workspace, the circle in the In column is blue

incorporated - once a job is finished, output files from the working directory are added to the project and shown in the Entries and Project Table

Project Table - displays the contents of a project and is also an interface for performing operations on selected entries, viewing properties, and organizing structures and data

Scratch Project - a temporary project in which work is not saved, closing a scratch project removes all current work and begins a new scratch project

selected - (1) the atoms are chosen in the Workspace. These atoms are referred to as "the selection" or "the atom selection". Workspace operations are performed on the selected atoms. (2) The entry is chosen in the Entries (and Project Table) and the row for the entry is highlighted. Project operations are performed on all selected entries

Working Directory - the location that files are saved

Workspace - the 3D display area in the center of the main window, where molecular structures are displayed