Frequently Asked Questions:


What are the input structures?

Our docking procedures calculate the interaction energy/score of two proteins in different orientations. They require separate protein structures - receptor and ligand. These individual structures may be multi-chain proteins. The user interface has input boxes to upload the protein structures for docking. These input structure files should be in standard PDB format, containing ATOM records.


What are the receptor and the ligand?

In protein-protein docking, the terms "receptor" and "ligand" describe the two interacting proteins. "Receptor" is the larger and "ligand" is the smaller protein in a complex.


How does GRAMM free docking work?

In the free docking algorithm, the receptor is in a fixed position, and the ligand is rotated and translated to determine the optimal binding pose. The procedure calculates the docking score, correlated with van der Waals energy of interaction, in each position, and selects the ligand pose with the lowest energy (see details in Proc Natl Acad Sci USA, 1992;89:2195 and Protein Eng, 1996;9:37 ).


Does GRAMM server take into account conformational changes during the docking search?

No, GRAMM uses a rigid body docking approach, which means that it assumes that receptor and ligand are rigid and do not undergo significant conformational changes upon binding. Smaller conformational changes are tolerated by lowering the docking resolution. However, generally, the rigid-body assumption limits the accuracy of the predictions for protein complexes that involve significant conformational changes upon binding.


How does GRAMM template-based docking work?

The template-based docking is performed by the alignment of the full structure of the target proteins (receptor and ligand) to the full or interface-only structure of the templates of binary protein-protein complexes. In the advanced options section, users can choose the full-structure or the interface-only algorithm (the default option is the full-structure). The template-based docking by default uses the library of 12,470 full structure templates and 12,430 interface templates from our DOCKGROUND resource. The success of the template-based approach depends on the availability of suitable templates that can be used to generate a model of the protein-protein complex. If no suitable templates are available, the procedure would not be able to generate an accurate model.


Is it possible to use custom templates for template-based docking instead of the default library?

Yes, users can define their own set of templates. We provide an option for selecting templates from binary protein-protein interactions extracted from the PDB and stored on our server. Upon entering the PDB ID, the interacting chains appear in the user interface for selection as templates. In case of a long list of custom templates, the user can enter comma-separated template IDs in the following format: xxxx_#A_#B where the first four characters are the PDB ID followed by two chain IDs as a binary complex (e.g., chain A & B of model 1 from PDB 12as is listed as “12as_1A_1B”).


How is the docked model stored/saved?

The docked models are stored as a protein-protein complex in PDB format named model_#.pdb (# = rank of the predicted model) with different chain identifiers (chain A is receptor and chain B is ligand). If there is more than one chain in each molecule (receptor or ligand), the unique chain IDs are labeled alphabetically starting with A in the receptor proteins.


What is the “receptor-ligand.res” file in free docking results?

The docking scores and the transformation matrix (rotation angles and translation vectors for generation of the docking poses from the initial coordinates) of each completed job are stored in the "receptor-ligand.res" file. The first column is the rank of the docked model, the second is the docking score/energy, and the next three columns are rotation angles, followed by three columns of the translation vector.


Is it possible to limit the docking search results based on known interacting residues?

Yes, by using docking constraints, GRAMM can improve the accuracy of predicted interaction, particularly for protein-protein complexes with known binding sites or where the general location of the binding interface can be expected. GRAMM server provides an option to filter/re-score the docking poses based on the user-supplied list of interacting residues of one or both proteins. Since the d accuracy of the predictions depends on constraints quality, the users need to provide confidence scores in the 0-10 range.


How can I use GRAMM to detect intermolecular energy funnel?

To detect binding energy funnels, one can use clustering of the docking poses generated at the scan stage based on some criterion of similarity (e.g., RMSD, or MM-score between the docked models). The basic assumption underlying the clustering approach is that the native structure of the complex corresponds to the binding funnel on the intermolecular energy landscape where the low-energy docking poses are clustered. The GRAMM server implements sequential clustering in which the lowest energy pose is designated as the representative structure of the first cluster and higher energy docking poses within the clustering threshold (based on Cα RMSD) are assigned to that cluster. In the current version of the web server, the clustering of docking poses is limited to free docking only. An advanced option is provided to select the total number of docking poses for clustering and the clustering threshold. For each cluster, a representative structure is linked to the download, and the other docking predictions are shown by the geometric centers for visualization on the results page.


Does the presence of hydrogens or any other HETATM in the coordinate file affect the docking results?

Our docking algorithms consider only the heavy atoms in the structure file. The hydrogen atoms and HETATM records are ignored. Thus, these atoms do not affect docking results.


Where is the docking score stored in the results files for free docking?

For free docking, the GRAMM algorithm calculates the energy of the interaction between two proteins in different orientations based on geometric fit. This score is correlated with the van der Waals interaction energy. The score/energy is in arbitrary units and is stored in plain text format in the "receptor-ligand.res" file provided to the user with other result files. The first column is the score-based rank of the docked model and the second is the docking score. This file also contains the transformation matrix (rotation angles and translation coordinates of ligand) for each docking pose.


What is the docking score in template-based docking and where is it stored in result files?

The template-based docking is performed by the alignment of the full structure of the target proteins to the full or interface-only structure of the templates (consisting of 12,470 full structures and 12,430 interface structures of binary protein-protein complexes from our DOCKGROUND resource (https://dockground.compbio.ku.edu). The scoring of the resulting docking models is performed by the combined scoring function (Biophys J. 2018;115:809). The file “TBD_results.txt” contains the docking score along with other scores used in calculating the combined score, such as alignment scores, contact potentials, etc.


What does “ERROR:[molecules] section not found in receptor-ligand.res” mean?

The submitted job failed due to the absence of ATOM coordinates. The GRAMM docking software is optimized and benchmarked for protein-protein interactions, not protein-ligand interactions. It does not generate docking prediction if the receptor or the ligand does not contain any amino acids.


How to cite the GRAMM webserver?

Singh, A., Copeland, M.M., Kundrotas, P.J., Vakser, I.A., 2024, GRAMM Web Server for Protein Docking, Methods in Mol. Biol., 2714:101-112.