Abstract Details

Poster 41: VHELIBS, a Validation Helper for Ligands and Binding Sites

Adrià Cereto Massagué1
1Nutrigenomics Research Group Biochemistry and Biotechnology Department Building N4, Campus Sescelades Universitat Rovira i Virgili Tarragona, Catalonia
Many users of structures from the Protein Data Bank assume the quality of its deposited structure models, but few models perfectly fit the experimental data, and there have been cases of deposited structures which had to be retracted due to their poor correlation to the experimental data. In order to avoid basing a research on a flawed model, a tool for easy model validation is needed.

Due to the interpretation step during modelling, it is very important to see if a model fits reasonably to its Electron Density Map, because otherwise it would mean the model does not wholly explain the EDM, thus making it unreliable. For drug design, the model quality of the binding sites and ligands of a protein are of particular interest, while the overall model quality or the quality of part of the model outside the binding site are not very relevant. A good way to assess how well a subset of model coordinates fits the experimental data is the Real Space R-value (RSR) parameter. The use of this parameter has been recommended by the X-ray Validation Task Force (VTF) of the Worldwide PDB. The RSR measures the similarity between the EDM built with experimental data and an EDM built directly from the model. The use of the EDM to validate the model will not catch all possible problems in the model, but they can show whether the model fits the data it was created from.
The Validation HElper for LIgands and Binding Sites (VHELIBS) is a software with a graphical user interface that aims to ease the validation of binding site and ligand coordinates for non-crystallographers (i.e., users with little or no crystallography knowledge) by checking how their coordinates fit to their corresponding electron density map. Thus, VHELIBS automatically labels as good those residues and ligands based on threshold values for several properties specified by the user, and enables the user to visually check the fitness quality of dubious residues/ligands to their corresponding electron density map. The values which can be used for filtering are: RSR (2 different thresholds), RSCC, occupancy-weighted B-factor, R-free, resolution and residual average occupancy. The user can also use PDB_REDO models instead of PDB models, which should provide better results regarding electron density fitting. Two default profiles are provided, one for use with PDB models and one for PDB_REDO models, but the user can adjust the threshold values to match his or her needs. Another interesting feature of VHELIBS is that the user does not need to know the list of PDB codes for his or her protein, as it supports UniProt names and access numbers as input, which are translated into their corresponding PDB codes and will ultimately yield which are the most reliable models for the binding sites and ligands of that protein.

Using VHELIBS it is a lot easier to see if the binding site and the ligand coordinates are reliable and, therefore, if they can be used for drug discovery purposes (for instance during structure-based pharmacophore development or as the target during a protein-ligand docking).

Return to Programme