Key takeaways

You have now explored the key concepts, tools, and metrics used to assess the quality of experimental macromolecular structures deposited in the PDB. Remember that every model is a representation of the experimental data and the way it was processed, and its reliability can vary. Your ability to critically evaluate these structures is vital for the success and validity of your research.

To confidently use a structural model, you need to synthesise information from several sources. Here are the key takeaways and guiding questions to help you “put it all together”.

Start with a global overview; get a general sense of the structure’s overall quality using global metrics.

• Check the resolution.
• Look at the Validation Slider on the PDB entry page for a quick visual summary and comparison to other structures. Are the key metrics (R-free, Clashscore, outlier percentages, RSRZ/Residue Inclusion) in the blue (better than average) or red (worse than average) regions?
• Consult the summary page of the PDB Validation Report for specific values and percentile ranks for metrics like R-free, clashscore, Ramachandran and sidechain outliers, RSRZ outliers, assignment completeness, restraint violations or Q-score.
• If your structure has ligands, quickly check the overall fit using metrics like Ligand RSCC and RSR.

Perform local assessment on regions of interest; global metrics can mask local issues. You must evaluate the quality of the specific parts of the structure relevant to your research.

• Use visual inspection with molecular viewers (like PyMol, ChimeraX or YASARA) to examine the model in 3D.
• Crucially, inspect the model’s fit to the experimental data. Does the model follow the density map? Are there signs of missing atoms or density where the model shouldn’t be?
• Check the local geometry of residues or ligands in your region of interest. Are there Ramachandran outliers, sidechain outliers, or unusual bond lengths/angles listed in the Validation Report for these specific residues? Is this unusual geometry supported by clear experimental data?
• Identify and assess the significance of atomic clashes in your region of interest using the Validation Report or molecular viewers.
• If studying a ligand, carefully assess its fit to density (RSCC/RSR values) and its stereochemistry.
• Is this region modelled despite weak or missing experimental density? If so, be aware of the lower confidence in its accuracy.
• For NMR structures, check for restraint violations in this region. These violations indicate that the model is inconsistent with the experimental data and suggest potential inaccuracies in the local structure.

Consider other important factors; think about factors that influence the model’s completeness, flexibility, and relevance.

• Are there missing regions in the model, especially in areas important for your research? Remember, these are often disordered or flexible regions not visible in the experimental data.
• Does the structure represent the correct conformational state of the molecule for your study (e.g., bound vs. unbound)?
• Be aware of how hydrogen atoms are handled by different experimental methods, as this can affect the interpretation of interactions and side chain identity.
• If dealing with very large proteins, remember that they might be assembled from fragments or domains solved separately.

The scientific paper provides essential context about the experiment, data, modelling choices, and the authors’ interpretation of the structure. This is crucial for understanding the strengths and limitations of the model in its biological context.

By systematically evaluating structures using this multi-faceted approach, starting global, checking locally, considering other factors, and consulting the original literature, you can gain confidence in the models you use and ensure their appropriateness for your research questions.

Remember that validation is an ongoing process. The wwPDB provides comprehensive Validation Report user guides on their website if you wish to delve deeper into specific metrics or sections of the report.

Sometimes, the quality of a structure can be improved through re-refinement using automated procedures like those available through PDB-REDO. This service re-refines structures deposited in the PDB using updated software and protocols and can sometimes result in models with improved validation statistics, especially for older entries or those determined at lower resolution. Comparing the original PDB entry with a corresponding PDB-REDO entry (if available) can be another way to assess potential model improvements.