- Course overview
- Search within this course
- Introduction to model quality assessment
- Global quality assessment
- Key takeaways
- Crosscheck with a crossword
- Further reading
- Acknowledgements
- Your feedback
All materials are free cultural works licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) license, except where further licensing details are provided.
Visual inspection
While numerical metrics are invaluable, nothing replaces the ability to see the model and its relationship with the experimental data in 3D.
Visual inspection is an indispensable part of local quality assessment. It involves using molecular visualisation software to examine the structure in detail. Common software includes Mol* viewer (available directly on the web browser), PyMOL, ChimeraX, Moorhen, and Coot (often used for model building and validation).
Here’s what to look for during visual inspection:
Examining the Fit to Experimental Data (for X-ray and cryo-EM)
For structures determined by techniques that produce a density map (X-ray crystallography and cryo-EM), the map is your most direct visual evidence of the molecule’s shape and position. The atomic model should fit snugly within this density.
What to check visually in the density:
- Backbone trace: Does the protein or nucleic acid backbone follow continuous density? Are there breaks or places where the backbone seems forced into a position without density support?
- Side chain / Base placement: Do the side chains of amino acids or bases of nucleotides fit neatly into distinct lobes of density? Or do they appear to stick out into empty space, or clash with the surrounding density?
- Ligand fit: Does a bound ligand fit neatly into the density within the binding site? Is there a significant positive or negative difference in density around the ligand, suggesting it is mismodelled, partially occupied, or not present?
- Ligand fit: Does a bound ligand fit neatly into the density within the binding site? Is there a significant positive or negative difference This refers to peaks (positive difference) or troughs (negative difference) in the electron density map, indicating where atoms might be present or absent compared to the current model. in density around the ligand, suggesting it is mismodelled, partially occupied, or not present?
Confident ligand fit in InhA complexed with N-(3-chloro-2-methylphenyl)-1-cyclohexyl-5-oxopyrrolidine-3-carboxamide (PDB ID: 4TZT).
This entry illustrates an example of a well-modelled ligand (468) snugly fitting within its binding site. The blue mesh represents the 2mFo-DFc electron density, which perfectly encapsulates the ligand atoms.
Poor ligand fit: Diclofenac in the binding site of C-lobe complex (PDB ID: 3ib0).
This entry illustrates a situation where a modelled ligand (Diclofenac) is not well-supported by the experimental electron density. This example highlights the importance of visually inspecting difference maps for clear evidence of ligand presence and accurate placement.
- Areas of weak or missing density: Are there regions (like loops or termini) where the density is very weak or completely absent? This indicates that these parts of the molecule are likely flexible or disordered and may not have a single fixed conformation in the crystal or sample. Models built into weak or missing density are less reliable. Researchers may still model these regions, perhaps based on prior knowledge or connecting segments, but their accuracy is lower. You must decide if the model quality in such a region is sufficient for your needs.
Carboxysome shell protein ccmK2 (PDB 2A1B)
This entry shows a C-terminal helix modelled with no electron density support.
Cryo-EM model (PDB ID: 7QFQ)
This entry illustrates a region of a macromolecular model where the atomic coordinates (ball and stick representation of residues 1198-1205) are not clearly supported by the experimental cryo-EM map