Evaluation of the quality of experimental structures

Electron density

RCSB PDBhttps://www.rcsb.org PDBehttp://www.ebi.ac.uk/pdbe

Antibody fragment

  • Download the structure 1FL6 from the PDB and its electron density map from PDBe
    • search 1FL6, click Download files → Map → EDS map
  • Rename 1fl6.ccp4 to 1fl6_map.ccp4
  • Drag and drop 1fl6.cif and 1fl6_map.ccp4 into PyMOL's main window
    • Alternatively, click File → Open
  • Create a mesh called ed1 according to 1fl6_map.ccp4, where lines delimit an electron density of 1.0, only plot within 1.6 Å of the structural model
    • isomesh ed1, 1fl6_map, 1.0, carve=1.6
  • Show the protein as sticks
    • show sticks

Quality of electron density maps and resolution

  • Download the structure 1KEL from the PDB and its electron density map from PDBe
  • Rename 1kel.ccp4 to 1kel_map.ccp4
  • Drag and drop 1kel.cif and 1kel_map.ccp4 into PyMOL’s main window
    • Alternatively, click File → Open
  • Display the sequence
    • set seq_view, on or Display → Sequence
  • define a selection named site1 made of residues 37 to 42 of the L chain of 1KEL. Then, display it as an isosurface mesh as previously.
    • sele site1, resi 37-42 & 1KEL & chain L
    • isomesh ed1, 1kel_map, 1.0, site1, carve=1.6
    • show sticks, site1
  • Look closely at site1. Show sticks and study, again, how the structural model and the electron density map match.
  • Let us look at the same patch of residues in 1FL6:
    • delete ed1
    • sele site1, resi 32-37 & 1FL6 & chain L
    • isomesh ed1, 1fl6_map, 1.0, site1, carve=1.6
    • hide sticks
    • show sticks, site1
  • 1FL6 has a resolution of 2.8 Å, and 1KEL has a resolution of 1.9 Å. Comment on the resolution and quality of both structures in relation with the quality of their electron density maps.

Protein-ligand interaction

  • Download the structure 2CZ8 and its electron density map from the PDB and its electron density map from PDBe
  • select the residue/object called FAD on chain A and call it inh1
    • sele inh1, resn FAD & chain A
  • display the electron density of the ligand and label its atoms by B-factor
    • isomesh ed1, 2cz8_map, 1.0, inh1, carve=1.6
    • inh1 → Label → b-factor
  • Going back to 1FL6:
    • select inh2, resn AAH & chain B
    • isomesh ed2, 1fl6_map, 1.0, inh2, carve=1.6
    • inh2 → Label → b-factor
  • And now back to 1KEL:
    • select inh3, resn AAH & 1KEL
    • isomesh ed3, 1kel_map, 1.0, inh3, carve=1.6
    • inh3 → Label → b-factor
  • B-factors describe internal motions, flexibility, or uncertainty on atom position.
  • Comment on the position of the ligand in all three structures. Which structure seems the most trustworthy when it comes ligand position.

    • More about low electron densities

    • Finally, study the structure and electron density of 1BN6. There are lot of low electron densities. What are these?

    Structure validation

    PDBsumhttp://www.ebi.ac.uk/pdbsum/ MolProbityhttp://molprobity.biochem.duke.edu/ SAVEShttp://services.mbi.ucla.edu/SAVES/

    X-ray crystallography

    • Let us look at 2 structures, both created using the same X-ray crystal diffraction data.
      • download structures 1CHR from here and 2CHR from PDB
      • using PDBsum, study their Ramachandran diagram (in the top right corner)
      • using PDB, look at the quality report of 2CHR in the “wwPDB Validation” subsection
      • using MolProbity and SAVES, compare both structures
        • results are available here (MolProbity data is only saved for 24h):
    • Load both structures into PyMOL. Load the electron density map of 2CHR too.
    • Study closely. Then focus on the first 50 residues. Is there anything wrong? Which structure is correct, and why?
    • Study residue R35, find polar contacts:
      • Action → Find → Polar contacts → To other atoms in object
    • In which structure is R35 in the most favourable conformation?
    • CryoEM studies provide atomic density maps, which can be treated similarly to electron density maps obtained from X-ray crystallography studies.

    NMR

    However, NMR data produces no such map, only structural restraints (distances, angles, secondary structure prediction from chemical shifts). NMR structures are computed as a structure bundle (at least 10 structures), which is then submitted to multiple cycles of temperature jumps. This protocol is called simulated annealing. If the NMR data quality is sufficient, all structures will converge towards a unique structure.

    • Let us look at some ribosomal proteins: 1RIP and 2LLU.
      • in PDBsum, study their Ramachandran diagrams
      • estimate structure quality using MolProbity: 1RIP and 2LLU
      • precalculated results (MolProbity data is only saved for 24h): 1RIP, 2LLU
    • Which structure has the best quality?
    • There is no unified definition of resolution in NMR. Thus, until ~10 years ago, providing a PDB entry with a resolution was unfortunately not standard practice. The simplest method to compute a resolution from an NMR structure bundle is provided below:
      • set all_states, on
      • intra_fit [your structure bundle]
    • PyMOL will return RMSD values for the alignment of every structure against the first structure, which is usually the structure of lowest energy. The average of these RMSD values is an estimation of the resolution.
    • Some will prefer computing the resolution as an average of RMSDs against an average structure. Other will prefer computing it as an average of pairwise RMSDs. All these methods usually give values in the same range.
    Previous chapter
    Next chapter