General model building workflow

Resources for Model Building/Refinement

(Courtesy of Nate Burman)

Note that this information will be specific to Cryo-EM maps.

General Strategy

• Decide on a map
• Decide on a model
• Dock the model into the map
  • If there are multiple subunits, you will need to manually create a composite PDB where each chain has its own unique identifier.
• Use Phenix to refine the model into the density
  • Don’t use SS restraints. Only use model restraints when you are sure that you’re close to the correct structure
• Use COOT to inspect the problem areas identified by phenix

Tutorials

• -2021 Phenix Workshop on Cryo-EM Map improvement:
  • Sharpening settings to use
  • Note that if you are using a Mac, you have a lot more bells and whistles. Could be worth getting this installed on a lab computer
• Phenix Workshop on Real Space Refinement for Cryo-EM.
  • As of 2023-10-30: This is recent and relevant information.

• General List of Phenix Tutorials,
  • Explanations for the various job types and strategies that are available for model refinement in this package.
• Model building in Coot
• Isolde tutorial

Useful Strategies to accelerate COOT and prevent runaway refinements

If you know that you are building into a secondary structural element, make sure that COOT has the proper restrains applied for local refinement

If you are building into an area with particularly weak density, you will want to lower the refinement weighting. This applies less weight to the observed density and more weight towards making a favorable structure.

Use the Refine/Regularize control to access the local refinement parameters.

Typically, it is good to enable torsion and rama restraints as a general rule for local refinement.

If the density is particularly weak in the area that you are working on, lower the refinement weighting. This will need to be experimented with manually on a per-map/per-region basis.

Order of operations for model building

1. Alphafold structure for each subunit of the complex
2. Dock the alphafold structures to the density
3. Make a composite PDB where each Alphafold structure is a unique chain
4. Refine the composite PDB into the density map. Enable reference model restraints for the initial pass.
5. From these refinement results, identify unstructured regions that cannot be modelled.
6. Trim each chain accordingly
7. Realspace refine without reference model restraints or secondary structure restraints
8. Identify areas that still have low CC scores, rebuild regions where density is present
9. Realspace refine
10. Identify Molprobity outliers
11. Cis peptides -> Bond distance violations -> bond angle violations -> Rama outliers -> Overlapping atom pairs