Restraining a Low-Resolution Structure Using A Homologous Higher-Resolution Model
In this tutorial, we will generate ProSMART external restraints for the deposited 3.5Å structure with PDB code 1ryx using the sequence-identical 2.15Å structure 2d3i, which adopts a different global conformation and was solved in a different space group.
A basic working knowledge of the CCP4i GUI is assumed.
1ryx.mtz - the deposited structure factors (coverted to MTZ format);
2d3i.pdb - the deposited high-resolution model that will be used as a reference structure.
Please click these three links to access the files individually and save them to a suitable location (e.g. your CCP4 working directory), or alternatively click here to get a .tar.gz containing all three files.
Automatic Generation Of External Restraints And Usage In Refinement
Follow these steps to automatically generate external restraints and use them in refinement using the REFMAC5 CCP4i GUI:
Open the CCP4i GUI.
Open the REFMAC5 task interface, which can be found in the Refinement module.
Provide the 1ryx structure factors and coordinates in the "MTZ in" and "PDB in" fields.
In the "Use Prosmart" field, specify "to generate external restraints to reference structures". Note that three additional input options have appeared.
Now provide the reference model 2d3i.pdb in the "Reference PDB in" field.
Essentially, that's all that's required in order to auto-generate ProSMART restraints using an exteral reference structure, and use them in refinement by REFMAC5! Indeed, clicking the "Run" button would not cause an error, and external restraints would be used in refinement.
However, it is always a good idea to really think about what options/parameters are being used when running ProSMART and REFMAC5, as suitable parameters will depend on the particular case.
For example, there is only one chain in both target and reference structures in this scenario. However, in cases where multiple chains are present in either structure, careful thought must be given to which chains are to be used as target and reference structures. This can be achieved using the ProSMART CCP4i GUI or by providing the REFMAC5 CCP4i GUI with suitable ProSMART keywords - more information on this topic can be found in the ProSMART documentation. Whether or not local NCS restraints are used is also of importance.
Certain parameters/options in the REFMAC5 GUI have to be adjusted in order to use external restraints successfully.
If you don't check these values, you will be unlikely to get good results when using external restraints. Now let's go back to the REFMAC5 CCP4i GUI:
Open the "Refinement Parameters" folder.
The two important refinement parameters that we will consider/change are:
Number of cycles - an increased number of cycles is often required when using external restraints in order to allow any conformational changes to happen, still allowing enough cycles for refinement to reasonably converge. Whilst 20-30 cycles may be sufficient in many cases, the example in this tutorial will use 40 cycles.
Set the number of cycles to 40.
Use hydrogen atoms - whether or not REFMAC5 generates hydrogen atoms can have a large impact on the success of refinement when using external restraints. Whether the overall effect is postive or negative can vary from case to case. Using hydrogens can aid refinement by helping to avoid clashes. However, this can have negative effects if local structure is not already in a sensible conformation, and can inhibit external restraints from causing conformational changes where appropriate. In the example in this tutorial, it is best to ignore hydrogen atoms.
Select ``Use hydrogen atoms: ignore even if present in file''
Note also that we will not use jelly-body refinement (at this stage), although whether or not to use jelly-body refinement should always be considered.
Now close the "Refinement Parameters" folder and open the "External Restraints" folder.
Setting each of these four options to sensible values is vital for the successful use of external restraints:
External restraints weight - controls how strong the external restraints will be applied relative to the data and geometry terms during refinement. Leave this field selected and set to 10.0.
Geman-McClure weight - controls robustness to outliers. As this value is raised, more restraints are treated as outliers and down-weighted. This will allow local conformational differences where there are dissimilarities between target and reference structures. Leaving the tick box unselected will let REFMAC5 choose the default value (typically 0.02). Leave this tick box unselected.
Maximum external restraint distance - should generally always be left set at 4.2Å. This is sufficiently large to allow local structure to be restrained at the secondary structure level, whilst sufficiently small so as not to enforce global rigidity, allowing conformational changes between target and reference structures.
Apply to main chain only - in some cases, you may only want the backbone to be affected by external restraints. However, it is almost always better to use external restraints for both main and side chain restraints. In this tutorial example, we shall disable the tick box so that restraints between both main and side chain atoms are generated.
That's it!!! Now it's time to re-refine the structure! Run the job, close the REFMAC5 GUI window, and return to the main CCP4i job list. While we're waiting for ProSMART and REFMAC5 to run, let's quickly consider how we could have done this using the command line instead of the REFMAC5 GUI.
How To Check That External Restraints Are Actually Being Used
Hopefully, by now our REFMAC5 job in the CCP4i GUI should have finished successfully.
It is worth checking the REFMAC5 logfile to ensure that the external restraints were used as expected.
Open the REFMAC5 logfile in CCP4i. You will notice that information regarding the execution of ProSMART is present at the top of the logfile.
Now scroll down through the logfile to where the first refinement cycle begins (which is identified by the text: "CGMAT cycle number = 1").
Just above this position, you should see two tables containing information regarding the external restraints:
--------------------------------------------------------------------------------
External restraints group : 1
External restraints file :input_keywords
Fail if one of the atoms involved in the restraints is missing in the pdb file
Use restraints for all defined atoms
Ignore restraints if abs(dmod-drest) > 50.00000 *sigma
Ignore retraints if input dist > 4.200000
Weight scale sigmas : 10.00000
Weight min sigma : 0.0000000E+00
Weight max sigma : 100.0000
GM parameter : 2.0000000E-02
Number of distances : 19248
Number of angles : 0
Number of torsions : 0
Number of planes : 0
Number of chirals : 0
Number of intervals : 0
--------------------------------------------------------------------------------
---------------------------------------
Standard External All
Bonds: 4754 19248 24002
Angles: 6510 0 6510
Chirals: 770 0 770
Planes: 859 0 859
Torsions: 3176 0 3176
---------------------------------------
In our case, 19248 external bond restraints are being used, which is correct.
Inspection And Interpretation Of Results
Compare the initial and final R factor and R free values. What do you notice? Do you think the final model is better than the initial model?
Whilst R-factors (and other global statistics) are generally considered good indicators of refinement success/quality, when using external restraints it is often more important to inspect the local structure and density than to worry too much about R-factors, at least to begin with.
Consequently, we should now look at model and density in Coot, before and after refinement.
Double-click on the REFMAC5 job in CCP4i, which should open up a qtRView results page.
Scroll down to "Output Files", and click the "Coot" button next to "Structure and electron density". Coot should be lauched, displaying the model and density corresponding to the structure re-refined with external restraints.
From within Coot, select "Fetch PDB & Map using EDS", type "1ryx" into the input box, and click "Get it". This should download and display the coordinates and density map corresponding to the original model (if the EDS method doesn't work, click here to access the MTZ file required to display the map).
Manual inspection is an important part of the process. At your leisure, compare the two models/maps by looking at different regions of local structure.
You can use Coot's Display Manager to switch models/maps on and off. Keep in mind the following questions:
Are there any features that are different?
Has the backbone moved?
Have any side chains changed conformation?
Can you see evidence that the external restraints have had any effect during refinement? If so, do you consider this effect positive or negative?
Now compare the Ramachandran plots of the two models in Coot (select "Validate -> Ramachandran Plot -> [model]").
Which of the models appears to have the most reasonable backbone geometry, overall?
Why do you think this is? Would you expect this? What solid conclusions can you draw?
In this particular case, there are various regions that have undergone noteable changes during refinement with external restraints. We shall now focus on two such regions:
Before we go any further, select "Draw -> Cell & Symmetry", then set "Master Switch: Show Symmetry Atoms" to "Yes".
Look at region around 472-477. Note how the backbone has changed conformation, and that there is a completely different interpretation of the electron density (this region requires manual remodeling).
Note also that density corresponding to a nearby symmetry-related chain has disappeared.
Now look at the active site region that includes the functional residues Tyr92 and Tyr191. The side chain for Tyr92 is not modeled. Note that after re-refinement with external restraints, density for the side chain of Tyr92 has become visible, currently occupied by Lys209. Also, the side chain of Tyr191 has adopted a different conformation.
Visualising and Editting ProSMART Restraints in Coot
In Coot, you should see a "Restraints" tab next to the "Extensions" tab. If not, follow these instructions:
Save the file user-define-restraints.scm (right-click -> "save link as") to a suitable location (e.g. somewhere in your home directory).
In Coot, select "Run Script" from the "Calculate" tab, and browse to find your user-define-restraints.scm file. Run this script.
This should cause a "Restraints" tab to appear next to the "Extensions" tab. If this does not happen, seek advice.
To visualise the ProSMART restraints:
Click the "Restraints" tab, and select the "Read Refmac Extra Restraints..." option.
Browse to the location of the external restraints file generated by ProSMART (this can be found by looking in the qtRView output for the REFMAC5 job) and click "OK".
You should see all of the restraints appear on the model (like a spider’s web).
Red and blue restraints are those that indicate large differences between the interatomic differences in the target and reference models.
Restraints involving individual residues can be deleted by selecting the "Delete an Extra Restraint…" option in the "Restraints" tab.
After editing the list of restraints, this list can be exported using the "Save as REFMAC restraints…" option in the "Restraints" tab. This restraints file can then be used for REFMAC5 refinement by selecting the "Use ProSMART: input existing restraints file" option in the REFMAC5 CCP4i interface, providing the exported restraints file in the "Prosmart restraints in" input field.
Strategy For Subsequent Refinement Iterations
If you want, try using Coot’s real space refinement and model building tools to get the two identified regions (residues 472-477 and 92, 191 and 209) into a reasonable conformation, and save this corrected file for further refinement.
Otherwise, click here to access a pre-prepared modified model. If you were trying to properly refine this structure, you would go through the whole structure in Coot manually addressing all issues before continuing with further refinement, but for the purposes of this tutorial we will continue anyway.
Now that the external restraints have been used to improve the model, we release the external restraints in subsequent refinement iterations. This allows the model to find it's own density, instead of re-inforcing the bias towards the externally-derived information (i.e. in this case 2d3i.pdb).
However, this has the negative consequence that we are no longer regularising refinement using external restraints.
Consequently, we now use jelly-body restraints in place of the external restraints, which are also regularisers, but do not present any further external bias.
For our next round of refinement:
Open a new REFMAC5 job in the CCP4i GUI.
Provide the original 1ryx.mtz structure factor file in the "MTZ in" fields, and provide the new modified model in the "PDB in" field.
Open the "Refinement Parameters" tab. Select "use jelly-body refinement", and set the "sigma" to 0.01. Again, we will "ignore hydrogens even if present". This time, set the number of cycles to 20.
That's it!!! Note that we are not using external restraints, so "Use Prosmart" has been left at "no". Now it's time to re-refine the structure - run the job!
After the job has finished running, re-inspect the regions of interest (residues 472-477, and residues 92 and 191) in the jelly-body refined structure.
You will see that the model sinks nicely into the density, suggesting that our re-modelling and refinement was reasonable (although of course it is not implied that these regions are now completely fixed - more model building/refinement and manual inspection would still be required).
Note that the effect of the external restraints has been maintained, even though the external restraints have been released.
In practice, further rounds of model re-building and refinement would be required.
These would be performed using jelly-body restraints.
It is recommended to use external restraints in early/medium stages of refinement, followed by subsequent rounds of jelly-body refinement (without external restraints).
Ultimately, the crystallographer should decide what strategy makes the most sense in their particular case.
Further to injecting external information obtained from reliable sources (e.g. homologous structures), external restraints can be used to help stabilize a desired conformation whenever the density alone is of too poor quality to maintain it.
In the next tutorial, we will continue with this example case, exploring how we can use ProSMART to analyse local conformational differences between two structures.
This can be useful for identifying which regions have changed during refinement with external restraints
ProSMART can also be used to compare models from different stages of refinement, or to compare homologous models from different sources/crystals.
Disclaimer - it is important to note that we are not criticizing the authors of the original deposited low-resolution model. Rather, due to improvements in macromolecular crystallographic refinement tools in recent years, we are now able to gain further insight that was not routinely possible at the time of original structure solution. Furthermore, we do not imply that the modifications applied to the model during this tutorial are necessarily accurate – this could only be confirmed/re-evaluated after performing extensive model rebuilding and refinement.