Back to the tutorial main page

ProSMART Tutorial - Part 2b

Introduction To ProSMART Structural Analysis With CCP4mg

Contents


Loading ProSMART Results Into CCP4mg

Results from ProSMART can be visualised using either CCP4mg or PyMOL. In this tutorial, we will use CCP4mg.

First, we must load the ProSMART results into CCP4mg. We can either load a pre-existing results directory into CCP4mg (e.g. as generated using CCP4i, or using the command line), or alternatively run ProSMART using all default settings from within CCP4mg. Running ProSMART from within CCP4mg is very convenient, but using the CCP4i GUI (or command line) is required for more functionality.

It is vitally important to remember to load the input models into CCP4mg before attempting to load the ProSMART results. Follow these steps:

Now that the ProSMART results have been loaded into CCP4mg, you will see a table detailing the residue-based structural conservation scores.

Various scores are provided pertaining to local backbone conservation ("Min", "Central", and "Rotate"), side chain conformation ("Side RMS", "Side AV", and "Max Dist"), and any identified rigid substructures ("Cluster1", "Cluster2", and "Cluster3", in this case).

Get familiar with the CCP4mg ProSMART analysis tool:

Above the table, you will see a box/tab specifying the chain-pair that these results correspond to ("1ryx/A 2d3i/A"). Had you ran ProSMART using multiple structures, or structures comprising multiple chains, then you would see multiple chain-pairs here.

Right of the table, you will see three buttons:

  • Show coloured by - allows colouring the chain-pair according to structural conservation scores.
  • Edit colours for - allows changing the colours and colour gradients.
  • Superpose - superposes the two chains, either using the global alignment, or one of the identified substructures/domains (if any - in this case, there are three "Clusters"). The global superposition will use all aligned residues, whilst the substructure/domain superpositions will produce a tight superposition in that particular region. Note that "Show coloured by" contains a colouring option corresponding to each of the substructures.

Note that you can sort the table by clicking on the column titles.

Visual Analysis Of Global Conformational Changes

We will now compare the target (1ryx) and reference (2d3i) structures, looking at domain motion in order to gain intuition regarding differences in their global conformations:

Now we can see something sensible. The rigid substructure is coloured yellow. Residues that closely 'belong' to the rigid substructure are coloured yellow, those that are more distant (in terms of relative coordinate frame) are coloured red. Residues are coloured on a gradient between yellow and red - you can see that the domain at the top is coloured orange, indicating that this domain has not rotated relative to the yellow substructure as much as the domain at the bottom that is coloured red.

We can see that the core of the structure seems relatively well-conserved between the two homologous chains, although it is hard to tell how well-conserved the other two domains are...

Now let's look at the second substructure:

We can now see that this domain seems relatively well-conserved, despite the large domain motion.

Now let's look at the third substructure:

Again this domain seems relatively well-conserved. However, "relatively well-conserved" isn't very scientific... now let's look at a more detailed analysis of structural conservation at the local level.

Local Structural Conservation Of The Backbone

At this stage, we will want to view the structures in the coordinate frame of the structural core - the first cluster:

We will now colour the structures according to the "Min" score, which is a conformation-independent measure of local backbone conservation. Note that this measure is independent of the global coordinate frame; it is not dependent on how you superpose the structures.

Residues that have a structurally-similar local environment are coloured yellow, gradually changing to red indicating comparative structural dissimilarity.

In order to get a deeper intuition regarding backbone conservation, it is useful to alter the colour gradient. This can be achieved dynamically in CCP4mg:

Get familiar with the CCP4mg ProSMART analysis colour editor:

From within the colour selection window, you can change the colours (if you don't like yellow and red!), the method of colour interpolation, and importantly alter the thresholds that define similarity (yellow) and dissimilarity (red), thus the colour gradient.

Try it out - play with these features and see what happens!

  • For the "Min" score, setting "Lower value" to 0.5 and "Upper value" to 1.5 is generally a safe bet in most situations. Different thresholds are appropriate for different scores, but we won't consider that further in this tutorial.
  • Switch between the three "cluster" superpositions, zoom in and inspect the structure - try to really understand the reason for the backbone colouring.

Note how useful this colouring is for helping you to quickly and easily identify which regions of structure are locally conserved and which aren't!

Local Structural Conservation Of Side Chains

For regions of the structure that sufficiently conserved in terms of backbone conformation, we are often interested in whether side chains maintain their conformation relative to the backbone despite any conformational changes such as domain motion.

To do this, we will consider the "Side RMS" score, which colours residues according to the RMSD of the corresponding side chain atoms after superposition of the local backbone:

Now we can see with more clarity which residues have side chains that adopt a similar conformation in the compared chains (yellow), and which adopt a dramatically different conformation (red).

We can see that many of the side chains adopt different conformations in the two models. Some of these differences may be real (i.e. differences in the crystal), and some may be due to incorrect modelling.

Get familiar with comparing side chain conformations

Since we are comparing side chains, it makes sense to display the side chains. To do this:

  • Within the CCP4mg "Display Table", click the "Ribbons" button corresponding to object "A/" under "1ryx", and change it to a side chain representation such as "Bonds" (note that you can have both representations simultaneously by cloning the object).
  • Repeat this action for "2d3i".
  • Since the side chain representations can get a bit overcrowded, try enabling "Depth cue fog" to allow for better viewing clarity (this is located in the toolbar above the structure).

Try it out - try switching between the different substructure superpositions, zooming and navigating the structure, looking at which side chains have conserved conformations and which do not.

Note that incomplete side chains are compared using whichever subset of atoms are present, by default.

Analysis Of Structural Changes That Occur During Crystallographic Refinement

So far in this tutorial we have focussed on comparing the target (1ryx) and reference (2d3i) structures. We will now consider the comparison of the target structure before (1ryx.pdb) and after (1ryx_prosmart_refine.pdb) refinement with external restraints (data files were made available at the start of part 2).

From this point on, we will assume knowledge of the procedures described earlier in this tutorial, regarding use of the ProSMART analysis features in CCP4mg.

Firstly, follow these steps:

The first thing to notice is that the global conformation of the model didn't change during refinement. This is worth noting, due to the use of external restraints to a homologous structure that is in a different conformation!

Important point - reference structures can be in different conformations:

External restraints can be generated using homologous structures that are in a different global conformation to the low-resolution structure you are trying to refine. Due to the local nature of the external restraints generated by ProSMART, these restraints will not unduly influence the target structure into a different global conformation during refinement.

However, there's nothing to say that local structure hasn't changed. Now we'll investigate whether there have been many changes to the backbone:

Important point - making use of ProSMART structural analysis in model building/refinement:

During part 1 of this tutorial, we identified an interesting region comprising residues 472-477 and 331-343 that requiring manual remodelling. However, it is tedious to systematically manually inspect every residue to see whether the structure has changed and requires manual attention. Using the ProSMART comparative analysis features, you can quickly and easily identify which regions are in most drastic need of attention (if any). Indeed, residues 472-477 and 331-343 could have been easily identified using this method. From looking at the ProSMART backbone conservation analysis, it is also immediately clear that there are various other regions in the structure that are in desperate need of manual attention.

Now investigate whether there have been many changes to the side chains:

In this case, a large number of the side chains have changed conformation, indicating that a lot of manual inspection and model rebuilding in Coot would be required in this case. In other cases, fewer side chains may be identified as having dramatically changed conformation during a given refinement cycle - in such cases, information from the ProSMART comparative structural analysis would dramatically speed up the process by identifying which residues to prioritise looking at in Coot.

Assessing Influence Of Reference Structures During Externally-Restrained Refinement

So far in this tutorial we have compared the target (1ryx) and reference (2d3i) structures, and compared the original (1ryx) and re-refined (1ryx_prosmart_refine) models. We will now consider the comparison of the re-refined model (1ryx_prosmart_refine) and the reference structure (2d3i). This will provide information regarding the degree of influence of the external restraints from the homologous reference model. When combined with the previous structural analyses in this tutorial, along with some manual inspection in Coot to determine causality, this information could help us to gain intuition regarding the usefulness/suitability of the external restraints, and be consequently used to hone the refinement protocol. For example, it may be desirable to use external restraints for some regions/residues, but not for others.

Firstly follow these steps:

The first thing to note is that the target structure has been neccesarily pulled towards the reference structure during refinement - the two structures have locally-similar backbones. However, there are a few regions that have not been pulled into the conformation of the reference structure. The analysis suggests that overall the backbone of the re-refined model is more locally similar to the reference structure than to the original target model.

Note that, when comparing the results from multiple ProSMART comparative analyses, it is imperative for the colour thresholds/gradients to be exactly the same in CCP4mg.

Now investigate the structural conservation of side chains between the re-refined model and the reference structure:

The first thing to note is that there are a substantial number of side chains that are in different conformations between the re-refined and reference models. This is important - this demonstrates that, despite using external restraints for all side chains, the external restraints have not pulled side chains out of their conformation if the density is strong enough to suggest that they should stay where they are. The analysis also suggests that more side chains are in the conformation of the reference structure than are in the conformation of the original target structure.

Important point - robustness to inappropriate external restraints during refinement:

In REFMAC5, external restraints are complemented by the Geman-McClure robust estimator function during refinement. This means that restraints with target values that are very different to the current interatomic distances will be automatically naturally weighted down during refinement. Consequently, inappropriate external restraints should have very little effect during refinement.

Also, good density should help to keep structure in place. If not, it could be that you need to reduce the external restraints weight.

Keep in mind that, under normal usage, it is expected for refinement with external restaints to result in a suboptimal model. After refinement with external restraints, refinement should be performed without external restraints but instead with jelly-body restraints. This should help the model to optimise nicely, sinking into the density whilst maintaining the positive influence of the external restraints.

Final Note

External restraints can pull regions into incorrect conformations, particularly when density is weak. Whilst this sounds like a bad thing, it is not! Indeed, this often has the positive effect of revealing new features in the density, and causing difference density to appear in regions where the "true" model should be placed (where the word "true" is used tentatively...). If this happens, and after manual inspection in Coot you conclude that the external restraints have had a negative affect in a particular region, then you should use ProSMART to regenerate the external restraints specifying for the region in question to be excluded from restraint generation, before re-attempting refinement with the updated set of restraints. Alternatively, you can apply the corrections manually in Coot, before continuing subsequent rounds of refinement with jelly-body restraints.


Back to the tutorial main page