Rebuilding Tutorial: Fixing Errors in 6m4j
Learn to use MolProbity and Moorhen to identify and fix structural problems
Warm-up: Reading Electron Density
The interactive viewer below shows a calculated map for an Ala-X-Ala tripeptide: three amino acid residues in a row, where the central residue (X) is unknown and must be identified from the map.
Look at the electron density below. (Left mouse rotates, Right mouse slides, Scroll wheel zooms.) Select the amino acid you think the sidechain density matches. If you guess wrong, you can try again. If the answer is ambiguous, click Improve Resolution to get more information. If you are stuck, click Reveal Answer.
Connected to classroom
Waiting for instructor to begin the warmup...
What amino acid is this?
Warmup Complete!
You have practiced reading electron density at multiple resolutions, interpreting distinctive density shapes like a Tyrosine sidechain and ambiguous shapes like Asparagine/Aspartate. In the levels ahead, you will apply this skill to evaluate and solve errors in a real structure.
Gather Your Party: File Acquisition and Preparation
Prepare your files and molecular viewers for the tutorial.
Go to MolProbity Next, click the
Fetch from PDB tab (next to File Upload), type 6m4j in
the PDB ID field, and click Fetch & Analyze.
Or jump straight to the pre-analyzed 6m4j results.
To fix a structure, you need to see the electron density, the experimental data that the model should fit. Download the map file for 6m4j:
Save this file somewhere you can find it. You will load it into Moorhen in the next step.
On the results page, click the Moorhen button in the 3D viewer panel (in the Mol*/Moorhen/Kin toggle). This opens a web-based molecular modeling tool where you can edit the structure.
Now drag and drop the 6m4j-sf.mtz file
you downloaded onto the Moorhen viewer. Moorhen will compute
electron density maps from the structure factors. You should see
blue density mesh appear around the model.
To adjust map settings in Moorhen, click on Maps in the Moorhen menu.
Map level can be manually controlled with the slider. An active map, whose level can be adjusted with Ctrl+Mouse Scroll, can be selected with a toggle.
Map colors can be set for accessibility or preference. Click on the small colored box next to each map name to open a color selection window.
Level 1: Your First Fix — A Simple Rotamer Flip
Rotamers are commonly-observed and energetically-favorable conformations of sidechains. A rotamer outlier indicates that a sidechain is in a conformation that is almost never seen in high-quality structures. Often this means the sidechain was built into the map "backwards". Moorhen can fix these cases with one click.
On the results page, go to the Residue Details tab. Check the Show Outliers Only checkbox above the table to filter down to problem residues. By default, the table sorts by MP-Triage score, which combines all of a residue's validation results into one number that suggests how urgently it needs fixing, so the residues most worth your attention come first. To find a specific residue, either click the Residue column to change sorting to sequential or type a residue number in the search box at top right.
Find chain A, residue 238 (THR). Notice the colored flags next to it. What kind of problems do they indicate?
Select the Kin mode above the viewer, then click on the A 238 THR row in the table to center the viewer on that residue. The gold highlighting shows the part of the model that has a problem:
Based on the table flags and the kinemage view, which of these outlier types does A 238 THR have? Select all that apply.
Rotamer outliers are marked with a gold trace over the sidechain. C-beta deviations are marked with magenta spheres at the CB atom. Covalent angle outliers are marked with red or blue fans.
See the Validation Guide for a complete description of visual markup.
Select the Moorhen mode above the viewer. Navigate to residue A 238 THR again by clicking on that row of the residue results table if necessary.
Rotate the model in the viewer and look at the density map. Note that, at this resolution, the end of the sidechain does not match the shape of the density. Use Ctrl+Mouse Scroll to adjust the map level in Moorhen for a clearer view.
Also note the difference density peaks near the sidechain. Difference density is a strong indicator of a modeling error, especially at better resolutions.
Recenter (middle click) and rotate so that you can look down the CB-CA bond. Tetrahedral centers like these prefer an evenly spaced “staggered” configuration. However, this sidechain is “eclipsed”, with the CB-OG1 and CA-C lines almost overlapping each other. This is energetically unfavorable.
Additionally, the residue has a CB deviation and a covalent angle outlier visible in the KiNG view. All these observations strongly suggest that the THR 238 sidechain outlier is a modeling error rather than a rare feature. It should be refit.
Right-click on any atom in the THR 238 residue in Moorhen and select Auto-fit Rotamer. Moorhen will try all common threonine rotamers and pick the one that best fits the electron density.
You should see the sidechain snap to a new orientation.
Autofit rotamer is at the top left of the right-click menu.
Its icon shows a black side chain surrounded by a green aura:
If you do not see the right-click menu, make sure you are clicking directly on the atoms of the residue, not on empty space. The text window at the top right of Moorhen shows which atom your mouse is pointed at.
Look at the sidechain and the density. Does the model fit better now? Congratulations, you just fixed your first validation outlier!
What happened when you auto-fit the rotamer?
Now find A 126 LEU, another rotamer outlier. Note that this residue is marked with only a rotamer outlier. You will have to rely on the map and structure context to judge whether this residue is a modeling error or a rare feature.
Examine the rotamer outlier by recentering so that you can look down the CG-CB bond in the sidechain. The CG-CD1 bond line is only partially eclipsed with the CB-CA bond line. This is unfavorable, but it is more plausible than the totally eclipsed conformation in the A 238 THR case above.
Examine the electron density. Is the sidechain a good fit to the density? Is there difference density nearby? Does the map provide convincing support for an unfavorable conformation?
Zoom out and look at the structural context of the sidechain. Is it located somewhere that might functionally justify an unusual conformation, like a ligand site? Does the sidechain look like it is closely packed with other sidechains in a way that would hold a strained conformation in place?
Neither the map nor the structural context justify keeping an outlier conformation here. Auto-fit the rotamer in Moorhen and compare the sidechain to the density before and after. Does the model fit the blue mesh better now?
Level 2: Interpreting Clashes
In Level 1, you evaluated sidechain rotamer outliers based on the electron density map and selected validation results. Now you will add an especially powerful validation tool to your evaluations: all-atom contacts. All-atom contact analysis shows steric clashes, hydrogen bonds, and van der Waals contacts. Fixing modeling errors often eliminates clashes and creates hydrogen bonds.
Navigate to A 286 THR in the residue table. Notice it has several flags visible there and in the Kin viewer: 2 clashes, a rotamer outlier, a C-beta deviation (0.44 Å), and 2 covalent geometry angle outliers.
C-beta deviation measures how far the first sidechain carbon (CB) is from its expected position. A large deviation indicates the backbone geometry or backbone-sidechain connection is distorted.
As in the Level 1 examples, multiple different validation metrics all pointing at the same residue are strong evidence of a modeling error.
Static contact dots are already available in the Kin viewer. Interactive contact dots can be activated in Moorhen and are useful for outlier interpretation.
Switch back to the Moorhen viewer and center on A 286 THR. From the main Moorhen toolbar, open the Models menu. Expand the Tools menu, then activate the Contact Dots toggle.
It will probably take a moment to calculate and display the dots. You should see pink spikes representing clashes, blue/green/yellow dots representing van der Waals packing contacts, and dark green clusters representing hydrogen bonds.
Rotate the view in Moorhen and inspect the contact environment of A 286 THR. Note that the CG2 atom has clashes with 2 different waters, and that those waters are well-supported by the map. Also note that the OG1 atom makes a small packing contact (favorable) with the planar surface of the 152-153 peptide bond.
Auto-fit the rotamer for A 286 THR, just like in Level 1. Watch the contact dots. What did you see?
Did fixing the rotamer also clear the clash spikes?
You should see the pink clash spikes disappear. In their place, the OG1 atom now forms two favorable hydrogen bonds. Additionally, the CG2 atom now has slightly better packing with the 152-153 peptide than the original OG1 did. All the contacts have improved and no new problems have appeared! This is strong evidence that the rebuilt rotamer is correct.
Return to the MolProbity residue results table, and make sure Show Outliers Only is active. Type “THR” into the Search box at top right. The table should now display all the Threonine residues with validation outliers.
Look at the validation tags for the THR residues. What patterns do you see? You should see that THR rotamer outliers in this structure are almost always accompanied by clashes and C-beta deviations. This pattern is typical of the kind of backwards THR sidechains you just fixed.
Search instead for VAL residues. A/B 187 VAL is also backwards; do you see the same pattern of outliers? Now search for LEU residues. A/B 18 and 126 LEU are backwards. Again, do you see a pattern in the rotamer outliers?
Why do these backwards THR sidechains make clashes more often than backwards VAL or LEU?
VAL and LEU both branch into two chemically similar carbon atoms at the end. THR branches into a carbon and an oxygen, which are chemically different and have different sizes. Flipping a THR is likely to put the OG1 oxygen into a position where it doesn't fit, causing a clash.
A/B 92 LEU and A 348 VAL do have both rotamer outliers and clashes, despite backwards LEU and VAL typically lacking clashes. However, not all rotamer outliers are backwards sidechains. These residues have a different kind of problem and therefore match a different pattern of outliers.
Another pattern to note: backwards VAL 187 has a CB deviation, much like all the backwards THR residues. However, the backwards LEU residues do not have CB deviations. This is because the LEU is longer than THR and VAL. The atom most affected by a backwards LEU is the CG atom, which is not part of the CB deviation check.
Use the LEU, THR, and VAL outlier searches to find a few more backwards sidechains to fix. Recommended residues to inspect and correct include A 241 THR and A 187 VAL.
As you work on additional cases, note that fixing rotamer outliers also fixes the associated clashes. (C-beta deviations are not immediately recalculated, but those are getting fixed, too.) Here, the rotamer outliers are the root cause and the clashes are a symptom. Always look for the underlying problem.
You have been correcting residues in the A chain of this structure. This is a homodimer, meaning it contains two copies of the same chain. Notice that the same outliers are almost always repeated in the B chain copy. This is common in structures with multiple copies, where the same errors get propagated or duplicated. You can fix the B chain for practice or refer to it as a “before” state for your work.
Level 3: Clashes and Alternate Conformations
Like Level 2, this problem involves a rotamer/clash combination, but with an added complication: alternate conformations. Sometimes crystallographers model two (or more) positions for the same residue, typically labeled A, B, and so on, to represent conformational flexibility, with each position at partial occupancy. When one conformer clashes with a neighbor, the problem can be easy to miss if you only look at the default conformer.
Whenever you have made corrections you like, and whenever you are about to attempt a difficult correction like this one, it is a good idea to save your progress. Do not rely on Undo to save you from a misclick!
Look for the Download PDB button above the residue results table. It should appear once you have made changes. Save a copy of the structure, with a unique name.
On the results page, look at the 3D viewer panel. In the row of outlier-type buttons above the viewer (Clashes, Rotamer, etc.), click Hotspots. A panel will appear below listing clusters grouped into Priority and Secondary tiers. In the Priority list, find the cluster containing B 159 GLU (score 14.6, 15 clashes) and B 190 MET (score 15.4, 18 clashes). The MP-Triage score puts the residues that most need fixing at the top. Scores above 3 are flagged red in the table.
These two residues are among the worst clashing in the entire structure; B 190 MET has the most clashes of any residue. Click the Focus button for the cluster to center the 3D viewer on them.
In Moorhen, look carefully at B 190 MET. You should see two positions for yellow sulfur atom, which Moorhen labels “B - Met 190 - SD :A” and “B - Met 190 - SD :B”, as well as the rest of the sidechain. These are alternate conformations (sometimes called “alt confs”), named A and B. The crystallographer modeled two orientations because the density supported both.
Now look at how they relate to B 159 GLU. It is specifically conformer B of MET 190 whose sidechain clashes with GLU 159. Conformer A may be fine on its own. Interpreting modeling errors in alternate conformations can be tricky because clashes and other diagnostic outliers may only be present in one conformer.
Alternate conformations affect contact analysis and clashes: an alternate conformation is only checked against atoms with the same alt label, or atoms with no alt label at all. Alt A and alt B never clash with each other, however much they overlap on screen, while an unlabeled atom is checked against every conformer.
Which part of the model would you try to fix first?
Since the clash comes from conformer B of MET 190,
it would be simple to delete that conformer entirely.
However, B - Met 190 - SD :B has clear support in the
map. In Moorhen, right-click on a B conformer sidechain atom and select
Change rotamers, which has an icon with a yellow
sidechain:
.
Manually cycle through methionine rotamers using the arrows
at the top of the Moorhen window until you find one that preserves the
current SD atom position and moves the CE atom to a clash-free
position. You will have to click the check mark to accept a change
before the contact dots will update.
What rotamer best preserves the SD position in density and avoids clashes?
The rotamer that best preserves SD and avoids clashes is mmt.
There is a case for mmp as the best rotamer overall, based on the map, but it results in clashes. See Step 5: Bonus round for more on how to make mmp work.
Since this structure is a homodimer, it may be informative to check the conformation in the other chain. Use the Sequences display at the bottom of the Moorhen window or the search function in the MolProbity results to navigate to A 190 MET A.
The A chain conformation has a clean alternate B and a modeling error in alternate A. The
terminal CE atom is placed in negative difference density, resulting
in a clash, while a positive difference peak is within reach. As in
step 3, use the Change rotamers tool
to find a promising rotamer that puts CE near the positive difference peak.
The rotamer mtt is a good candidate, but is not centered in the density.
Select mtt, then use the Refine Residues tool
to refit the area.
Across the two chains, you have performed fixes resulting in similar alternate A positions, and two slightly different alternate B positions. This would be a good time to send the model back to refinement. The map should improve slightly, making it easier to identify the correct position for alternate B.
If you are following a live tutorial, do not attempt this task. It is more challenging, delicate, and time-intensive than a live workshop permits.
Save your work before attempting ambitious corrections like this. Moorhen’s live interaction with MolProbity is convenient, but the full version of Coot may be better suited to certain difficult tasks. Moorhen/Coot experience may be required.
Return to chain B MET 190. The mmt rotamer you built in step 3 avoids clashes, but may not actually be correct. Try building alternate B as the mmp rotamer, which places the CE atom near a positive difference density peak. This results in a clash with HOH 686, but that water has a lower density peak than most other waters in the area, suggesting that HOH 686 could be modeled as partially occupied in the A conformer only. The Met 190 mmp rotamer also clashes with Glu 159. However, Glu 159 is surrounded by several strong peaks of highly suggestive difference density. The glutamate probably has an alternate conformation that pairs with the alternate Met 190!
Try assigning partial occupancy and an alternate A name to B HOH 686. Also try building alternate conformations for B Glu 159. Take care when assigning alternate names, and remember the scoring rules from step 2: Alt A is never checked against alt B, and atoms without alt names are checked against all conformations. This will be challenging. Good luck!
Level 4: Is This Really Cysteine?
Not all problems are about geometry and contacts. Sometimes an entire molecule is modeled in error. This requires detective work beyond the structure, investigating the experimental conditions.
6m4j is titled “SspA in complex with cysteine.” The depositors claim a free cysteine molecule is bound in the active site. Find it in the results table: A 402 CYS (listed as a ligand).
MolProbity flags it with 2 clashes and an angle outlier (score 7.7). That is suspicious for a small, simple ligand.
In Moorhen, center on A 402 CYS. Over the previous Levels, you have built up several visual displays that add informtion to the structure. However, this case is easier to understand with simplified information. Turn off the contact dots by opening the Models menu, then the Tools dropdown, and switching the toggle. Turn off the difference density by opening the Maps menu, then clicking the eye symbol next to the difference map name. This should give you a clear view of the ligand and the map.
Cysteine (CYS) has just 6 non-hydrogen atoms in a short chain. Look carefully at the shape of the density around the modeled cysteine. Does the density match the shape of a cysteine molecule? If the density blob looks bigger or more branched than that, it might be something else.
Now that you have a clear sense of the density, toggle the contact dots to reveal that there are two nearby waters that clash severely with each other. Water-water clashes are often an indication that waters have taken the place of covalently bonded atoms.
Look up 6m4j on the PDB website and check the crystallization conditions. Click the Experimental tab on the PDB page to find the crystal growth conditions. What buffer and precipitants were used?
Based on the density and crystallization conditions, what is this ligand most likely?
Compare the density shape to each candidate molecule:
Validation can flag geometry problems, but identifying a wrong molecule requires thinking beyond the model. Look at the density, especially unoccupied parts of the density, and think about what chemicals were present in the experiment. Always check crystallization conditions when a ligand looks suspicious.
In your own work, keep in mind what reactions might occur in your system to change a ligand, and what contaminants might enter your experiment from elsewhere in the lab.
In Moorhen, delete the problem molecules from the model.
Right-click on an object, then select Delete item
.
Confirm to remove the residue or atom. Repeat this process until all
of the intended ligand density is free from atoms.
The structure is now ready for a better choice of ligands.
If you are following a live tutorial, do not attempt this task. It is more challenging, delicate, and time-intensive than a live workshop permits.
Save your work before attempting ambitious corrections like this. Moorhen’s live interaction with MolProbity is convenient, but the full version of Coot may be better suited to certain difficult tasks. Moorhen/Coot experience may be required.
Based on step 3, the ligand with PDB code FLC is a good candidate for this density. Try placing FLC into the model with Coot or Moorhen tools, then moving and refining it until it fits the density.
Boss Fight: A Sequence Register Shift
This is the hardest kind of error to fix. A register shift means a whole segment of the protein backbone is threaded through the density with the wrong residues in the wrong positions. MolProbity cannot tell you directly that it is a register shift, but the pattern of errors is unmistakable.
Click the Hotspots button above the 3D viewer (as you did in Level 3). In the Priority list, the cluster around A 259-266 contains the highest-scoring residue in the structure and is devastating:
- A 259 SER — score 17.2, twisted omega (-60°; the peptide bond angle, normally ~180°), 14 clashes, 2 bond outliers, 3 angle outliers (covalent geometry strained beyond normal limits)
- A 260 ALA — score 13.6, twisted omega (-123°), Ramachandran outlier (backbone dihedral angles in a forbidden region)
- A 263 THR — rotamer outlier, 16 clashes
- A 264 GLU — rotamer outlier, 7 clashes
- A 265 PHE — cis omega, Ramachandran outlier, C-beta 0.57 Å
For detailed explanations of each metric, see the Validation Guide.
Notice what types of errors are present:
- Twisted/cis omegas — the peptide bonds are distorted to accommodate wrong residues
- Ramachandran outliers — the backbone is being forced into impossible angles
- Rotamer outliers — sidechains cannot find reasonable conformations
- Bond and angle outliers — even the covalent geometry is strained
- Massive clashes — atoms are being forced on top of each other
What pattern of errors suggests a register shift rather than individual mistakes?
Fixing a register shift requires rebuilding the affected segment. There are two main approaches:
Option A: Delete and rebuild from scratch
- Delete the mis-registered residues
- Rebuild the backbone through the density with the correct sequence assignment
- Refine the rebuilt segment
- Re-fit the sidechains
Option B: Gradually drag atoms into place
- Working from one end of the shifted region, drag individual atoms or small groups of atoms into the correct density
- Use real-space refinement after each move to regularize the geometry
- Work your way along the segment, nudging the chain into the right register one residue at a time
Option A is more dramatic but can be faster for large shifts. Option B is gentler and works well when the shift is small (one or two residues) or when you want to preserve parts of the model that are already close. In practice, experienced crystallographers use whichever approach best fits the situation, or a combination of both.
This is advanced work that goes beyond what we can cover step-by-step here, but recognizing the problem is half the battle. In a real refinement, you would use tools like Coot or Moorhen to manually rebuild, then re-refine with your refinement program (REFMAC, Phenix, etc.).
Chain B has a similar pair of Priority clusters covering B 255-269 (B 264 GLU scores 15.3). The same register shift error was propagated to both copies of the protein, as we saw with the rotamer outliers in Level 2.