Part 2: Rebuilding in KiNG

A. Thr 77

Fire-up KiNG and read in the model

For this tutorial you will use phenix.king, the version of KiNG that is distributed with Phenix. For other contexts, you can download a stand-alone KiNG application from the kinemage web site.

  1. KiNG is a Java-based kinemage graphics viewer, portable to all platforms capable of running a Java virtual machine and amenable to the addition of new plug-in modules. You've already seen the stripped-down web version, KiNGlet, in Part 1.
    KiNG has numerous tools available; we'll use the Structural Biology subset to rebuild two poorly-fit threonines in 1BKR. In Part 1, you assessed the quality of this model using the MolProbity web service and downloaded a Reduce-protonated PDB file and a multi-criterion kinemage file. You also downloaded the 2mFo-DFc and mFo-DFc maps for 1BKR from the EDS.
  2. In this rebuilding demo, we are using a very high-resolution structure for didactic purposes -- problems are rare in such cases, but when they do occur their nature and their correction can be seen and understood extremely clearly.
  3. In your working directory, type "phenix.king 1bkrFH_reg-multi.kin.gz"

Adjust the kinemage view and read in the map

  1. The multi-crit kinemage contains much information, most of it abstracted in the default view; the only contacts shown are the bad overlaps. Use the kinemage display buttons on the right panel of the KiNG window to prepare for the rebuild: turn off the Cαs and turn on mainchain, sidechain, H's, and water.
  2. Go to Thr 77 using Edit > Find point... Use either the Zoom slider or right-mouse vertical-drag to zoom in. (Adjust the slab with the Clipping slider or right-mouse horizontal drag.) Note the big ball that marks the 0.349Å deviation, and then turn off the "Cbeta dev" button.
  3. Read in the 2mFo-DFc ED map (1bkr.omap.gz) using Tools > Structural Biology > Electron density maps... The map will be displayed in the KiNG graphics window, along with the map's settings widget shown at left (which you'll want to move to a better location). The 10Å default size is OK and the default gray for the first map contour is a decent color. You may want to increase the contour level to 1.4.

Choose a new rotamer

MolProbity identified Thr 77 as an outlier both for rotamer and for deviation. You may be able to tell for yourself that some of the bond angles are distorted and chi1 is eclipsed (look down on the sidechain from its end). Turn the contact dots off (the "dots" button on the panel) and look at the beautiful density versus the model placement, to see that there's definitely a problem with Thr 77. Which atom seems to be fit worst? Since we know that the current model has a non-rotameric conformer, let's start the refitting process by seeing if any of the 3 real rotamers will match the density and give decent contacts.

  1. KiNG uses our Penultimate rotamer library (Lovell, 2000). One way of gaining access to rotamer selection is via Tools > Structural biology > Sidechain Rotator. In the resulting file-selection dialog, navigate to and open the 1bkrFH_reg.pdb file. The kinemage will update with some new buttons: 2 near the top of the list represent the "frozen" and "molten" models used in the rotation. The Model Manager Widget is floated on your screen; in it, turn on "Probe dots", to enable interactive contact display. Select the Thr77 residue for rotation by a Ctrl-click, option-click, or middle-click on one of its atoms. This will pop up another widget, for control of the sidechain dihedral rotations. On this widget, you'll find a dial for each torsion angle, a listing of rotamers for this residue (with % occurrence), a rotamericity score for the current conformer, and a checkbox for sidechain idealization (toggle to see how this setting affects the model, but leave it turned on).
  2. The dials work by dragging around the edge of the circle, left-mouse for fast and right-mouse for slow change. Move one slightly, to trigger update of the contact dots for this starting position (or click the ">" in the Model manager next to the Probe command line).
  3. Click on each listed rotamer (p, t, or m, in the sidechain-control window) to evaluate its match to the electron density and its interactive probe dots (red clashes and pale green H-bonds). [Make sure the "dots" button for the original multi-crit model is turned off, to allow you to evaluate the new contacts.] [Not calculating contacts with water because they are unsure and might have been fit into other density or noise.] Right-click on the and narrow the clipping to isolate the sidechain model and density clearly. Look down the -Cα bond, but rotate your view as needed to get a feel for the fit of each rotamer. [At this stage, matching the shape and orientation of the density is much more important than being centered on it.]
  4. The m rotamer gives the best agreement to the electron density shape. But it needs some improvement: the whole sidechain is shifted to one side, and the CG methyl clashes with the Lys 78 N amide. Adjustment of chi angles will not correct this problem (try it). We need some way to recenter the Cbeta back into its density, which in turn will alleviate the clashes. For this, we will use the "Backrub".

Backrub the Thr 77 mainchain

We've now got what appears to be the right rotamer for Thr 77, but the entire residue needs to be tilted back into the density. For this, KiNG has a unique and effective backbone adjustment tool, the "Backrub". [For more information about the backrub motion, see Davis et al. (2006).]

  1. Click the "BACKRUB mainchain" button on the A 77 THR rotator widget to invoke the backrub tool for this residue. The BACKRUB widget will pop up; take a look at the information and controls that it displays. Beneath the three rotation dials are reports for the Tau deviation, Ramachandran status, and phi,psi angles for the subject residues. Fundamental to the backrub tool are the three labeled dials that control movement of the central residue (Thr77) and its two neighboring peptides. The primary backrub motion rotates the entire dipeptide as a rigid unit around the axis between the two Cαs at i-1 and i+1; a small shift of the backbone is leveraged into a much larger change of the central sidechain position. The two secondary motions (less often needed, and at smaller amplitudes) each rotate one of the flanking peptides around the axis between its end Cαs, which helps to maintain H-bonding and carbonyl O position.
  2. You may want to get a feel for the effect of the three motions of backrub, before trying to get a good fit. To do so at this time, turn off the map and the frozen model in order to concentrate on the movement of the orange sidechains and brown mainchain of the molten model. Play with the three dials of the Backrub widget (especially the primary, center one) to get a feel for their connection with the model. Return the dials to their original values when done.
  3. Now, let's use the Backrub and Sidechain rotator dials to make the model better fit the density. First, make sure that the dials are in their original positions by matching the black and gray numbers for each dial. Be sure the frozen model and *1bkrFH_reg model are on. (A good position: Center on of Thr77 (right-mouse click on the ) and then decrease the map coverage to 4Å at top of the map Presets widget.)
  4. Start out by pulling the of Thr77 back into its density by using the A 77 THR dial (middle dial) on the Backrub widget. Something around 7 to 9° is a reasonable movement.
  5. Turn on the purple 3σ map contours, to see that the Og1 is now in the higher-density arm as it should be. Now use the chi1 dial on the sidechain rotator to move both Og1 and Cg2 more toward the centers of their lobes of density. A chi1 angle near -50 to -55° seems close.
  6. In a real refinement situation, we would quit here - this is close enough to know we've found the right local minimum, and the refinement program will twiddle the rest. However, for this lesson continue to optimize the rebuild. Extend the map coverage to about 8Å, and try the two individual-peptide adjustments to center the backbone COs in their density and maintain H-bonding. Rotate the OH to point at nearby density that might be a water peak (no atom shown, hence no dots, since it's a symmetry related one).

Apply and save the changes

Backrub tool: Click the "Finished" button. Answer "Yes" to keeping the changes.
Sidechain rotator: Click the "Finished" button. Answer "Yes" to keeping the changes.
In the Model manager window: Choose File > Save PDB file... and in the file-selection dialog give your new PDB file a name and click save.

B. His 42

To see density for the His 42 flip from Part 1, enter 42 cg in Edit/Find point, get a good view, and turn on the "H-bonds" button to remember its contacts in this conformation that was preferred by Reduce. The ring electron density is quite clean, with no evidence of disorder. In the map dialog, move the purple contour level up to 4.7σ, where you should see differences between peak levels for N vs C atoms that corroborate Reduce's flip choice (in spite of model bias from having refined it the other way around). In clean density, the distances between neighbor atoms are quite sensitive indicators of whether or not they form an H-bond.

C. Thr 101

  1. In Edit/Find point, type: 101 cb, zoom in, and turn to get a good view. Note that the Cb is not in the density. Read in the mFo-DFc map, and choose the "Coot Fo-Fc" option onder "Presets" in the map widget. The pair of +/- difference peaks around Cb are even clearer, and the + peak at the methyl suggests it's really the Og.
  2. Rebuild Thr 101, as you did for Thr 77. Include a small backrub and a rotation of the OH, to get a good H-bond to the backbone CO on the previous turn of the α-helix. This is a common motif for Thr or Ser, but the person who fit this probably didn't know that.
  3. Another possible reason this misfit might not have been caught is if the aniso B's were used too early. For Thr 101 Cβ, the anisotropy elongates the density between the real and the fitted positions. Find either the 1bkr.pdb or 1bkrFH_reg.pdb file and bring it up in a text editor. Scroll down to Thr 101. The isotropic B is the last field on the ATOM record. How does the 101Cb B compare to Cg and Og, and to other heavy-atom sidechain B's nearby in sequence? Is this normal? Aniso B's have 6 values: the first 3 are the amount of motion along each principal axis of the atomic displacement ellipse, while the last 3 specify its orientation. Look at how much the first 3 terms differ for most atoms here: very seldom as much as 2:1. For 101 Cb, the anisotropy is about 5:1, but it's quite normal for Cg and Og. Is that physically possible? Aniso B's can help model real flexibility at high resolution, but provide extra parameters that also model (and mask) errors.

D. (Optional: Arg 32 -- a trickier case)

  1. Turn on dots, contacts, and overlaps, but turn off the H-bonds, and find point 32 ne; zoom, turn, and clip for a good view of the Arg guanidinium and neighboring atoms. Save and name the view under Views/Save current view, so you can return any time. Arg 32 is a poor rotamer and clashes with hoh 1075 occupying part of what should be the sidechain density. Turn off the "dots" button, and Ctrl-click, option-click, or middle-click on an Arg 32 atom to enable fitting. [If nothing happens, check that Tools > Structural biology > Sidechain rotator is selected.] KiNG brings in the new model with the same dihedrals but idealized geometry, so the difference from the pre-existing model shows its degree of geometrical distortion. Turn on "Probe dots" in the Model manager if it isn't already. Since the sidechain model matches the density shape fairly well, use the chi-angle dials to move the model into the density.
  2. Note that even when you get the hh12 atom close to a backbone CO, you see no H-bond dots. Click on hh12 or nh2 to see that they were assigned zero occupancy (they were, indeed, originally not in the density!) As a workaround for this, scroll left in the command-line box for the Probe dots until you see the flags (such as "-both -stdbonds"), and type -noocc into that flag list (make sure it has a space on either side). Update by clicking on the ">" box next to the command line, and you should then see contact dots to hh11.
    Work on rebuilding, to get a nice H-bond and a fairly good density fit - or, if you're having trouble and are short on time, try setting chi angles near 65, 150, -55, -85. This is still a "poor" rotamer as scored in the sidechain dialog, but is in all other ways enormously preferable to the original.
  3. No H-bond contacts are shown to the nearby waters. That's a deliberate default, since misplaced waters like 1075 otherwise clash overpoweringly. To see those interactions, scroll right in the command line until you see the words "not water" and delete them, then update. Now you should see 3 guanidinium-to-water H-bonds, plus the mess around hoh 1075. You can delete the bad water (although you'll still see its clashes in this KiNG session unless you put "not water" back in the command line). Under Tools > Edit/draw/delete, choose "Punch one point", pick the water (which should disappear), then choose "Do nothing" to turn off the punch function. The modified PDB file could be saved using the File menu of the Model manager.
To see how MolProbity tools have been integrated into PHENIX and Coot, please continue with Part 3. The PHENIX GUI makes structure validation and correction very efficient and simple. The MolProbity tools that have been used at the MolProbity website for many years have been incorporated into the PHENIX package. The PHENIX GUI has been linked to COOT allowing interactivity between the GUI and where structure correction can be made. This allows the user to analyze local validation metrics and then go directly to the problem area in COOT where corrections can be made. The tutorial for this section (Part 3), on how one does validation in Phenix, is linked here: http://www.phenix-online.org/documentation/tutorials/molprobity.html

http://kinemage.biochem.duke.edu/teaching/workshops/CSHL2015/
Jane & Dave Richardson