[Kinemage Web Site]

Structure Validation Workshop

Using All-Atom Contact Analysis for Model Diagnosis & Repair

[MolProbity Server]

Scope and Objective

In this part one of the tuturial, you will use the MolProbity web service to add hydrogens to an example PDB structure, 2SIM. You will assess the model's quality via flip-state of HIS, GLN, and ASN sidechains, with Ramachandran plots, comparison against a rotamer database, a graph of C-beta deviations and also make and download from the webservice a kinemage file for the all-atom contact analysis of Part 2.

Part 1: Running the MolProbity web service to add H, calculate contacts, and assess geometry

(done on file 2sim - a neuraminidase at 1.6Å resolution).


  1. Make a working directory for this practical, if you don't already have one.
  2. Open a browser configured with Java 1.3 or later (e.g. a recent Mozilla or Netscape). Go to the web site at http://kinemage.biochem.duke.edu/, and choose MolProbity on the navigation bar. In the intro page, choose to use the KiNG viewer then proceed (with the standard MolProbity version for now - you can try the new alpha-test site later if you want).
  3. On the main page, enter "2sim" in the field for choosing a PDB file and upload. Check what is reported about the file, and continue.
  4. Back on the main page you should see a Java thumbnail kinemage of the C-alphas colored N to C; drag with the mouse to rotate it and admire the beta-propeller structure with a bound dansyl ligand.

  --  click on a thumbnail to see a larger screen capture in a separate window  --  

Add hydrogens & evaluate Asn/Gln/His flips

  1. Scroll down to see the available tools (some are grayed-out until H atoms are added). Choose the Add H option, which will now run the Reduce program on your file; don't change any of the advanced options. This process takes 10-20 seconds on 2sim, since the analysis optimizes entire local H-bond networks.
  2. When done, a report is shown on those Asn, Gln, and His sidechains which Reduce decided should be flipped by 180 degrees, including scores and their differences. Continue, and Flipkin will now run Reduce with the opposite N/Q/H orientations and make kinemages to show you the alternatives. When done, you are offered those kinemages for either viewing or download.
  3. Choose to view the Asn/Gln flips. When it has loaded in the Java graphics window, pull down the list of views and go in turn to each one marked with an asterisk (those that were flipped). You may want to enlarge the window. Animate between the two orientations by pressing "a" on the keyboard (or the Animate arrows in bottom of the right panel) evaluate the pale green H-bonds vs the red clashes, and decide whether you agree with Reduce's choice. All N,Q,H have flip green/red (good/bad) pairs: in the Views marked with a star the green ones have been flipped by Reduce.
    Asn 206 is a no-brainer, while Gln 194 is a bit subtler. Close the browser window containing the graphics display to view the flip score report. The check boxes give you the option of undoing a flip, in the rare but definitely possible case where you feel the algorithm made a misjudgment. These are all sound choices in 2sim, however, so just continue (choose "Regenerate H..." but nothing will be done) and return to the main page.

  --  click on a thumbnail to see a larger screen capture in a separate window  --  

Assess Geometry - Phi,Psi

  1. Our updated geometrical-validation tools (in the green box) can be run either before or after adding hydrogens.
  2. Click on 'Ramachandran plots'. The Java kinemage comes up showing all residues. Animate to the general-residue case, with a green contour around the favored area that includes 98% of the high-resolution, low-B data in an updated reference set and a blue contour around the allowed region that includes 99.95% of the high-quality data. 2sim has two residues outside the allowed contour, which constitute outliers that should be examined with real care (Ser 230 is especially bad, but can be fixed).
  3. Animate thru the other 3 cases (Gly, Pro, and pre-Pro), which show no outliers for 2sim. Close the window with the text Ramachandran report.

  --  click on a thumbnail to see a larger screen capture in a separate window  --  

Assess Geometry - Rotamers

  1. Sidechain conformations cluster quite closely around local minima in the multidimensional chi-angle space. We use a smoothed plot of values found for high-resolution, low-B residues to define a contour that encloses 99% of that good data. Residues outside of that contour may be genuine (e.g. if they have enough compensating H-bonds to keep them there), but they should be examined individually.
  2. On the main page, choose the 'Rotamer analysis' option in the green box, for a list of all residues in the structure with dubious sidechain rotamers. For 2sim over 8% of the sidechains are worse than the 1% level, which is much more than it should be at this resolution.
  3. Look at the listed chi values; can you see for many of them why they are unfavorable? We will work on Thr 159 in Part 2. Return to the MolProbity main page.

  --  click on a thumbnail to see a larger screen capture in a separate window  --  

Assess Geometry - C-beta

  1. We have found that the deviation of the model Cbeta atom from ideal position is a single quantity that encapsulates much of the critical information from bond-angle deviations.
  2. On the main page, choose the 'C-beta scattergram plot' option in the green box, for a plot of all the Cbeta deviations in the structure shown on top of an ideal-geometry residue.
  3. 2sim has about 8 or 10 residues with anomalously large Cbeta deviations. Click on a few of those points to identify them - many are branched-Cbeta residues; we will work on Thr 159 in Part 2.
  4. Close the browser window containing the scatterplot.

  --  click on a thumbnail to see a larger screen capture in a separate window  --  

Calculate All-Atom Contacts

  1. Now choose the 'Find all-atom contacts' option in the yellow box, which will run the Probe program on the entire structure, in either a graphics or a list mode.
  2. First run the 'clashlist' version. 2sim has a relatively poor clash score, which is why it makes a good example (at that excellent resolution, people sometimes believe that the data will take care of everything). The clustered list can be used to find places that need attention.
  3. Contact dot kinemages are generated by selecting the radio-button for Probe on this page. In this case, you should also check the top three boxes in the right-hand column of advanced properties: H-bonds, VDW contacts, and C-beta deviations will be added to the kinemage graphic files. An option to be used later, if you want, is to generate a "Multi-Crit" kinemage. This option is available in the blue-boxed "Other Tools (under development)" section on the main page.
  4. Choose to view the sc-sc and sc-mc contacts. (If your browser was not able to run King, you could download this big file and view it in Mage or KiNG locally on your own computer).
  5. Turn off wide contacts, close contacts, small overlaps, water, and H-bonds (but not H's) to get an overview of the problem areas. In front near the middle is a cluster of 3 bad clashes (red) near a very big Cbeta deviation (gold ball). (If you have lost track of "front", go to menu item Views/1 Unnamed view). Use a right-mouse-click or turn on PickCenter to center on that gold ball (it should be Thr 159, with a Cbeta deviation of .496Å), and zoom way in. The clashes surround the Thr methyl group, which is much too close to a backbone NH and 2 waters. Rotate to look down the Ca-Cb bond, to see that the chi1 angle is eclipsed. This residue looks very bad by many independent criteria (it was also on the bad rotamer list), but we will easily repair it in Part 2.

  --  click on a thumbnail to see a larger screen capture in a separate window  --  


Close the browser window containing the KiNG graphics, return to the main page, and scroll down on the main MolProbity page to where the files are listed. Download 2SIMH.pdb and 2SIMH-contact.kin into your working directory, logout from MolProbity, and exit the browser.

  --  click on a thumbnail to see a larger screen capture in a separate window  --  

Jane & Dave Richardson
previous: Introduction next: Part 2