C687 Tutorial: Model Evaluation Tutorial


Use model evaluation tools in the ProStat menu of the Homology Module of InsightII and the InsightII Viewer Module to evaluate the model of 3znf.pdb. See also the notes about Making Graphs at the end of this tutorial.

Answer the following questions. SAVE YOUR ANSWERS TO THESE QUESTIONS IN A FILE NAMED zf_model_eval.txt. Save at least one Ramachandran plot in a InsightII folder named Ramachandran_Plot.psv (other objects may also be saved in this file). See the WWW/Alignment/Homology Assignment page for details.

The Homology module is currently licensed only on splatter. Therefore, unless you are working on splatter, it is necessary to run the program remotely on splatter, then display on your local monitor.

To run remotely on splatter,
xhost splatter
telnet splatter (then login)
insightII (to start the program)

This tutorial is MUCH FASTER if you use splatter. If splatter is not in use, choose this woorkstation!

1. How many residues have alpha-helical backbone torsion angles?
Do one or more of the following:

2. How many residues have beta-sheet backbone torsion angles?
Do one or more of the following:

3. How many residues have backbone torsion angles that are NOT alpha-helical or beta-sheet backbone torsion angles?
Do one or more of the following:

4. Do certain residues have preferred side chain torsion angles?
Use the Homology/ProStat/Residue_Dihedral menu to measure Chi_1, Chi_2, Chi_3, and Chi_4 angles. Make graphs of these angles to visualize your numerical results.

5. Do certain residues have preferred combinations of chi1 and chi2 side chain torsion angles?
Use the Homology/ProStat/Residue_Dihedral menu to measure Chi_1 and Chi_2 angles. Make a graph of these angles to visualize angle correlations.

6. Are all peptide bonds planar?
Use the Homology/ProStat/Struct_Check menu to Check_Torsions. Check only the Omega angles. Select Color_Molecule to visualize the results on your model. Adjust your N_Std_Dev threshold to a low value until you visualize a change in color in your model.

An Explanation of N_Std_Dev
If you measured the Omega dihedral angle (also known as the CA-N-C-CA dihedral angle or peptide bond dihedral angle) in many protein structures, you would find that this dihedral angle has an average value of 180 degrees, and a very small deviation. In fact, if you plotted all angle values, it would look like a bell curve; the width of the bell curve is the standard deviation. I don't know for sure, but let's say that 1 standard deviation for this angle is 3 degrees.

When you select a N_Std_Dev value in this menu, and select a default color and highlight color, you tell the program to color with the highlite color all CA-N-C-CA atoms that deviate more than N_Std_Dev deviations away from the average; all other angles that are within N_Std_Dev of the average are colored with the default color. This highlites "bad" parts of you model so that you can visualize & fix them.

The program uses a N_Std_Dev value of 3 as the default. According to statistical theory, 95% of all measurements should be within 3 standard deviations of the average. (Statistitians say that measurements within 3 standard deviations have met the "95% confidence limit".) When you try this part of the tutorial with a conformer from the PDB, you'll find that nearly all measurements are within 3 standard deviations of their averages (yet it's interesting that some measurements---dihedral angles for end residues for example---are not within 3 standard deviations!).

7. Do all residues have the right chirality?
Use the Homology/ProStat/Struct_Check menu to Check_Torsions of CA_chirality. Select Color_Molecule to visualize the results on your model. Adjust your N_Std_Dev threshold to a low value until you visualize a change in color in your model.

8. Are all bond lengths and angles "acceptable"?
Use the Homology/ProStat/Struct_Check menu to Check_Bonds and Check_Angles. Select Color_Molecule to visualize the results on your model. Adjust your N_Std_Dev threshold to a low value until you visualize a change in color in your model.

9. Are there any van der Waals "bumps"?
Use the Measure/Bump menu to measure intramolecular bumps.

10. Identify ranges of residues that constitute alpha-helical secondary structures.

11. Identify ranges of residues that constitute beta-sheet secondary structures.

12. How do the Chou-Fasman and GOR-II secondary structure predictions agree with the Kabsch-Sander secondary structure elements of the model?
Use the Homology/By_Residue/SecondaryPred menu and make graphs of the secondary structure prediction of your model's sequence. Choose a file name of znf_cf.htbl for the Chou-Fasman prediction algorithm, and znf_gorII.htbl for the GORII prediction algorithm. Compare the graphs to the contents of the files.

Color Secondary Structure
green alpha-helix
red beta-sheet
purple reverse turn
white random coil (GorII only)

Evaluating Secondary Structure Prediction Graphs The absolute value of the propensity for a region of residues to form a particular secondary structure is not useful. Instead, a significantly greater propensity for the most likely type of secondary structure relative to the other less likely secondary structures is important.

Secondary structure prediction methods are known to be only moderately accurate. To improve your confidence in a secondar structure prediction, PERFORM A MOLECULAR MODELING EXPERIMENT: try more than one prediction algorithm (e.g., Chou-Fasman and GorII) and compare their results.

The contents of your znf_cf.htbl and znf_gorII.htbl files list the numbers used in these graphs and provide further details.

13. Are there any buried polar groups or buried hydrogen bond donors or acceptors?
Use the Homology/ProStat/Access_Surf menu to measure the Area, Fractional_Area, and Polar_and_NonPolar areas of Heavy_Atoms_Only of your model. Use a Solvent_Radius of 1.4, standard VdW values, and choose a unique Run_Name. Once the calculation is finished, review the results of these files in a UNIX window: Evaluate whether polar atoms and potential hydrogen bond acceptors & donors have small surface areas.
Evaluating Buried Polar Atoms The information provided by InsightII to evaluate buried polar atoms is numerical, and not visual. Therefore, evaluating buried polar atoms is not straightforward. Also, for this tutorial, almost all residues are exposed to solvent, because the zinc finger structure is relatively small (only 34 residues).

The name_area.tab file lists surface areas for each residue. This table can be used to identify polar residues that have relatively low surface areas, and therefore may be buried. To filter for particular residues, type
more name_polarity.tab | grep restype For example, to filter for all LYS residues in the file znf_polarity.tab, type
more znf_polarity.tab | grep LYS

The name_polarity.tab file lists surface areas for each atom. This table can be used to identify polar atoms that have relatively low surface areas, and therefore may be buried. To filter for particular atoms, type
more name_polarity.tab | grep restype | grep atomtype For example, to filter for all NZ atoms at the ends of LYS side chains in the file znf_polarity.tab, type
more znf_polarity.tab | grep LYS | grep NZ

14. What do the hydrophobicity profiles indicate about this model?
Use the Homology/By_Residue/Hydrophobicity menu, graph your the hydrophobicity profile of all residues of your model with a Smoothing Window Size of 5, and don't set levels (use the default values). Make a graph for each type of scale. Note that some algorithms use positive values for hydrophobic regions, and other scales use negative values for hydrophobic regions. The absolute values of hydrophobicities are not important (in fact the absolute scales of different algorithms have very different maxima and minima), but the relative values & shapes of the hydrophobicity curves are important.


Making Graphs

Follow these steps to make a Ramachancran plot:

  1. Create a InsightII table with the data that you want to graph (e.g., measure Phi and Psi angles using the Homology/ProStat/Residue_Dihedral menu). Manually edit the table if necessary (edit blank cells, insert new columns, etc).

  2. Click the LEFT mouse button on the top cell of the colum that contains your X-axis data (e.g., Phi)

  3. Hold down the <Crtl> button and click the LEFT mouse button on the top cell of the colum that contains your Y-axis data (e.g., Psi).

  4. Click on the Graph icon at the bottom of the table window (the middle icon). A grap will appear.

  5. Click on the Graph icon on the left of the InsightII window (two icons below the blue & white "Biuosym module" icon).
    • Change the Label of the Graph to "Ramachandran Plot"
    • Change the CharSize of the Axes and Graph to 0.04.
    • Change the Color of the X axis, Y axis, Box_segment 1, and Box_Segment 2 to light blue.
    • Change the Color of the Title to green.
    • Change the Color of the plot to yellow.
    • Use Modify_Display ito modify the display of your graph. Turn off bar, line and graph, and turn on point. Select a Point Symbol, and change the Symbol Scale to 10.0. Set Graph to graph_name:1
    • Set the Threshold of the X axis and Y axis to a Min Value of -180 and a Max Value of 180. Turn on Zoom_Axis.
    • Divide your plot into four quadrants: Change the Tick Mark so that Extended tick marks Start at -400 and End at 400 with a Mark Step of 200. Use a Labe Display of none. Add these makrs to the X axis and Y axis.

Back to  |  C687 Spring 1999  |  Courses & Instruction  |  MolViz Home  |
Send comments to chemvis@indiana.edu
Last updated: 01/23/2001