Determining a Protein Structure Using Homologous Proteins (Homology Modeling)

Oscar Moran, Istituto di Biofisica, CNR, Genova (Italy)

moran@ge.ibf.cnr.it
Based on Tutorial For Deep View (Swiss-PdbViewer), Version 3.7b2, Gale Rhodes, Department of Chemistry,
University of Southern Maine, Portland, Maine, USA. Modified with permision of the author.

Introduction

Homology modeling is a form of structure determination of proteins, based on the assumption that
proteins that are homologous in sequence are similar in structure. With nothing more than a modern
computer and software, and knowing only the amino-acid sequence of a protein, you can often
roughly determine its structure.
Homology modeling involves finding homologous proteins whose structure is known, and then
building your sequence onto the homologous proteins as templates. This process is called threading
the unknown protein onto the reference proteins. Much of threading can be automatic, but applying
some judgment in regions where the protein homologies are weakest can greatly improve the
model. These areas of weakest homology are usually surface loops, which require most of the
manual labor of modeling. The goal in improving the loops is to align as many residues of the
threaded model as possible with those of the reference proteins, and to minimize the length of gaps
(visible in the model as long bonds) in regions where the model has fewer residues than the
references.
After threading, the model is likely to harbor chemically unreasonable features, such as parts that
clash with each other, or very long bonds across gaps. Some, but not all, clashes can be repaired
with Deep View. Usually, however, the last stage of homology modeling is energy minimization, an
automated process by which a computer program allows the model to settle into a lower energy
conformation as similar as possible to the threaded conformation.
In the tutorial provided by the programmer of Deep View, the starting point is the amino-acid
sequence of a protein asigned to each group, whose three-dimensional structure is unknown. You
will find at least one homologous proteins and thread your model onto them, improve the fit of
several loops, fix clashes, and prepare for energy minimization.

Exercise
Here is an outline of what you will accomplish in this exercise. In the following paragraphs, you
will find more guidance about how to carry out each step. A sequence in fasta format will be
assigned to each group
 1. Drawing on what you learned from the homology modeling tutorial, find and use an appropriate
template to make a homology model of mouse lysozyme C.
2. Submit your final model to SwissModel for optimization.
3. Compare the optimized model with the raw model.
4. Assess the capacity of the optimized model for binding the substrate NAG units that are bound
to 1LZSChA.

The following instructions are not as detailed as those of a tutorial. You may have to search menus
for tools you have not used before. In some cases, the point is to figure out how to get a desired
result or make a desired measurement.
I. Making the Raw Homology Model
- Start SPDBV. No windows are open, but Deep View menus are available.
- SwissModel: Load Raw Sequence To Model
- Browse to the file MousLys.fst and open it. It appears as a large alpha helix.
- SwissModel: Find Appropriate ExPDB Templates
Your net browser (Netscape or Explorer) should start and provide a form for you to submit. Notice
that your fasta sequence is being sent to a server in Geneva for a template search. ExPDB templates
are single domains or single chains extracted or exerpted from PDB files expressly for use as
templates in homology modeling.
In a few seconds, a list of possible templates appears. Read the entries and find the first one that is a
native model (not a mutant). You can read more about this model by following the link in the Parent
PDB column, which takes you to the PDB and the file from which the ExPdb file is obtained. The
Blast scores tell you how closely the sequences of the templates match those of your target protein.
Small Blast scores are better. Think of them as a probability that the match between unknown
sequence and the template is coincidental. Clearly, the correspondence results from evolution, not
concidence.
- In the Download ExPDB column, click PDB code to obtain the template. It should come to your
computer, and you can open it in SPDBV by file:open pdb.
- Color: Layer to give the models distinct colors.
- Fit: Best (With Struct. Align). This threads the unknow sequence model onto the template, and
aligns their sequences. Blink between the two models to see how similar they are.
- Prefs: Alignments. This preference dialog controls the appearance of sequence alignments for text
files and printing. Check all boxes under "For Files:" to make a full-featured alignment file. Click
OK.
On the left side of the Sequences Alignment window, click the small document icon below the ?.
The alignment file appears, with the feature you set in the Alignment Preferences. It shows the two
sequences aligned, with asterisks (*) marking identical residues, dots marking similar residues, and
no marks under residues that are dissimilar. In addition, h and s mark residues in helical and sheet
conformations.
- Save: Alignment. Give the filename seqVstemp.txt. Retain this alignment file to hand in. (This file
can be opened and printed in any text editor or word processor. For proper alignment, you need to
apply a monospace font like Courier or Monaco.)
Wind: Sequences Alignment and Wind: Layer Infos. These two windows are handy for working
with multiple models and for their sequences.
-Save your model and template (File:save:project) before any further action.
Now you can try to minimise your model. See at preferences. Establish a minimisation protocol (i.e.
20 steps with stepest decendents and 20 steps with conjugated gradient)
- Fit: Best (with Struct. Align). A dialog box asks you to specify the reference protein and the
protein to align with it. Use the pop-up menus to specify the template as reference, and the
unknown protein to align. Click OK. (You will see a warning that you are aligning to a model-in-
progress, which is never done while building a new model, but it's OK for this comparison.) Then
realign all the model sequences with Edit: Generate Structural Alignment. Now the new model and
the template are superimposed.
To assess potential ligand binding by your new model, you must merge the ligand moleule of the
template with the model to produce a new combined model, so that protein and NAG models are
part of the same layer. Here's how:
1. Make the unknown protein layer active, and Select: All.
2. Make template active. Select: All and then Select: Inverse Selection. The result is that no
groups are selected in template.
3. Make template active. Select only the ligand at the bottom of the control panel.
4. Check the Sel column of the Layer Infos window to confirm that 131 groups of the unknown
protein, and only the ligand group(s) of template are selected.
5. Edit: Create Merged Layer From Selection.
A new layer named _merge_ appears. In the cyc column of Layer Infos, check only the template
and _merge_. Blink to compare them.
To complete this assignment, save the _merge_ layer as modeln.pdb.