The Homology Module allows you to build a 3D model of a protein whose 3D structure has not been experimentally determined IF you already know the structure of one or more homologous protein(s).

The protein(s) whose structure(s) is/are aleady known are referred to as the "reference" or "real" protein(s). The protein whose structure you are trying to build is called the "model", "unknown", or "sequence" protein.

The steps involved in this process are:

1. Obtain the sequences of both the unknown and the reference proteins.
2. Align these sequences using your favorite alignment tools.
3. Obtain a structure file for each reference protein
4. Read the sequences and structures into InsightII
5. Identify which sequence corresponds to which structure
6. Find structurally conserved regions (SCRs) in the reference proteins (only possible if there is more than one reference protein)
7. Copy the coordinates of the conserved regions in one of the reference proteins to the model protein.
8. Propose structures for the loops or variable regions (VRs) between the SCRs.
9. Make corrections to sidechain conformations as appropriate.
10. Refine the structure my energy minimization and/or molecular dynamics.

Unknown protein:
>976347 Human zinc finger homeodomain protein (Res 724-750)
KPFRCEVCNYSTTTKGNLSIHMQSDKH

Reference protein:
>3znf.pdb Human enhancer binding protein zinc finger
RPYHCSYCNFSFKTKGNLTKHMKSKAHSKK
(minimized average NMR structure)

Note 1: This is a relatively simple case because:

1. Zinc fingers are small and very highly conserved domains. In our example 14/27 residues are identical and there are no gaps in the alignment although the unknown protein is slightly shorter than the reference protein.
2. The alignment is obvious.
3. We will only use a single reference protein.

Step 1: Getting Started

1. The Homology module is currently licensed only on chemvgx and splatter. Therefore, when working on other machines it is necessary to run the program remotely on chemvgx or splatter, then display on your local monitor. Half of the class should choose each machine. e.g. To run remotely on chemvgx,
   xhost chemvgx
   telnet chemvgx (then login)
   cd to appropriate directory
   insightII
   
2. Choose the Homology module and take a few minutes to browse through the menu options

Step 2: Reading in and Aligning Sequences

The sequence alignment and pdb files needed in this tutorial are in the directory /ruser/instruct1/stone/C687/homology.

1. molecule-get
   Get 3znf.pdb
   The name of the molecule you get is ZNF
2. Sequences-get
   Choose alignment and get zf.align
   A sequence window should appear containing two sequences
3. Sequences-copy
   Copy from: ZNF1
   Copy to: ZNF
   This generates a new sequence which is identical to the ZNF1 sequence but is understood by the program to be the sequence of the displayed protein ZNF whose structure is known.
4. Sequences-delete
   Delete ZNF1 (which is no longer useful)
5. There are many different manipulations you can do with the sequences. One of the simplest is to align two sequences automatically within Insight.
   Alignment-Pairwise_sequence-Automatic
   Specifiy the 2 sequences and execute
   Note that they are now perfectly aligned within the sequence window.
6. Now spend the next ~15 minutes experimenting with the various options you have to manipulate the sequence. Most of these are mouse driven options. They are described in detail in the help menu invoked within the sequences window. They basically fall into 2 categories distinguished by specifying the mode in the sequences window:
   
   • Ways of moving a sequence or insert, moving, or deleting gaps in a sequence.
   • Ways of creating, changing, moving, and deleting boxes.
   Boxes are used to specify regions of 2 sequences that are aligned and regions of a sequence that should be mapped onto a particular region of a known structure. Boxes cannot contain gaps!

   This section is going to be particularly important for people who are doing any homology modeling in their projects.

7. Once you have become reasonably comfortable with the mouse driven options, go back to where you were after point 5 above. You can do this either by deleting boxes and realigning the sequences or just by deleting everything and starting over.

Step 3: Assigning Coordinates fo the Unknown Protein

1. Draw a box around the region of the 2 sequences that is perfectly aligned. Under boxes-freeze specify the box number (boxes are numbered starting from 0). Freezing prevents the box from being changed and is necessary before you assign coordinates. Frozen boxes are colored red.
2. Sequences-AssignCoords
   Give box number and choose bump check (which will look for steric violations as the new structure is generated). Spend some time looking at the new structure and comparing it to the old structure.

   Note: When you assign the coordinates for proteins that are not perfectly aligned, the process is somewhat more complicated. Basically it involves:

   • Creating and freezing boxes for EACH region of conservation without any gaps. Where there is a gap, the boxes should end at least two residues away from that gap so that the residues that span the gap in the model structure are not too harshly constrained.
   • Assigning coordinates for EACH boxed region.
   • Modeling the strctures of the "loop" or non-conserved regions that span the gaps. This can be done by; (a) searching the PDB for conformations that match the regions surrounding and including the loop regions then copying their conformations; or (b) alternatively by generating a number of possibilities de novo and then "manually" sorting through them for those that seem most sensible. This whole process is tedious and inherently biased but obviously necessary. Most of the uncertainty in your final structure will correspond to these regions and they are often the most interesting regions biochemically. If you plan on doing any homology modeling during your project or in the future, you should take the time now to learn about these methods. The best way is to run throught the on-line Homology pilot tutorial. Also check out the Homology manual.

Step 4: Fixing the Geometry

1. Comparison of the the unknown and reference protein structures. Take 10 minutes to look at various details of the two structures which should be superimposed on the screen and observe the similarities and differences.

   In particular, look at the confomations of the backbone, the sidechains of residues that are identical in both proteins, the sidechains of residues that are similar in the two proteins, and the sidechains of residues that are quite different in the two proteins. Think about how the program has assigned conformations to each of these parts of the unknow protein.

2. In regions where the structure of the unknown protein does not seem to be optimal, we can change it either by changing specific dihedral angles or by global energy minimization.
3. First try changing rotamers of dihedral angles. Experiment with the Residue-Manual_Rotamer and Residue-Auto_Rotamer options.
4. Then try doing an energy minimization. You should save your work now, then log out of the remote CPU and start up Insight on your local computer to do this exercise. Otherwise it will take forever!

Currently, the name of the moleucle starts with a "$" character. You must change the name of the molecule before proceeding: Click on Object/Rename, an drename the molecule.

Under Biopolymer, choose a forcefield and fix potentials and charges.

Under Discover

Constraint-Fix

to fix the backbone conformations of all resiudes and the sidechain conformations of 5, 8, 21, and 27.

Parameters-Minimize

Steepest descent algorithm
2000 iterations
Derivative of 1.0

Run

run
Local host
Interactive