DYNDOM (CCP4: Supported Program)

NAME

dyndom - determine dynamic domains when two conformations are available

SYNOPSIS

dyndom command_file

DESCRIPTION

For complete and up-to-date information, see the DynDom Home Page.

DynDom is a program that determines protein domains, hinge axes and amino acid residues involved in the hinge bending. It is fully automated.

You can use DynDom if you have two conformations of the same protein. These may be two X-ray structures, or structures generated using simulation techniques such as molecular dynamics or normal mode analysis.

The application of DynDom provides a view of the conformational change that is easily understood. The conformational change may be quite complicated in detail, but by using DynDom you can visualize it as involving the movement of domains as quasi-rigid bodies. The analysis of a conformational change in terms of domain movements only makes sense if the interdomain deformation is at least comparable to the intradomain deformation. You can use DynDom to access this, but the results could be misleading if this is not the case.

INPUT AND OUTPUT FILES

Input

command_file: The name of the command file should be given as the sole argument to dyndom. This file specifies the PDB file names, chain identifiers, if they exist, and the control parameter settings (see below and examples).
Two PDB files: The names of the two PDB files containing the two conformations should be given in the command file.

Output

foo_info: A file that gives all the relevant information on the conformational change in terms of domain motions.
foo_arrow: An "arrow file" that gives the coordinates of atoms in PDB format that create an arrow to represent the interdomain screw axis for display with RasMol. This file is appended to the PDB file of the first conformer.
foo_rotvecs: A rotation vector file in PDB format for display using RasMol. Each "atom" is located at a point whose coordinates are the components of the rotation vector respresenting the rotation of a particular backbone segment. The number of atom gives the number of the residue at the centre of the backbone segment.
foo_rasscript: A RasMol script file for display of the protein, its domains, its interdomain screw axes and residues involved in the interdomain motion. It is also used to display the rotation vectors.
foo_shscript: A short C-shell script file for convenient display of the protein or rotation vectors using RasMol.

To display the rotation vectors, simply enter the command:


   foo_shscript rotvecs

To display the protein to view the domains etc, simply enter the command:


   foo_shscript protein

Control Parameters

clusters

This sets the maximum number of clusters of rotation vectors the program looks for. This can be set high for the reason described in the previous section. Default=20

iterations

This sets the maximum number of iterations the clustering algorithm is allowed to determine the clusters. It normally requires far less than 100 iterations. Default=100

window

A sliding window is used to generate backbone segments whose rotation vectors are calculated for the clustering algorithm. The longer the window the better local intrasegment rotations, which may have nothing to do with the domain motion, are eliminated. A shorter window, however, is preferable if one wants to accurately identify those residues involved in the interdomain bending. In any case the length of the window should be short compared to the minimum domain size. The length of the window is specified by the number of residues. The length of the window must be odd as the central residue must be defined. Default=5

domain

This sets the minimum domain size in number of residues. It depends on your feeling on how big a domain should be, to be called a domain. It can be set as small as you like, but it wouldn't be logical to make it smaller than the window length. Default=10% of total number of residues

ratio

This value gives the minimum value for the ratio of interdomain displacement to intradomain displacement for a domain pair. For a precise definition of this value see the DynDom main reference. There is no clear cutoff for this value, but the lower it is for a domain pair, the less sensible it is to analyse their motion in terms of an interdomain motion. The program calculates this value for every prospective domain pair and outputs it to the screen. If you set the minimum to a value much less than 1.0 you may end up analyzing meaningless results. Default=1.0

Description of the Basic Methodology

Determination of Dynamic Domains

The program first determines the "dynamic domains." First, a whole protein best fit of the two conformations is made. Then, rotation vectors of residues or short main-chain segments are determined. A clustering algorithm is then used to identify clusters of rotation vectors. Groups of residues forming these clusters form possible dynamic domains.

Determination Hinge Axes

Groups of residues are only accepted for the analysis of hinge axes if they satisfy a criterion based on the ratio of the interdomain displacement to intradomain displacement with another group of residues with which there exists a physical connection. If this is the case the two groups of residues form dynamic domains and their interdomain motion is meaningful. The axes determined are in fact interdomain screw axes. This is based on the theorem of Chasles which states that the general displacement of a rigid body is a screw motion. The location of the interdomain screw axis tells us something about the kind of motion allowed by the interdomain connections. It is possible for the interdomain screw axis to be located far away from the interdomain connections if they are very flexible. Only if the interdomain screw axis is located near to those residues involved in the interdomain bending (defined below) can we think of the axis as a hinge axis. In such a case we call the axis an, "effective hinge axis" and the residues are said to be acting as "mechanical hinges."

Determination of Residues Involved in Interdomain Bending

If one domain is fixed in space with the other rotating, then one will see a rotational transition in the connecting region between the two domains. One can define the residues involved in the interdomain bending to be those at the interdomain boundaries, as found by the clustering algorithm, plus those neighbouring residues whose rotations deviate at least one standard deviation from the average of the domain to which they belong.

Closure and Twist Motion

Axes can be classified into two extrme types: those parallel to the line joining the centres of mass of a pair of domains, and those perpendicular to this line. The former are called twist axes, the latter, closure axes. Any axis can be decomposed into components parallel or perpendicular to this line and a percentage measure of the degree of closure motion can be defined from the square of the projection on the closure axis.

How DynDom Works

DynDom uses the K-means clustering algorithm to find clusters of rotation vectors. The number of clusters found is specified by the user, but should be set high as DynDom automatically finds the largest number of clusters for which one may reasonably regard the conformational change in terms of domain motions.

A cluster in rotation space may not correspond to a cluster in real space, but rather a fragmented region. Such a fragmented region one would not normally call a domain. DynDom splits up any clusters that do not correspond to heavy atoms connected through a network of distances of 3.5 angstrom or less, into domains. Any domain that comprises fewer residues than the minimum domain size set by the user is not included in any further analysis. If all the domains from any single cluster are smaller than the minimum domain size, the program stops.

For every domain larger than the minimum size, the program checks which are connected directly through the backbone (not through another domain), and calculates the ratio of interdomain displacement to intradomain displacement for every connected domain pair. If any domain pair has a ratio less than the minimum set by the user, then the domains at this level of clustering are not analysed for interdomain screw axes, etc. The program goes on (providing the maximum number of clusters set by the user is not exceeded) to find the domains corresponding to the next level of clustering, and so on. The program finds the largest number of clusters for which all connected domain pairs have a ratio for their interdomain displacement to intradomain displacement that is larger than or equal to the minimum specfied value. It is these domains that are analysed for the interdomain screw axes, etc.

EXAMPLES

Example script for adenylate kinase can be found in $CEXAM/unix/runnable/

dyndom.exam (Runnable unix script.)
adenylate.command (Command file.)

AUTHOR

Steve Hayward

REFERENCES

Method:
S.Hayward, A.Kitao, H.J.C.Berendsen, Model-Free Methods of Analyzing Domain Motions in Proteins from Simulation: A Comparison of Normal Mode Analysis and Molecular Dynamics Simulation of Lysozyme Proteins, Structure, Function and Genetics, 27, 425, 1997.
DynDom main reference:
S.Hayward, H.J.C.Berendsen, Systematic Analysis of Domain Motions in Proteins from Conformational Change; New Results on Citrate Synthase and T4 Lysozyme Proteins, Structure, Function and Genetics, 30, 144, 1998.
Applications:
B.L.de Groot, S.Hayward, D.van Aalten, A.Amadei, H.J.C.Berendsen, Domain Motions in Bacteriophage T4 Lysozyme; a Comparison between Molecular Dynamics and Crystallographic Data Proteins, Structure, Function and Genetics, 31, 116, 1998.