Next Previous Contents

2. Theoretical basis

A detailed description of the theoretical methods used in DomainFinder can be found in Ref. 1. This section provides a brief summary that everyone should read before using DomainFinder.

Since dynamical domains are by definition large regions in a protein, DomainFinder does not look at all atoms, but only at the C-alpha atoms, which define the backbone. Specifically, DomainFinder looks at changes in the relative positions of C-alpha atoms; these changes describe the deformation of the protein while it undergoes a specific motion. To characterize the deformation of each part of the protein, DomainFinder calculates a deformation energy for each atom, which depends on the changes in the distance of this atom from each of its close neighbors. A low deformation energy indicates relatively rigid regions, which are candidates for dynamical domains, whereas high deformation energies indicate flexible regions. Dynamical domains are identified by grouping the rigid regions according to their overall motion; a dynamical domain is thus by definition a sufficiently rigid region whose parts move in a sufficiently similar way. This procedure requires two input parameters that the user must provide: a deformation threshold, up to which a region is considered sufficiently rigid, and a similarity threshold, up to which the motion of subregions is considered sufficiently similar.

In order to analyze protein motions, DomainFinder needs some input data describing them. One possibility is the use of two different conformations; the difference between these conformations describes one possible motion. Typically these conformations are obtained experimentally (by crystallography or NMR), and differ by the presence or absence of certain ligands. This mode of operation amounts to an analysis of experimental data.

However, there are only few proteins for which two substantially different conformations are known experimentally. DomainFinder can also use calculational methods to obtain the necessary information about domain motions, using only a single conformation as input. The technique used by DomainFinder is a variant of normal mode analysis which has been optimized for the purpose of domain motion study. It is described in detail in Ref. 2, the updated form of the interactions used in DomainFinder 2.0 is described in Ref. 3 Even if multiple experimental structures are available, the normal mode approach is in general preferable, because it yields clearer domain delimitations and more detail. This is not surprising, since two conformations necessarily contain only a small subset of the full protein dynamics.


Next Previous Contents