advertisement: compare things at compare-stuff.com!

Discussion

Antibodies have been used in biomedical research for many years and have started to be used in applied medicine. They have potential roles in fields as diverse as cancer therapy[Chester & Hawkins, 1995], biosensors[Killard et al., 1995] and catalysis[Hilvert, 1994]. They are also important molecules in autoimmune diseases such as rheumatoid arthritis and systemic lupus erythematosus[Brinkman et al., 1990]. Research and medical applications of antibodies have been made possible by recent advances in library-based in vitro production of monoclonal antibodies[Winter et al., 1994], computer based structure prediction from primary sequence[Chothia et al., 1989,Martin et al., 1991], and the availability of an increasing number of experimentally determined three-dimensional antibody structures.

Analysis of the antibody combining site has four ultimate goals:

1. The ability to tailor the combining site to modify its binding or to introduce novel function,
2. de novo design of antibodies to target antigens,
3. Identification of antigens to which antibodies bind,
4. The ability to dock antibody and antigen.

Together with methods for modelling the antibody combining site, our analysis of combining site shape and contacting residues is applicable to all these areas.

The position and orientation of a residue relative to the centre of the combining site are the major determinants of its propensity to bind antigen. Our analysis of contact residues allows tailoring and design via mutagenesis experiments to concentrate on those residues which are most likely to interact with antigen. Correlations of CDR amino-acid composition with antigen type are much stronger when the analysis is restricted to common contacting residues rather than the complete hypervariable surface (unpublished results).

As we have shown, contact residue analysis also allows us to identify the region of the antigen combining site most likely to interact with antigen and therefore restricts the area of the molecule which must be considered in docking experiments. In addition, mean burial data, $\overline{\rho}$, could be used as part of a potential function to score models in an antibody-antigen docking algorithm.

It has long been proposed that antibodies, binding to protein antigens, are characterised by a flat combining site, while those binding to peptide and DNA antigens have a groove-like combining site and hapten-binders a cavity[Webster et al., 1994].

Our analysis of the shape of the combining site has shown that these broad shape classes of antibody combining sites can be identified using a fractal measure. However, the correlation with antigen type is not exact. In antibodies whose crystal structures are available, both complexed with antigen and uncomplexed, there is evidence for conformational change[Stanfield et al., 1993]. Our analysis has shown that whilst changes in the gross topography do occur, they are not of great magnitude (i.e. never from cavity to planar or vice versa). Therefore the shape analysis, used cautiously, should prove useful in the design of antibody combining sites for specific antigen targets and in the identification of potential antigen types for autoimmune antibodies.

In summary, we have shown that antibody-antigen interactions, although random and combinatorial in the origin of antibody diversity, are not completely chaotic. Antigens tend to bind to the antibody residues located at the centre of the combining site where the six CDRs meet and we have provided a `contact definition' of the CDRs which may prove useful in antibody design and mutagenesis. Our detailed analysis of combining site shape has largely confirmed the proposed correlation between shape and antigen type although we have identified a number of noteworthy outliers. Thus our analysis adds to the understanding of antibody-antigen interactions and the mechanisms by which antibodies recognise their targets.