| advertisement: compare things at compare-stuff.com! |
The antibody repertoire is capable of recognising an almost infinite number
of previously unencountered molecules, from small organic compounds to
large macromolecular complexes. The structural basis for this recognition
is surprisingly invariant; random genomic splicing of the light and heavy
chain genes creates variability in sequence and length in just six loops
(complementarity determining regions or CDRs) supported on a structurally
highly conserved 
-sheet framework. Improvements in binding affinity
are achieved through somatic hypermutation of the CDRs and clonal expansion
on exposure to antigen. The CDR residues number only 
70 of the
total 230 residues of the Fv antibody fragment (the smallest
fragment able to bind antigen). Furthermore, the CDRs generally adopt only
a limited number of canonical backbone
conformations[Chothia et al.,
1989], determined by their loop length and
certain key, `structurally determining' residues. Thus antibodies provide
a system for the study of molecular recognition by proteins where
structural variability is conveniently restrained, but where many examples
of different inter-molecular interactions are available.
Since the first complexed antibody-antigen crystal structures were solved, numerous studies[Rees et al., 1994,Webster et al., 1994,Wilson & Stanfield, 1993,Wilson & Stanfield, 1994, for reviews] of molecular recognition have focussed upon them. There are now (July 1995) 26 different complexed Fab and Fv structures in the Brookhaven Protein Data Bank[Bernstein et al., 1977] and another 19 in uncomplexed form. In a recent review, Wilson and Stanfieldwilson:abag2 have presented the variable contributions of each CDR to the antigen-buried surface, whilst Padlan et al.padlan:contacts have published a by-residue summary of antigen contacts with an emphasis towards sequence variability and humanization.
A number of groups have explored how detailed atomic interactions give specificity and affinity to antibody-antigen binding[Arevalo et al., 1994,Wong et al., 1995, for example]. However there is also some interest in the broad-scale assembly of the combining site. Webster et al.webster:interactions have suggested a subjective classification of antibody combining sites by general surface topography and antigen type. The objective computational analysis of combining site shape has so far been limited to just 5 light chain surfaces[Gerstein, 1992].
In this study we perform a detailed residue-level analysis of antigen contacts and investigate the contacts made by different sized antigens and different canonical loops. In addition we have developed an automated method for classifying surface topographies and support the suggestion that the general topography of antibody combining sites is correlated with antigen type. Finally, the application of our findings to the prediction of antigen contacting residues in uncomplexed antibody structures is discussed.