|advertisement: compare things at compare-stuff.com!|
The data set of antibody structures used in this study includes all the complete lightheavy chain dimers available in the Brookhaven Protein Data Bank (PDB)[Bernstein et al., 1977] January 1995 release and pre-releases up to July 1995, with three exceptions. Antibody 26/9 (PDB code 1frg[Churchill et al., 1994]) has been excluded since it is very similar (91% and 88% identity in the light and heavy chains respectively) to 17/9 (1him[Schulzegahmen et al., 1993]) and has the same antigen. Also excluded are the pre-release structures for antibodies F9.13.7 (1fbi[Lescar et al., 1995]) and JEL103 (1mrc-f[Pokkuluri et al., 1994]) which have large numbers of missing atoms. Where more than one structure was available for a given antibody, the highest resolution structure was chosen. All water molecules have been removed from the coordinate files. The data set is presented in Tables 2.1 and 2.2.
All sequences are numbered according to Chothia et al.chothia:canon using an alignment with a consensus sequence and the program kabatnum by Andrew Martin. This is not in accordance with the scheme used by Chothia et al. in their later papers[Chothia et al., 1989, and more recent papers] where they place the insertion at L31. However, examination of the CDR-L1 conformations shows that L30 is the correct site for the insertion.
Three antigen size classes (Table 2.1 column 6) have been introduced in order to simplify the text. Although the classes were defined by simple cluster analysis of the antibody-antigen interface areas, they correlate with the antigen types as follows: small (haptens); medium (peptides, carbohydrates, nucleic acids); large (proteins, cyclosporin A).