Sequence and structural analysis of antibodies.

The work presented in this thesis focusses on the sequence and structural analysis of antibodies and has fallen into three main areas. First I developed a method to assess how typical an antibody sequence is of the expressed human antibody repertoire. My hypothesis was that the more \humanlike" an antibody sequence is (in other words how typical it is of the expressed human repertoire), the less likely it is to elicit an immune response when used in vivo in humans. In practice, I found that, while the most and least-human sequences generated the lowest and highest anti-antibody reponses in the small available dataset, there was little correlation in between these extremes. Second, I examined the distribution of the packing angles between VH and VL domains of antibodies and whether residues in the interface in uence the packing angle angle. This is an important factor which has essentially been ignored in modelling antibody structures since the packing angle can have a signicant eect on the topography of the combining site. Finding out which interface residues have the greatest in uence is also important in protocols for `humanizing' mouse antibodies to make them more suitable for use in therapy in humans. Third, I developed a method to apply standard Kabat or Chothia numbering schemes to an antibody sequence automatically. In brief, the method uses proles to identify the ends of the framework regions and then lls in the numbers for each section. Benchmarking the performance of this algorithm against annotations in the Kabat database highlighted several errors in the manual annotations in the Kabat database. Based on structural analysis of insertions and deletions in the framework regions of antibodies, I have extended the Chothia numbering scheme to identify the structurally correct positions of insertions and deletions in the framework regions.