Jan T. Pedersen

Antibody modeling: Beyond homology

The immune system is capable of producing at least 109 different antibody specificities. There are currently about 800 variable-region sequences known, whereas there are only 16 deposited Fab structures. Thus, the rate of sequence acquisition far outstrips the rate of structure determination. While for many protein families the presence of 16 highly homologous structures would enable accurate modeling of new family members by homology, the hypervariability of combining site sequences and structures precludes this. To solve this problem we have developed an algorithm (CAMAL) that exploits the knowledge base of protein structures but adds to it the power of conformational search and the subtlety of energy screening. The accuracy that this algorithm is able to achieve leads to the possibility that, with further development, extensive use of X-ray crystallography for antibody structure determination will become unnecessary.

Antibody-combining sites. Extending the natural limits.

The antibody repertoire is very large with at least 109 different antibody specificities, yet there are currently only 800 variable-region sequences known and < 23 Fab structures deposited with the Brookhaven Protein Data Bank. To engineer the antibody-combining site rationally, we need to define the rules that govern antibody structure. To understand the process of antibody-antigen recognition, we need not only to predict complementary determining regions accurately, but to simulate accurately the interaction of antibody with antigen. We have made progress in the modeling of antibody-combining sites and in the simulation of antibody complex formation. The combination of these approaches will allow us to extend the natural limits of antibody-combining sites in a more rational manner.