Abstract
Human leukocyte antigen (HLA) recognizes antigenic fragments and presents them to T cells. HLA is polymorphic. There are over 2000 different HLA alleles at present and the number is constantly increasing. However, antigen binding studies are limited to a small proportion of these alleles; the binding specificities of most alleles are unknown. Several research groups have attempted to partition different HLA alleles into groups. In this chapter previous classifications are reviewed and we present two chemometric approaches to classifying class I HLA alleles. The program GRID is used to calculate interaction energy between protein molecules and defined chemical probes. These interaction energy values are imported into another program GOLPE and used for principal component analysis (PCA) calculation, which groups HLA alleles into supertypes. Amino acids that are involved in the classification are displayed in the loading plots of the PCA model. Another method, hierarchical clustering based on comparative molecular similarity indices (CoMSIA) is also applied to classify HLA alleles and the results are compared with those of the PCA models.