Definition of MHC supertypes through clustering of MHC peptide binding repertoires

Abstract

MHC molecules, also known in the human as human leukocyte antigens (HLA), display peptides on antigen presenting cell surfaces for subsequent T cell recognition. Identification of these antigenic peptides is especially important for developing peptide-based vaccines. Consequently experimental and computational approaches have been developed for their identification. A major impediment to such an approach is the extreme polymorphism of HLA, which is in fact the basis for differential peptide binding. This problem can be mitigated by the observation that despite such polymorphisms, HLA molecules bind overlapping set of peptides, and therefore, may be grouped accordingly into supertypes. Here we describe a method of grouping HLA alleles into supertypes based on analysis and subsequent clustering of their peptide binding repertoires. Combining this method with the known allele and haplotype gene frequencies of HLA I molecules for five major American ethnic groups (Black, Caucasian, Hispanic, Native American, and Asian), it is now feasible to identify supertypic combinations for prediction of antigenic peptide, offering the potential to generate peptide-vaccines with a population coverage ≥95%, regardless of ethnicity. One combination including five distinct supertypes is available online at our PEPVAC web server (http://immunax.dfci.harvard.edu/ PEPVAC/). Promiscuous peptides predicted to bind to these five supertypes represent around 5% of all possible peptide binders from a given genome. © Springer-Verlag 2004.