Using HLA binding prediction algorithms for epitope mapping in HIV vaccine clinical trials

Current HIV vaccines are designed to elicit both T-cell and B-cell responses. A common endpoint in any T-cell based vaccine trial are measurements of vaccine-induced T-cell responses such as breadth and magnitude [7]. In order to measure such endpoints blood samples are collected at multiple timepoints. Current immunological assays for measuring T-cell responses are functional assays in which peripheral blood mononuclear cells (PBMC) is incubated with target peptide(s) and then the release of various cytokines such as IFN-? are measured. The major limiting factor in these mappping studies is sample availability, as each of these tests requires an order of 100K live cells. Therefore current mapping strategies use a group-testing approach in which responses to the immunogen are first measured using peptide pools that span a full protein, and are then further refined using sets of mini-pool and finally a peptide matrix [23]. In this paper we explore the idea of using HLA binding predictors to improve the efficiency of epitope mapping protocols in vaccine trials. Given information about participant's HLA alleles, we attempt to predict vaccine induced T-cell responses at various levels of refinement, based on the current group-testing hierarchical mapping approach. Using extensive epitope mapping data from a cohort of 12 acutely infected HIV infected individuals, we show that using state-of-the-art HLA binding predictors, significant improvements in mapping efficiency can be obtained.