A major challenge in unrelated hematopoietic stem cell transplantation (HSCT) is the prediction of permissive HLA mismatches, ie, those associated with lower clinical risks compared to their nonpermissive counterparts. For HLA-DPB1, a clinically prognostic model has been shown to be matching for T cell epitope (TCE) groups assigned by cross reactivity of T cells alloreactive to HLA-DPB1*09:01; however, the molecular basis of this observation is not fully understood. Here, we have mutated amino acids (aa) in 10 positions of HLA-DPB1*09:01 to other naturally occurring variants, expressed them by lentiviral vectors in B cell lines, and quantitatively measured allorecognition by 17 CD4(+) T cell effectors from 6 unrelated individuals. A significant impact on the median alloresponse was observed for peptide contact positions 9, 11, 35, 55, 69, 76, and 84, but not for positions 8, 56, and 57 pointing away from the groove. A score for the "functional distance" (FD) from HLA-DPB1*09:01 was defined as the sum of the median impact of polymorphic aa in a given HLA-DPB1 allele on T cell alloreactivity. Established TCE group assignment of 23 alleles correlated with FD scores of <= 0.5, 0.6 to 1.9 and >= 2 for TCE groups 1, 2, and 3, respectively. Based on this, prediction of TCE group assignment will be possible for any given HLA-DPB1 allele, including currently 367 alleles encoding distinct proteins for which T cell cross reactivity patterns are unknown. Experimental confirmation of the in silico TCE group classification was successfully performed for 7 of 7 of these alleles. Our findings have practical implications for the applicability of TCE group matching in unrelated HSCT and provide new insights into the molecular mechanisms underlying this model. The innovative concept of FD opens new potential avenues for risk prediction in unrelated HSCT. (C) 2015 American Society for Blood and Marrow Transplantation. OI Maiers, Martin/0000-0002-0198-2064