In fully HLA-matched allogeneic hematopoietic cell transplantation (HCT), the main mechanism of the beneficial graft-versus-tumor (GVT) effect and of detrimental graft-versus-host disease (GVHD) is believed to be caused by donor cytotoxic T cells directed against disparate recipient minor histocompatibility antigens (miHAs). The most common origin of disparate miHAs is nonsynonymous single nucleotide polymorphism (nsSNP) differences between donors and patients. To date, only some 30 miHAs have been identified and registered, but considering the many different HLA types in the human population, as well as all the possible nsSNP differences between any 2 individuals, it is likely that many miHAs have yet to be discovered. The objective of the current study was to predict novel HLA-A- and HLA-B-restricted miHAs in a cohort of patients treated with nonmyeloablative conditioning allogeneic HCT (matched related donor, n = 70; matched unrelated donor, n = 56) for a hematologic malignancy. Initially, the cohort was genotyped for 53 nsSNPs in 11 known miHA source proteins. Twenty-three nsSNPs within 6 miHA source proteins showed variation in the graft-versus-host (GVH) direction. No correlation between the number of disparate nsSNPs and clinical outcome was seen. Next, miHAs in the GVH direction were predicted for each patient-donor pair. Using the NetMHCpan predictor, we identified peptides encompassing an nsSNP variant uniquely expressed by the patient and with predicted binding to any of the HLA-A or -B molecules expressed by the patient and donor. Patients with more than the median of 3 predicted miHAs had a significantly lower 5-year overall survival (42% vs 70%, P = .0060; adjusted hazard ratio [HR], 2.6, P = .0047) and significantly higher treatment-related mortality (39% vs 10%, P = .0094; adjusted HR, 4.6, P = .0038). No association between the number of predicted miHAs and any other clinical outcome parameters was observed. Collectively, our data suggest that the clinical outcome of HCT is affected not by disparate nsSNPs per se, but rather by the HLA-restricted presentation and recognition of peptides encompassing these. Our data also suggest that 6 of the 11 proteins included in the current study could contain more miHAs yet to be identified, and that the presence of multiple miHAs confers a higher risk of mortality after nonmyeloablative conditioning HCT. Furthermore, our data suggest a possible role for in silico based miHA predictions in donor selection as well as in selecting candidate miHAs for further evaluation in in vitro and in vivo experiments.