Recent advances in DNA-based typing have led to the detection of a continuously growing number of HLA alleles. For this reason, HLA matching in transplantation of hematopoietic stem cells from unrelated donors has become increasingly complicated. When there is no genotypically identical sibling and there are several alternative potential donors that all have a mismatch at a relevant HLA locus, until now no rating system has existed indicating different levels of allogenicity. In order to find a theoretical approach to this problem we propose a rating system ('dissimilarity index') based on structural data of HLA class I molecules, and on published data about frequencies of naturally occurring amino acid exchanges. For demonstration we employ our rating system for the comparison of the HLA-A*23 and A*24 groups, both of which allelic products are subdivisions of the serological HLA-A9 family. Remarkable differences between the subtypes were revealed, which were superior to a simple sequence comparison. More surprisingly, it was uncovered that some alleles of the A*24 group showed fewer differences to A*2301 than to alleles within their own subtype group. Sequence similarity matching may serve as a starting point for the clinical evaluation of acceptable mismatches within the HLA-A9 family and serve as a model for other HLA class I groups.