Urban Christine, Schmidt Alexander H, Hofmann Jan Andreas
DKMS, Tübingen, Germany.
Front Med (Lausanne). 2020 Feb 18;7:32. doi: 10.3389/fmed.2020.00032. eCollection 2020.
In the setting of hematopoietic stem cell transplantation, donor-patient HLA matching is the prime donor selection criterion. Matching algorithms provide ordered lists of donors where the probability of a donor to be an HLA match is calculated in cases where either donor or patient HLA typing information is ambiguous or incomplete. While providing important information for the selection of suitable donors, these algorithms are computationally demanding and often need several minutes up to hours to generate search results. Here, we present a new search kernel implementation for Hap-E Search, the haplotype frequency-based matching algorithm of DKMS. The updated search kernel uses pre-calculated information on donor genotypes to speed up the search process. The new algorithm reliably provides search results in <1 min for a large donor database (>9 Mio donors) including matching and mismatching donors, even for frequent or incomplete patient HLA data where the matching list contains several thousand donors. In these cases, the search process is accelerated by factors of 10 and more compared to the old Hap-E Search implementation. The predicted matching probabilities of the new algorithm were validated with data from verification typing requests of 67,550 donor-patient pairs.
在造血干细胞移植的背景下，供体与患者的HLA配型是主要的供体选择标准。匹配算法提供供体的有序列表，在供体或患者的HLA分型信息不明确或不完整的情况下，计算供体成为HLA匹配的概率。虽然这些算法为选择合适的供体提供了重要信息，但它们对计算要求很高，通常需要几分钟到几小时才能生成搜索结果。在此，我们展示了一种用于Hap-E Search的新搜索内核实现方式，Hap-E Search是DKMS基于单倍型频率的匹配算法。更新后的搜索内核使用预先计算的供体基因型信息来加速搜索过程。新算法能够在不到1分钟的时间内，为包含匹配和不匹配供体的大型供体数据库（超过900万个供体）可靠地提供搜索结果，即使对于匹配列表包含数千个供体的常见或不完整的患者HLA数据也是如此。在这些情况下，与旧的Hap-E Search实现方式相比，搜索过程加速了10倍甚至更多。新算法的预测匹配概率通过67550对供体-患者对的验证分型请求数据进行了验证。
