The following links list how search determinants are assigned to each HLA allele by the different registries; The Anthony Nolan, the NMDP, BMDW, Canadian Unrelated Bone Marrow Donor Registry, France Greffe de Moelle, BBMR and also the ZKRD approach to search determinants.

Zentrales Knochenmarkspender-Register Deutschland (ZKRD)

ZKRD's Approach to Search Determinants

The term “search determinant” (SD) has not been unequivocally defined so far. There are several potential applications for SDs

  1. plausibility checking between the serological and DNA assignment within a given HLA type
  2. selecting potential matches between donors and a patient where one has been typed by serology and the other one by DNA methodology
  3. speeding up the match process by using “search determinants” as keys values in conjunction with a database and an matching algorithm

We think that for 1 and 2 the serology dictionary (with possible extensions; see below) should be sufficient. Application 3 is interesting and the only one implemented at ZKRD. Essentially, here the description of SDs alone is of little help since they only can be interpreted in the context of the matching program.

Assignment of Search Determinants

ZKRD uses an algorithmic approach to determine search determinants for HLA alleles. This algorithm is based on the existing relationships between antigens (1) and uses the official allele/antigen correspondence file (2) from the WMDA reference website (3) as primary data source.

The existing relationships between antigens according to the existing relationships between antigens (1) can be visualized as so-called rooted trees. Every broad specificity is the root node of such a tree and the related split/associated antigen are inner nodes or leaf nodes of the appropriate tree. Every inner node is the root node of a subtree.

Examples: B16 - tree


  • B16 is the root node
  • B39 is an inner node with child nodes B3901 and B3902
  • B38, B3901 and B3902 are leaf nodes
Other examples / structures of antigen-relationship trees:



Every antigen belongs exactly to one tree. There are degenerated tree structures, e.g. trees with one single node (e.g. A3 tree) or two nodes (e.g. B7 tree).

Algorithm for the computation of search determinants:

  • Given:  Allele A
  • Goal:   Possible antigens (SDs) of allele A
Algorithmic steps:
  • Get the corresponding antigen(s) S for allele A from file (2).
    The special value "?“ is returned if no defined serology is available for the given allele A (e.g. B*8202 or B*8301). In this case the steps 2-4 have to be omitted and the first two digits of the allele code are used as SD.
    The special value "0“ is returned if the allele A is non or low expressed. In this case the steps 2-4 have to be omitted and the value "0“ is used as SD.
    For all other returned values S repeat steps 2-4.
  • Locate the appropriate relationship tree containing the node S.
  • Locate the node S within the relationship tree.
  • The following antigens are possible for allele A:
    • S itself
    • all nodes along the path from S to the root node (may not exist)
    • all child nodes of the subtree with root node S (may not exist)

B*4011   =>   B40 (from 4a), B60 (from 4c), B61 (from 4c)
B*3901   =>   B3901 (from 4a), B39 (from 4b), B16 (from 4b)


For multiple-allele-codes the steps 1-4 have to be repeated for every included allele. The process may result in multiple SDs for each (multiple) allele code. Selection and prioritisation of donors with identical or overlapping SDs is done in a separate routine.

Extending the data source:

The antigen assignments of the WHO correspondence file (2) are “conservative“, which means that it does not include expert assigned antigens or non-rare deviating antigens. It would be easy to add an additional column to the correspondence file (2) to provide additional antigens which should be regarded as possible. It is important that the additional column must only contain the exceptions / extensions to the other 3 columns containing the official WHO assigned antigen(s). The major advantage of such an additional column would be that all necessary data would be globally available within one single file. These data can/should be used for phenotype checking as well. Note that such an extension of the correspondence file will not influence the algorithm described above. In the meantime we are considering to implement a local extension to the correspondence file in order to accommodate certain needs of clients esp. in the international context.

Werner Bochtler and Carlheinz Müller
Ulm, 2006-03-26




Further Information

For more information about the database, queries (including website) or to subscribe to the IPD mailing lists please contact IPD Support.

Please see our licence for our terms of use.