The research project BIO MIS focuses on three different biomedical problem domains, each dealing with huge amounts of data. In order to validate new input data and create a proposal system this data is used to build up an adaptive knowledge-base. The knowledge is structured by creating clusters of similar items and searching relationships among them. In the classification process system input is compared to the generated cluster profiles. If it affirms similarity it is assigned to the according cluster and, in that way, adapts the knowledge-base to recent circumstances. The clustering process is non-trivial since we mainly have to deal with non-metric laboratory data and cannot use standard clustering algorithms. Our new clustering approach is exemplified on the "LabExpert" package; a system part of one subproject implementing a validation system for large-scale laboratory data.