Selection of Collections Based on Available Data and Their Quality

Problem:

When biomedical researchers require material and data for their studies they search samples and collections in biobanks. This search can be based on phenotypical or genotypical data of donors, therapeutical data, lifestyle data, etc. Increasingly, this search does not only aim at pointers to samples but also to available data associated with the samples, which can be used in the studies. Here data quality is of essential importance and support for the specification of quality requirements is needed.

Solution:

We present a metadata model which captures the data available for collections in biobanks in a generic and semantic way based on reference ontologies to overcome the problem of different data representation formats and encodings. The prototype implementation used LOINC Codes as reference ontology. This representation is augmented with a representation of essential data quality characteristics on the sample and on the collection level. Search in this metamodel yields which collections contain the specified data in the requested quality.

Discussion:

The metamodel provides an infrastructure for search for appropriate biobanks and collections for a specified requirement data and data quality requirements. It is by far more expressive than biobank directories and does not require nor is it restricted to a minimal data model. It is very flexible and allows to use ontological hierarchies to extend and/or restrict the search. The explicit inclusion of data quality allows to avoid to chase collections which are ultimately not usable. Empirical evaluations of the efficiency of the metamodel for study preparations are planned.