Model-based approach for optimizing the storage management of bio-repositories

• First ontology-based framework for exploring optimization and simulation approaches to bio-repository storage management.
• Ontology mapping provides flexible, but systematic support of the modelling process.
• Transferable to different domains.

Systematically collected samples of human tissue and fluids (e.g. blood, urine) play a key role in biomedical research. Biomaterial banks aim at collecting such biomaterials systematically and providing long-term availability of samples, which have to be kept constantly frozen at low temperatures. Therefore, storage management has to meet critical constraints in order to maintain high quality of the samples. The project aims at supporting the application of optimization and simulation algorithms for storage management. The approach has to provide means for adapting the algorithms flexibly and systematically to different or changing settings of biomaterial management and, therefore, aims at implementing a model-based framework to specify and run the algorithms.The process of establishing an optimization/simulation model for a given Biomaterial bank supported by the project starts with acquiring separately a model (1) of biomaterial handling and (2) of the optimization or simulation approach. The models for specific situations are derived from generic ontologies. A subsequent model mapping step assigns components representing aspects of biomaterial handling and storage to roles in a suitable optimization or simulation scheme. The mapping then enables the semi-automatic generation of optimization and simulation code. In the case of offline optimization problems, the framework produces program code that can be processed by an existing generic solver for mixed integer optimization programs (Zuse Institute Mathematical Programming Language code to be processed by the solver Solving Constraint Integer Programs). In the case of online problems, i.e., problems with an ongoing production of new information relevant for the optimization, the approach generates Java code suitable for processing by an existing simulation kernel (SimKit).The mapping module has been verified by reproducing a given solution for the Capacitated Facility Problem. Preliminary results of simulation runs show antagonistic effects of sorting bio-specimens by sample type: Sorting decreases the number of freezer opening operations while it increases fragmentation. The project is the first to adopt ontology-based modeling and model mapping for systematically exploring optimization and simulation approaches to biomaterial storage management. The results of simulation runs not only demonstrate the feasibility of the approach, but yield first practical insights (e.g. the effect of sorting samples).