The concept of translational research is always hampered by the problem that most of the disease phenotypes do not have a mono causal origin. Therefore most treatment schemes based on one to three drugs are not really productive for most of the patients even if the patients are carefully selected from the responder group. Here the array techniques has inspired many research groups to develop algorithms deriving interaction networks or regulatory networks from this type of data to better get rid of the complexity of the biochemical interactions. The challenge is to find networks and to select the group of master nodes which might be good targets for a balanced multi-drug treatment. This means not only to measure one data type with array techniques but to join array data from multiple platforms and different data levels. Our goal is to integrate these data types to form networks with a predictive character for osteosarcomas.

The existing web platform CAPweb/VAMP from the Institute Curie is based on a Java web-client and R. This platform is focused on array data analysis and visualisation, can be extended by additional R modules and is therefore an excellent choice to implement further algorithms for data integration and network prediction. We are now establishing algorithms beyond a pure association of effects like permutation procedures for optimal rank orders of effects in a given subset of 16 factors which can be assembled to bigger units and selection procedures of gene expression signals by gene dosage concepts.

The presented approach is sustainable because the platform can be constantly extended and improved. On the other hand this platform is end-user suitable. This is the best way to bring theoretical concepts to the bench scientist. As a consequence translational research will become more real and complex systems more feasible.