In a molecular tumor board, therapy decisions for cancer patients are taken based on a set of different data types (clinical data, radiological data, histopathological data, genomic data, gene expressuib data, and other high-throughput data). In addition, specific features are prioritized using bioinformatic methods (e.g. specific mutation and gene expression patterns), which are then used together with database research (mostly performed manually) to compile therapy suggestions and discuss in a disciplinary manner. However, it stays unclear how progressive knowledge and data from external sources can be integrated in the activity of a molecular tumor board as a standard-driven process. On the other hand, there is no systematic acquisition of previous patients and recommendations, that would make comparison of current and previous similar cases possible. In order to refer to big numbers of patients' cases, an interconnection between moelcular tumor boards in different locations should be established. The goals here are as follows:

• The conceptioning of a modular platform, in which reproducibility, documentation and update mechanisms can be realized alongside the processes.
• Creating concepts to bind clinical routine with research data.
• Reinforcing the automation of data integartion and the reprocessing including external knowledge.
• Developing of machine learning processes for the analysis in context of MTBs, for example the usage of unstructured knowledge (e.g. Doctor Reports), geenration of complex signatures to model therapy Therapieansprechen and identification of similar patients.
• The implementation of a prototype and open-source software for a modular, reproducible software platform of a MTB.