AI systems for integrative
Multi-Omics Data

Molecular data plays a crucial role in oncology for the identification of new therapeutic and diagnostic targets. This group investigates AI approaches in molecular databases encompassing genomics, epigenomics, transcriptomics, proteomics and metabolomics of clinical partners. The aim is to identify subpopulations of patients with clinically relevant survival phenotypes or to predict treatment response. In the study design, a dedicated data base is created to train AI models and there is close collaboration with other junior medical groups, in particular with the Collaborative Development Group, to share and validate multi-omics data.

While group #4c focuses on image data, the emphasis here is on multi-omics data. Existing proteomics data from other funding programs (BMBF CancerScout) form a comprehensive basis. The research focuses on developing novel AI algorithms that take into account the specific requirements of medical applications. These should efficiently interpret molecular data from various sources, taking into account a large knowledge base, e.g. ontologies and protein-protein interaction networks (proteome and transcriptome). A particular focus is on explainability and confidence of predictions, with robust estimation of uncertainties and privacy preservation. The integration of the platform in the CCC-N ensures that data and algorithms are driven in a clinically relevant way and meet the requirements of medical applications.