Biomedical knowledge graphs serve as powerful data integration platforms and are widely utilized in early drug discovery by biotech and pharmaceutical companies. Knowledge graphs facilitate efficient retrieval of existing data and augment this with predictive inference, leveraging advanced analytical tools such as graph neural networks (GNNs). In this presentation, we will demonstrate the application of predictive approaches to two representative tasks in drug target characterization. The first application involves the analysis of high-throughput cell imaging, where GNN-based modeling aids in evaluating the novelty of experimentally determined genetic associations. The second application showcases the use of GNNs to develop predictive AI models for early drug target safety assessment (TSA), integrated into Evotec's internally developed TSA platform.