Due to the advances of imaging and high throughput technologies for data acquisition in biomedicine an increasing amount of data is produced. In many applications, data is associated with uncertainty, often due to specific properties and limitations of data aquisition, eg. the resolution of some imaging modality. In addition, integration of data from different sources is essential to optimally support the knowledge discovery process. This talk will introduce the challenges and chances of data integration in the context of two concrete applications in neurosciences and proteomics. Combining uncertain information from different sources may reduce uncertainty and thereby effectively support knowledge discovery.