It is becoming increasingly clear that RNA-binding proteins are key elements in regulating the cell's transcriptome. Thus, unraveling the interaction network of the RNA-binding proteins by determining their binding sites is becoming an increasingly important topic. There are several high-throughput methods available to detect binding sites such as CLIP-seq. Since not all possible binding sites are covered due to differential expression in tissues and developmental states, the main problem is to come up with good motif descriptions to find missing binding sites and to evaluate the binding strength. Our new approach GraphProt uses an advanced machine learning approach based on our graph-kernel, and is able to use both structural profiles as well as detailed 2D-structures, and predicts missing binding sites with a high accuracy.