Interpreting the rapidly increasing amount of experimental data requires the availability and representation of biological knowledge in a computable form. The Biological expression language (BEL) encodes the data in form of causal relationships, which describe the association between biological events. BEL can successfully be applied to large data and support causal reasoning and hypothesis generation. With the rapid growth of biomedical literature, automated methods are a crucial prerequisite for handling and encoding the available knowledge. The BioNLP shared tasks support the development of such tools and provide a linguistically motivated format for the annotation of relations. On the other hand, BEL statements and the corresponding evidence sentences might be a valuable resource for future BioNLP shared task training data generation. In this paper, we briefly introduce BEL and investigate how far BioNLP-shared task annotations could be converted to BEL statements and in such a way directly support BEL statement generation. We present the first results of the automatic BEL statement generation and emphasize the need for more training data that captures the underlying biological meaning.