Complexity of post-genomic data and multiplicity of mining strategies are two limits to Knowledge Discovery in Databases (KDD) in life sciences. Because they provide a semantic frame to data and because they benefit from the progress of semantic web technologies, bio-ontologies should be considered for playing a key role in the KDD process. In the frame of a case study relative to the search of genotype-phenotype relationships, we demonstrate the capability of bio-ontologies to guide data selection during the preparation step of the KDD process. We propose three scenarios to illustrate how domain knowledge can be taken into account in order to select or aggregate data to mine, and consequently how it can facilitate result interpretation at the end of the process.