Annotating data with concepts of an ontology is a common practice in the biomedical domain. Resulting annotations (data-concept relationships) are useful for data integration whereas the background ontology can guide the analysis of integrated data. Formal Concept Analysis (FCA) allows to build from a binary context a concept lattice that can be used for data analysis purposes. However annotated biomedical data are not binary and a binarization procedure is required as a preprocessing, coming with classical problems, e.g. a trade-o between expressivity and the large number of induced binary attributes. Interestingly, pattern structures o er a general method for building a concept lattice from any set of objects associated with partially ordered descriptions. In this paper, we show how to instantiate this general framework when the space of descriptions is based on an ontology. We illustrate our approach with the analysis of biomedical annotations and we show its capabilities for knowledge discovery.