In this paper, we present a data-mining approach in gene expression matrices. The method is aimed at extracting formal concepts, representing sets of genes that present similar quantitative variations of expression in certain biological situations or environments. Formal Concept Analysis is used both for its abilities in data-mining and information representation. We structure the method around three steps: numerical data is turned into binary data, then formal concepts are extracted and filtered with a new formalism. The method has been applied to a gene expression dataset obtained in a fungal species named Laccaria bicolor. The paper ends with a discussion and research perspectives.