Molecular biology produces and accumulates huge amounts of data that are generally integrated within graphs of molecules linked by various interactions. Exploring potentially interesting substructures (clusters, motifs) within such graphs requires proper abstraction and visualization methods. Most layout techniques (edge and nodes spatial organization) prove insufficient in this case. Royer et al. introduced in 2008 Power graph analysis, a dedicated program using classes of nodes with similar properties and classes of edges linking node classes to achieve a lossless graph compression. The contributions of this paper are twofold. First, we formulate and study this issue in the framework of Formal Concept Analysis. This leads to a generalized view of the initial problem offering new variants and solving approaches. Second, we state the FCA modeling problem in a logical setting, Answer Set programming, which provides a great flexibility for the specification of concept search spaces.