The rapid growth of data stored in molecular biology-related databases has stimulated the development of integrative tools for retrieval and presentation of the data in the form of, e.g., biological association networks. We argue that a general and specific knowledge-based approach may provide substantial support for automated reconstruction of networks which otherwise tend to be large and, eventually, unreadable. This knowledge-based approach introduces a novel strategy with the potential to greatly enhance the explanatory power of automatically generated biological association networks. We discuss the motivation for a study of the process an expert employs while building a network, and suggest that a series of expert sessions be used as a case library for future reference. An example of a biological problem and the shape of its solution is described, and the types of knowledge involved are discussed.