In this talk I shall analyze network mining and bisociation from a logical and probabilistic inference point of view. This is inspired by the work on ProbLog for link mining [De Raedt, Kimmig, Toivonen, IJCAI 2007], which is - in turn -based on Biomine [Sevonen et al. DILS 06]. The talk shall introduce a probabilistic semantics for networks and databases and use it to clarify notions of deduction, abduction and explanations, induction, analogy, abstraction and spread of influence. All notions will be illustrated in the context of the Biomine network.