RelEx interaction extraction rules. The three RelEx rules search the dependency graph for paths between protein names indicating an interaction. Found paths are filtered according to criteria such as whether certain interaction terms appear in them. In the examples the protein names are marked in bold, and the possible interaction words in italics. Rule 1 extracts protein1-relation-protein2 structures from the dependency graph. In the basic case the rule requires that the path includes a subject dependency (e.g. nsubj). Rule 2 extracts relations that are represented by a sequence of noun phrases connected to each other via prepositional dependencies (prep). Rule 3 searches for structures of the type relation-between-protein1-and-protein2 connected through prep_between and conj_and dependencies.