Well, for example, knowing which genes are associated with a biological mechanism can help to restrict the search for gene targets. If you want to develop a drug against malaria you will look at the genes involved in immunity, blood development, etc..

Moreover, if you know which genes interact together and how, you can more or less predict what will be the effect of blocking the activity of an enzyme or the interaction between two proteins.

There are a lot of examples in the literature about using pathways to design better drugs... for example L. G. Boros and T. F. Boros Use of Metabolic Pathway Flux Information in Anticancer Drug Design

While @Giovanni's answer is correct I can add a bit more information about what we were actually asked to do around WikiPathways by pharma.

1. By making WikiPathways content available as RDF we were able to load in the Open PHACTS linked data resource. In this way you can immediately look for things like "drug that hit targets (from ChEMBL) that are in pathways (from WikiPathways) that contain genes related to specific diseases (from DisGeNET)". (This was part of the original Open PHACTS IMI project).
2. By improving the way we capture directionality of interaction in pathways you can look for drug targets that are up or downstream from your target of interest. (This is part of the second stage of the project).

Finally the BridgeDb identifier mapping framework we use at WikiPathways and in the PathVisio pathway analysis tool turned out to be very useful for semantic web identifier mapping needed in the Open PHACTS project.