Resource discovery in a distributed environment is always a challenging issue. It is even more difficult to provide an efficient query routing mechanism while still able to support complex query processing in a decentralised P2P environment. This paper presents an adaptive approach to P2P resource discovery. It separates the routing of queries from query matching mechanism so that an effective combination could be explored. Three properties of scientific research communities provide the grounding for the method: the existence of common interest groups, the willingness to share resources of common interests and the transitive relationship in the sharing behaviour. By exploiting these properties, search queries can be efficiently forwarded to those who are more likely to have the answers to improve the quality of search results and to reduce the network traffic. Experimental results have provided some evidence to confirm the efficiency of this adaptive approach.