In this paper, we propose a new random walk-based clustering algorithm for detecting functional modules in protein-protein interaction (PPI) networks. It has been tested on two yeast PPI networks. Greater precision, better homogeneity and higher modularity were achieved in comparison with the results produced by the recently developed RRW clustering technique and the well-known CFinder algorithm. A much higher level of true positives were observed in the clustering results. The analysis indicated that the proposed method can not only detect overlapping modules but also be potentially used to identify functional modules with different topological structures, which may not be highly connected.