Biological signal transduction of protein-protein interaction (PPI) networks occupies an important position in life activities because it affects the regulation mechanisms of growth, development and metabolism. To further understand the realization of cellular biological functions, we need to predict the directionality of PPIs. This study combines protein semantic similarity and overlapping clustering algorithm to improve the network propagation method for orientating signal transduction networks. This orientation algorithm was applied to an integrin-mediated PPI network (containing 1166 PPIs). As a result, 484 unoriented PPIs (forming protein complexes) and 682 oriented PPIs (interactions involving signal transmission) were obtained. Finally, in order to evaluate the effectiveness of this method, it was compared with the original method based on network propagation, and acceptable results were obtained.