A protein complex is a group of two or more proteins that interact to perform specific biological functions. Protein complexes are essential for cellular processes, and their understanding is still limited. Therefore, the detection of protein complexes is an important topic in bioinformatics, often using protein-protein interaction (PPI) networks. There are many approaches that can detect tightly connected subgroups of proteins in PPI networks that correspond to known reference protein complexes or that may be the subject of further investigation based on a biological assessment of their significance. Accurate detection is still problematic due to the complexity of PPI networks and the dynamic and asymmetric nature of protein interactions. However, PPI networks usually do not capture such properties. The novel approach presented in this paper exploits the asymmetric relationship between pairs of proteins based on analyzing the network structures in their neighborhood. We show that a simple mDepStar method exploiting this structural asymmetry outperforms most of the six state-of-the-art detection methods using only the topological properties of the network. Comparisons were performed on two well-known PPI networks and a newly constructed BioGRID network with a new SGD24 reference set for Saccharomyces cerevisiae. In addition, we introduce a new MR-score inspired by the F-measure that incorporates the most commonly used Maximum Matching Ratio.