摘要：The prediction of protein function is one of the most challenging problems in bioinformatics. Several studies have shown that the prediction using PPI is promising. However, the PPI data generated from high-throughput experiments are very noisy, which renders great challenges to the existing methods. In this paper, we propose an algorithm, MFC, to efficiently mine maximal frequent dense subgraphs without candidate maintenance in PPI networks. Instead of using summary graph, MFC produces frequent dense patterns by extending vertices. It adopts several techniques to achieve efficient mining. Due to the imbalance character of PPI network, we also propose to generate frequent patterns using relative support. We evaluate our approach on four PPI data sets. The experimental results show that our approach has good performance in terms of efficiency. With the help of relative support, more frequent dense functional interaction patterns in the PPI networks can be identified.
