Identifying spurious interactions in the protein-protein interaction networks using local similarity preserving embedding

Over the last decade, the development of high-throughput techniques has resulted in a rapid accumulation of protein-protein interaction (PPI) data. However, the high-throughput experimental interaction data is prone to exhibit high level of noise. In this paper, we propose a new approach called Local Similarity Preserving Embedding(LSPE) for assessing the reliability of interactions. Unlike previous approaches which seek to preserve a global predefined distance matrix in the embedding space, LSPE tries to adaptively and locally learn a Euclidean embedding under the simple geometric assumption of PPI networks. The experimental results show that our approach substantially outperforms previous methods on PPI assessment problems. LSPE could thus facilitate further graph-based studies of PPIs and may help infer their hidden underlying biological knowledge.