Parallel SPICi

In this paper, a concurrent implementation of the SPICi algorithm is proposed for clustering large-scale protein- protein interaction networks. This method is motivated by selecting a defined number of protein seed pairs and expanding multiple clusters concurrently using the selected pairs in each run; and terminates when there is no more protein node to process. This approach can cluster large PPI networks with considerable performance gain in comparison with sequential SPICi algorithm. Experiments show that this parallel approach can achieve nearly three times faster clustering time on the STRING human dataset on a system with 4-core CPU while maintaining high clustering quality.