Several studies have shown that the prediction of protein function using PPI data is promising. However, the PPI data generated from experiments are noisy, incomplete and inaccurate, which promotes to represent PPI dataset as an uncertain graph. In this paper, we proposed a novel algorithm to mine maximal dense subgraphs efficiently in uncertain PPI network. It adopted several techniques to achieve efficient mining. An extensive experimental evaluation on yeast PPI network demonstrated that our approach had good performance in terms of precision and efficiency.

Biclustering is one of the important techniques for gene expression data analysis. A bicluster is a set of genes coherently expressed for a set of biological conditions. Various biclustering algorithms have been proposed to find biclusters of different types. However, most of them are not efficient. We propose a novel algorithm MRCluster to mine constant row biclusters from real-valued dataset. MRCluster uses Apriori property and several novel pruning techniques to mine biclusters efficiently. We compare our algorithm with a recent approach RAP, and experimental results show that MRCluster is much more efficient than RAP in mining biclusters with constant rows from real-valued gene expression data.

Mining functional modules with biological significance has attracted lots of attention recently. However, protein-protein interaction (PPI) network and other biological data generally bear uncertainties attributed to noise, incompleteness and inaccuracy in practice. In this paper, we focus on received PPI data with uncertainties to explore interesting protein complexes. Moreover, some novel conceptions extended from known graph conceptions are used to develop a depth-first algorithm to mine protein complexes in a simple uncertain graph. Our experiments take protein complexes from MIPS database as standard of accessing experimental results. Experiment results indicate that our algorithm has good performance in terms of coverage and precision. Experimental results are also assessed on Gene Ontology (GO) annotation, and the evaluation demonstrates proteins of our most acquired protein complexes show a high similarity. Finally, several experiments are taken to test the scalability of our algorithm. The result is also observed.