technical report Inferring Crosstalk Modules of Combined Networks by Using Multi-relational Clustering


Signal transduction pathways have typically been considered as linear cascades of separate entities. However it has become increasingly clear that signalling pathways are extensively inter- connected and are embedded in networks with common protein components and crosstalk with other networks. Constructing the crosstalk networks is crucial to understand the mechanisms un- derlying the cell signalling. In this paper, we have presented a new computational method to detect the crosstalk networks between signalling networks based on the human protein-protein interaction network and multiple relational data. The combination of multiple relational data is advantageous to infer those intricate networks. The NF-kB and p53 systems are one of most important networks in discovering the cellular signal transduction and human disease. The proposed method has been applied to find out the networks crosstalk between these two central networks. Conducting 3-fold cross validation, we obtained a high clustering case likelihood for both the k-means method and the EM method. The better performance of EM methods demonstrates that soft-clustering is more sensibly than hard-clustering in detecting crosstalk networks with sharing parts. In addition to the computational validation, the biological analysis shows the plausibility of the findings and suggests several testable biological hypotheses to study the functional networks crosstalk between signalling networks and the its association with human disease.

Paper Details


P. Nguyen,  A. Ihekwaba,  C. Priami