EM Clustering based Approach to Decipher Functional Modules in Cross-Talking Signaling systems

It has become increasingly clear that signalling pathways are extensively interconnected and are embedded in networks with common protein components. These components do not exist in isolation but may gather together to form crosstalk modules. Constructing these crosstalk modules has emerged as a good method to understand the mechanisms underlying the propagation of transduction signals in cell. In this paper, we have presented an advancement of the method, which is chiefly used to integrate multiple topological and functional data to detect crosstalk modules between nuclear factor kappa B (NF-kB), p53 and theG1/S phase of the cell cycle systems. Applying the Expectation Maximization (EM) clustering algorithm, our results were comparable to the k-means algorithm. The EM algorithm as a soft clustering method is able to distinguish overlapping parts among clusters, and here we show that it is potentially more effective than the k-means algorithm in detecting the cross talking modules involved in the network interactions between the two systems NF-kB and p53. In addition, the biological analyses support our findings, and propose testable hypotheses to which the functional networks are involved in along with their associated human diseases.