RSS-Feed abonnieren
DOI: 10.1055/s-0032-1330820
Analysis of Cancer Signaling Motifs
In the meanwhile, large sets of cancer genomes have been sequenced by deep level sequencing methods. To gain an understanding of these outcoming large lists of cancer specific mutations, methods have been developed to embed them into the context of cellular signaling interactions. However, the entire system of cellular signaling is too complex to allow its modeling as a whole. Thus, simple building blocks (network motifs) have been defined and studied to describe signaling events in cellular networks. We model signaling flow through triangular motifs when nodes in these networks are disturbed and compare the modeling results to known cancerous gene mutations (loss-of-function). We have developed a neural network to simulate signaling flow in triangular motifs from its upstream/input proteins (nodes) to its downstream/output nodes. We estimate the substantiality of the nodes in each triangle by comparing the simulation results with and without loss-of-function mutation of its nodes. The objective for both scenarios is to minimize Hamming distances of the computed output and the expected output of downstream proteins using a mixed-integer linear programming approach. Neural edge weights are optimized to fulfill this criterion for each reasonable combination of input and output patterns resulting in distributions of score values for every triangle and perturbed node. These distributions are analyzed to estimate the impact onto the network when losing the function of the nodes. We compared our results to a list of experimentally confirmed cancer mutated genes. Interestingly, we find a high enrichment of our hits in the proteins from these cancer mutated genes. Our network model is generic and enables mapping loss of function nodes in cancerous signaling networks to their substantial role in signal processing. In the future, this concept may be used to support distinguishing driver mutations from passenger mutations when analyzing deep level sequencing data from tumor material.