A Method For Finding Essential Protein Based On Perturbation Propagation

In the field of computational bioinformatics, essential protein, as a kind of indispensable protein for cellular life, is widely used to identify drug targets and pathogenic genes. It is also important for synthetic biology. As consequence, how to identity the essential protein efficiently becomes a hot issue in this field. At present, there are mainly two kinds of method for finding the essential protein. One is based on biological experiments, and the other is based on the node centrality of complex network. The former method is very time-consuming and not applicable for all species. While the latter one gets attention because of high efficiency. However, in practice, the efficiency of the latter method runs into the bottleneck, because only the static feature of topological structure of network has been taken into consideration but not the dynamic progress of protein-protein interaction network.With law of mass action, an perturbation propagation experiment has been conducted by combining the data of protein-protein interaction network and that of protein concentration. Then, a coupling matrix of concentration is built after observing whether the network node is disturbed or not. Finally, the new influence of nodes is reviewed in accordance with the impact of those nodes being disturbed. The reason why we evaluate the impact of nodes through perturbation propagation is that the change of the concentration of essential protein will cause the disorders of biological functions. The effect of its concentration change is more important than that of other non-essential protein concentration. Through perturbation propagation, we can clearly perceive the influence of one node to other nodes in networks. By this means, we are able to evaluate the influence of nodes, greater influence, and larger probability of essential protein.According to the experiments on the protein-protein interaction network of yeast and that of insulin in mice, compared with other five common centralities, the method proposed in the paper can identify more essential protein, and the intersection of essential protein found by this method and other method is minor. As a ranking mechanism, we also calculate the robustness of two kinds of network to our method and other centrality methods. In order to make comparison, the method of random ranking is applied in our experiment as well. It shows that the robustness of network to our method is similar to random ranking, which indirectly confirms the conclusion that the intersection of essential protein found by this method and other method is minor.Compared with other methods, the method proposed in the paper can identify more essential protein, which is an efficient improvement for the existing methods. Meanwhile, the paper also provides a new idea about node centrality measurement, that is, we can evaluate the centrality of nodes from the perspective of perturbation propagation in accordance with the difference of actual network.