The Modeling of Decomposable Gene Regulatory Network Using US-ELM

In the method of constructing gene regulatory network, the correlation model and Bayesian network model have attracted wide attention of researchers with their respective advantages. However, due to their inevitable disadvantages, the effect of using a single method to constructing gene regulatory networks is not ideal. The combination of these two models may be better to construct gene regulatory network. Therefore, this paper proposed a decomposable gene regulatory network modeling method (DGRN). The DGRN combines three methods including the correlation model, unsupervised extreme learning machine (US-ELM) and Bayesian network model (BN) which used to construct initial network, decompose network, and optimize network structure respectively. Initial network are constructed by correlation model and decomposed by US-ELM. The Bayesian network model is used to optimize the structure of the decomposed gene regulatory subnetworks. The experimental results show that the proposed method can improve the computational efficiency and the scale of the network while ensuring good construction results.