Reconstruct feedback control of cell cycle-regulated networks of the yeast by neural network computing

Cells continuously recycle their gene expressions. In order to understand the expressions of cell cycle-regulated genes, time series expression profiles provide a more complete picture than single time point expression profiles. However, these time series expression profiles raise new challenges for computer scientists and statisticians. One of these challenges is the reconstruction of the regulatory connections between genes, proteins, or other gene products. Recently, some analytic methodology or techniques have been constructed to model such time series data to discover gene regulatory networks. But most of these researches do not take account of the feedback control mechanism within a regulatory network. In our approach, a hybrid method is applied to reconstruction of cell cycle-regulated networks to determine gene interactions in gene expression data, especially to deal with the feedback mechanism of some particular genes. By using radial basis function neural network (RBF) and recurrent neural network (RNN), experiments conducted on real world microarray expression data verify that this approach is sufficient for fitting the data set and reconstructing the feedback regulatory networks.