By using the analysis idea and methods of complex network

we adopted a normalized Laplace matrix and a K-mean clustering method to detect multiple communities in a gene regulatory network (GRN).At the same time

gene-gene interactions both inside a community and among communities were also described.In addition

in order to reflect the actual gene-gene interaction process and to improve the modeling precision of GRN

a time-series spectrum analysis method was used to estimate delay of gene expression before the process of community detecting.Experimental results of the cell cycle-regulated genes of yeast show that the proposed construction method of GRN not only can reflect the actual gene-gene interaction process much exactly but also can provide details about the gene regulator model.