A transcription-translation model of gene networks and a method to reconstruct it from gene expression data are proposed. The model is a hybrid system based on the Glass network with continuous-time dynamics and logical interactions. Transcription-translation dynamics is introduced into the Glass network. The reconstruction of gene networks is reduced to the problem of estimating logical functions from binary representations of quantities of mRNAs and proteins. The reconstruction method is applied to the gene expression data of circadian rhythms. The response characteristics of the reconstructed gene network to periodic stimuli are analyzed. The results suggest the existence of a receiver gene that responds to an external signal, consistently with biological knowledge. (C) 2008 Elsevier Ltd. Ail rights reserved.