Many genes have unknown function and/or regulation, so it is important to generate in silico predictions that can guide experimental work. I propose a method for probabilistic inference of functionally related, transcriptionally regulated gene modules. The method is based on integrative data analysis of transcription factor binding motifs, frequent patterns of temporal association of gene expression, epigenetic state and cellular localization in differentially stimulated time-course data. Datasets for testing include differentially stimulated yeast and MCF-7 human breast cancer cells. The validation relies of GO and protein interaction annotations, as well as known transcription factor networks in yeast and human.