Microarrays allow monitoring of thousands of genes over time periods. Recently, gene clustering approaches specially adapted to deal with the time dependences of these data have been proposed. According to these methods, we investigate here how to use prior knowledge about the approximate profile of some classes to improve the classification result. We propose a Bayesian approach to this problem. A mixture model is used to describe and classify the data. The parameters of this model are constrained by a prior distribution defined with a new type of model that can express both our prior knowledge about the profile of classes of interest and the temporal nature of the data. Then, an EM algorithm estimates the parameters of the mixture model by maximizing its posterior probability.