One of the main objectives of systems biology is to learn network topologies from gene expression time series. The ultimate goal is to infer the dependency between random variables in the form of a network. To deal with this task, the class of dynamic Bayesian network (DBNs) models have been widely applied. The essential assumption for DBNs is that the regulatory processes are homogenous, so that the interaction parameters are constant over time. However, this assumption is too restrictive for many real-world applications. To overcome this issue, non-homogenous dynamic Bayesian networks (NH-DBNs) are considered as the best alternatives. In this work, we applied the class of NH-DBNs models to infer the network structure from gene expression time series. Particularly, we will discuss the results for two types of gene networks, firstly, the yeast network (true structure of the network is known) and secondly, mTOR network (true structure of this network is unknown). In the presentation we will put the main focus on the mTOR network, as we currently struggle with the data analysis. All our network models are based on the assumption that measurements have been taken at equidistant time points. But this assumption is not fulfilled for the mTOR data, where after a stimulation of the system the measurements were taken after 0, 1 ,3, 5, 10, 15, 30, 45, 60 and 120 minutes.