Possible IOHMM topologies in DREM 2.0. The basic topology for a DREM 2.0 IOHMM is shown. The hidden states represent the network nodes (in blue) that we are interested in. The observations (black nodes) are the gene expression ratios which are given to the model, these are dynamic and dependent on the time point. The protein-DNA interaction data (green nodes) are used as supervised input data to construct the network. (a) In the original DREM formulation only one static input node is connected to all hidden nodes. In DREM 2.0 the nodes in the input layer can be dynamic and dependent on the time point with a topology either fully dynamic (b) or a mix of static and dynamic input (c).