The major goal of computational biology is to derive regulatory interactions between genes from large-scale gene expression data and other biological sources. There have been many attempts to reach this goal, but the field needs more research before we can claim that we have reached a complete understanding of reverse engineering of regulatory networks. One of the aspects that have not been considered to a great extent in the development of reverse engineering approaches is combinatorial regulation. Combinatorial regulation can be obtained by the presence of modular architectures in regulation, where multiple binding sites for multiple transcription factors are combined into modular units.
When modelling regulatory networks, genes are often considered as "black boxes", where gene expression level is an input signal and changed level of expression is the output. We need to shed light on reverse engineering of regulatory networks by modelling the gene "boxes" at a more detailed level of information, e.g., by using regulatory elements as input to gene boxes as a complement to expression levels. Another problem in the context of inferring regulatory networks is the difficulty of validating inferred interactions because it is practically impossible to test and experimentally confirm hundreds to thousands of predicted interactions. Therefore, we need to develop an artificial network to evaluate the developed method for reverse engineering. One of the major research questions that will be proposed in this work is: Can we reverse engineer the cis-regulatory logic controlling the network organised by modular units?
This work is aiming to give an overview of possible research directions in this field as well as the chosen direction for the future work where more research is needed. It also gives a theoretical foundation for the reverse engineering problem, where key aspects are reviewed.