Return to Research

Network Controllability Project

Networks are all around us. The first example crossing one’s mind might be the World Wide Web, but probably not the only one. Our world is full of social structures, networks, where individuals are connected with each other by different means such as mobile phones or transportation. A network can be represented by nodes and by edges, where edges describe connections between the nodes. Imagine airports with flight connections. There is continuous flow of people travelling through the cities, and the biggest airports are the system hubs. Networks can also be microscopic, for example metabolic networks or gene regulatory networks in a cell. Common to all is that the state of the system typically changes with time. Hence, we call them dynamic networks. With many systems, it would be in our interest to steer them or even control them toward some desired state. However, this has turned out to be a very challenging task.

Within this project we aim to better understand the fundamental principles of network controllability. We develop novel and extend the existing methods on controlling networks. In particular, we are interested in large and complex biological networks, which are extremely difficult to control. Therefore, parallel to full control we are investigating also how such systems could be controlled partially. This kind of approach might, indeed, enable many practical applications, which otherwise would be infeasible simply from engineering perspective.
We exploit the control principles in order to predict the best way to engineer the networks, which we aim to control. In particular, we apply the methodology to two biological systems, cancer and renewable energy production. Malignant growth tend to arise when there is ‘too much energy in a system’. Hence, controlling cancer would mean steering the tumor towards non-growing state. Our ambition is to re-program cancer networks by identifying cell’s own pathogenic addictions and other vulnerabilities. Multi-targeting those identified nodes can improve anti-cancer therapeutics. In renewable energy production we optimize the metabolic flux to obtain maximal fuel yield. While the practical goals with energy aspect are rather opposite, the ultimate goal in both systems is to be able to control them. Obviously, the generic approaches and tools can be used in other applications. They also serve as a complement to traditional micro-scale engineering.

This research project is truly multidisciplinary with four research groups involved. Group leaders and respective links:

Ion Petre

Tero Aittokallio

Patrik Jones

Krister Wennerberg

Controlling Directed Protein Interaction Networks in Cancer

Control theory is a well-established approach in network science, with known applications in bio-medicine and cancer research. We build on recent results for full and structural controllability of directed networks, which gives a set of driver nodes able to control the whole network, or an a-priori defined part of it, respectively. We develop a novel approach for the structural controllability of cancer networks and demonstrate it
for the analysis of breast, pancreatic, and ovarian cancer. We build in each case a signalling transduction (STN) protein-protein interaction (PPI) network and focus on the so-called “essential proteins” specific to each cancer type in our study. We show that the cancer essential proteins are efficiently controllable from a (relatively small) computable set of driver nodes. Moreover, we adjust the method to find the driver
nodes among FDA-approved drug-target nodes. Interestingly, we find that while many of the drugs acting on our driver nodes are part of known cancer therapies, some of them are not used for the cancer types analyzed here; also some drug-target driver nodes identified by our algorithms are not known to be used in any cancer therapy. Overall we show that a better understanding of the control dynamics of cancer through mathematical modelling could pave the way for new efficient therapeutic approaches and personalized medicine.

The algorithm used in this research can be downloaded from following links:

Pipeline NetControl4BioMed

Basing on our algorithm, we have implemented a pipeline (that can be downloaded and installed as a stand-alone software) and the online service (i.e., a publicly available web interface for an instance of the pipeline installed on our servers). The pipeline performs an automatic generation of intracellular molecular interaction networks (by combining publicly available pathway data) and identification of driven nodes (that also can be targeted by FDA approved drugs) for a set of target genes/proteins defined by the user.

Old pipeline source code is available here:

Old online service based on the pipeline is available here: