Complex systems

Complex systems can be found in physics, biology, business and social sciences, to name just a few examples. Molecules and their interactions, metabolic cycles of the human body, the networks connecting people on social media, or companies from the same industry doing business deals with one another – these are all examples of complex systems.

One molecule on its own is a relatively simple system. But if you take a great number of them together, they have emergent characteristics which, for example, determine whether they are solid, liquid or gaseous. Human metabolism involves countless interactions between proteins, enzymes, genes and many other factors, which, together, keep the body alive. The collective behaviour of people on social media can create trends; information such as memes or fake news gets widely distributed.

While many other scientific disciplines focus on a specific part of a system and its nature, the science of complex systems looks at a system in its entirety. In this context, researchers abstract real-world systems in order to simplify them and identify their essential characteristics. This helps them to determine the basic laws which even very different systems have in common. In this process, researchers often use networks as tools that help them to mathematically describe these systems.

Networks

A network is a collection of points (nodes) which are connected by lines (edges). The nodes represent individual components of a system, e.g., molecules, human beings or companies. The edges represent the interactions between components.

Networks come in many different shapes: The connections can be evenly distributed among the nodes, so that the nodes, on average, have the same number of edges. We may, however, also find a concentration of many edges at just a few nodes. The edges may also contain directional information, illustrated by arrows, in order to illustrate directional interactions, such as the selling of a product in a trade network. A network may contain different types of nodes and edges for the purpose of illustrating even more complex systems. Furthermore, networks are not necessarily static. They may change over time as nodes and edges are added or removed.

The science of complex systems uses illustrations like this to model systems and make statements about them based on mathematical processes. This enables researchers to identify interconnections between genes and diseases, assess the robustness of computer and power networks or retrace the way in which particularly well-connected individuals and organisations disseminate information via social media.

This approach also has its limits, though: if the information about the real-world systems to be modelled is insufficient, for example, this makes it impossible to illustrate them correctly. The sheer number of nodes, edges and their characteristics may also make it difficult for researchers and their computer systems to handle them and gain insights from them.

Network medicine

In medicine, too, networks play an important role and can be used in a variety of ways. Researchers use three categories of networks: Molecular networks for modelling the interactions between proteins, enzymes and other substances that play a role in human metabolism. Disease networks for identifying how certain genes or metabolites and their various combinations are connected to certain diseases. Population networks for understanding the spread of infections during a pandemic, for example.

Molecular networks

Molecular networks can roughly be divided into two categories: networks representing physical interactions and networks illustrating more abstract functional interconnections.

Protein-protein interaction networks model the biophysical behaviour of proteins in the human body, which forms the basis of many bodily functions. Because of the molecules’ composition and folding, the mechanisms which determine how a protein interacts with other proteins are very complex and highly specific. Only specific proteins can fulfil specific tasks. If these intricate networks are disrupted, for example by an incorrectly folded protein, this can give rise to various diseases ranging from diabetes to cancer. Over the last decades, a huge amount of data has been collected about protein-protein interaction networks in the human body. Nevertheless, researchers surmise that only 30 percent of all interactions have yet been documented. They use modern algorithms to analyse data and predict network connections that have not yet been studied.

Metabolic networks constitute the next level; in addition to proteins, they also include other components of human metabolism, as well as genes and their expression, i.e., the proteins produced by a certain gene. The most comprehensive model of the human body’s metabolic network that is currently available comprises thousands of molecules, reactions and genes. Such models enable researchers to modify the nature, concentration and types of interaction between various elements of the network in order to simulate their effects. This helps them to better understand disease pathogenesis and, to a certain extent, predict the effect of drugs.

Gene regulation networks focus on the interactions between various genes and how they determine the up- and down-regulation of gene expression. Specific proteins and pieces of RNA that are based on certain genes are responsible for controlling the expression of other genes, based on environmental influences and interactions with other molecules. This may happen in various different ways, which means that the network model must contain a myriad different types of nodes and edges, making it extremely complex. Gene regulation networks help researchers to understand human development from the egg to adulthood and diseases connected thereto.

Gene co-expression networks illustrate more abstract functional relationships between various genes. Genes are the network’s nodes, which are connected by an edge if both of them are significantly activated under certain circumstances, for example in the event of a disease, and produce proteins. These networks do not indicate direct causal relationships, but can help researchers to identify groups of genes that are regulated by the same mechanisms or are responsible for certain metabolic processes. This, in turn, may help to explain complex phenomena such as autism spectrum disorders or chronic inflammatory bowel diseases, which result from the interactions of many different genes.

Gene interaction networks are based two genes being interconnected, meaning that a modification of both genes at the same time produces a different effect than the modification of only one of them. A functional connection of this kind can show researchers which genes are involved in certain biological processes, thus helping them to better understand these processes and identify new therapeutic measures, such as the application of certain drugs.

Disease networks

Disease networks are abstract illustrations of reality that go one step further than molecular networks, being based on various interconnections between individual diseases.

Diseases are the nodes of these networks, which may be connected on the basis of their genetic root causes. Several thousand human diseases have already been mapped at this molecular level. The results have shown that diseases rarely present individually, but in groups of diseases with similar genetic causes. It has also become clear that diseases which belong to the same group share similar metabolic processes and gene expressions. Researchers can use these findings to develop new therapies. Drugs that have been developed for a specific disease may, for example, also be used to treat a disease that belongs to the same group.

Diseases can also be connected in a network if their clinical symptoms are similar. These data clearly show that similar symptoms indicate the same genetic causes and an increased interaction between relevant proteins.

In a third type of disease network, two diseases are connected in the event of co-morbidity, i.e., if the two diseases tend to occur at the same time. Diseases which are found at the centre of such a network have a higher mortality rate, for example. These networks may also be used to identify side effects of drugs and establish whether a disease is more likely to be caused by genetic or environmental influences.

Population networks

Population networks provide different illustrations of interactions between human beings. They can thus be used to model and predict the spread of infectious diseases like swine flu or Ebola.

The structure of population networks is based on real-life events and determines how they work and which insights can be derived from them. These networks may be based on social relationships, as in families or groups of friends, the places where people spend time, like their homes or places of work, on shared political convictions, or on transport networks, which may range in scale from individual cities to the whole globe.

Multi-layered social networks of this kind can be used not only to illustrate the spread of infectious diseases, but also to explain the social factors that influence the occurrence of non-infectious diseases such as obesity. Furthermore, these networks can also model the spread of ideas, attitudes and the behaviour of human beings.

The Ludwig Boltzmann Institute for Network Medicine was publicly launched in February 2024 and took up its work in the same year. It is integrated into the University of Vienna and has an annual budget of up to 1.5 million euros. The Institute has been set up for seven years, with the option of an extension for a further three years.

Jörg Menche is the Institute’s Scientific Director. There are three Research Groups which study networks in the medical context at three levels and actively exchange ideas and data with one another. The Research Group leaders are Jörg Menche, André Rendeiro and Julia Guthrie. There is an additional Cross-Sectional Group responsible for visualising networks and exploring the data in them.

The teams at the Ludwig Boltzmann Institute for Network Medicine are multidisciplinary in order to ensure that the diverse topics can be comprehensively studied, that scientific results can be translated into clinical applications, and that practical resources are developed for the research community. The team members come from fields as diverse as bioinformatics, computer science, physics, mathematics, architecture and digital art. Digital artists are part of the Cross-Sectional Group and produce visualisations that make complex networks accessible in virtual reality.

Jörg Menche

Jörg Menche’s team primarily focuses on network medicine at the molecular level and, in particular, on protein-protein interaction networks. To date, the researchers have, for example, studied the causes of cardiovascular diseases, one of the leading causes of death worldwide. They use networks that link the diseases to the responsible genes, proteins and other factors to disentangle the intricate network of interactions that gives rise to these diseases. The team thus contributes to the development of more efficient treatment methods.

A further study looks at rare diseases, which are often caused by a single mutation in a single gene. Networks help the researchers to understand the consequences of such a mutation across several scales in the human body. The resulting complex network comprises more than 20 million connections between various genes and helps researchers to understand and illustrate the influence of mutations associated with thousands of rare diseases on different bodily systems.

After studying physics and completing a doctorate at the Max Planck Institute of Colloids and Interfaces, Menche worked with Albert-László Barabási at Northeastern University and Harvard Medical School in Boston before founding his own research group at the CeMM Research Center for Molecular Medicine of the Austrian Academy of Sciences in 2015. In 2020, he became a full professor at the University of Vienna, where he holds a joint position at the Centre for Molecular Biology (Max Perutz Labs) and the Faculty of Mathematics.

André Rendeiro

André Rendeiro’s Research Group focuses on networks at the level of cells, groups of cells, tissues and organs. The researchers seek to understand the architecture of various organs; not just their anatomical layers, but from the level of cells to the level of tissues and, eventually, to the level of the entire organ, and how the architecture based on these levels influences health and diseases. To this end, they use data on RNA and metabolites as well as information on the position, type and organisation of the studied cells provided by high-resolution imaging. Combining this information with demographic, biographical and clinical data of individual patients helps to explain the formation of tissues and the development of diseases.

Rendeiro completed his doctorate in molecular medicine at the CeMM in Vienna. In the laboratory of Christoph Bock, he developed methods for cell profiling with single-cell resolution, which he then applied to leukaemia. Between 2020 and 2022, he developed computer-assisted methods for analysing complex imaging techniques at Weill Cornell Medicine in New York. He also worked on the first tissue-based, single-cell mapping of lung diseases during the COVID-19 pandemic.

Julia Guthrie

Julia Guthrie and her team study networks at the population level using anonymised data from medical records of up to millions of patients. They seek to understand the basic molecular connections between different diseases in order to find new fields of application for existing drugs.

In one of her most recent studies, Guthrie classified more than 180 rare auto-immune and auto-inflammatory diseases, such as inflammatory bowel diseases or multiple sclerosis, based on the gene mutations causing these diseases and their effects on molecular interactions in the human body. Together with her colleagues she has shown that these diseases can be categorised in groups defined by similar interactions between molecules in the patients’ metabolism. When comparing their results with clinical data, they were able to prove that patients suffering from diseases of the same group responded to the same drugs. This could open up new options for therapy.

Guthrie also studied the connection between chronic inflammatory bowel diseases and genetically predisposed defects in the immune system. She was also involved in the development of the virtual reality platform for the visualisation of complex networks that was developed by Menche’s Research Group.

Guthrie studied molecular bionics and molecular medicine in Budapest, Uppsala and London. She completed her doctorate in the field of rare diseases and network medicine as part of Kaan Boztug’s group at the CeMM, in close collaboration with Jörg Menche’s group.

Cross-Sectional Group for Visualisation

The Cross-Sectional Group for Visualisation aims to re-think the future of data research and exploration and deals with the issues that result from the visual complexity of large networks. The latter’s high number of nodes and edges makes it impossible to illustrate them in a comprehensible way on paper or on screen. The members of the Cross-Sectional Group thus work on developing different digital tools to enable other scientists to study data in a new way, using virtual reality on their software platform “VRNetzer”.

They make it possible to experience networks in a three-dimensional way, allowing people to move through them in the virtual space and enlarge them to better see the details. These visualisations have already been exhibited at the Ars Electronica Festival and the Ars Electronica Center, allowing the scientists to make their research accessible to the public.

The Cross-Sectional Group also deals with machine learning algorithms that can be used to manipulate large networks in virtual spaces in real time and study them. The Group also works on network-based large-language models of the same kind as ChatGPT’s algorithms. These are intended to help researchers define hypotheses, as they can use a chat interface to learn more about a network and then use their expertise to assess the information gathered.

What does the new Ludwig Boltzmann Institute for Network Medicine mean for you and your work?

Menche: The Institute’s goal is to go beyond academic proof-of-concept studies and translate the knowledge gained from basic research into practical applications. As soon as we have understood something fundamental in basic research, we can start thinking about how it can be translated into more specific clinical questions in cooperation with partner organisations.

That’s the Institute’s mission as defined by the Ludwig Boltzmann Gesellschaft. Together with long-term funding, organisational support and academic leeway, this allows us to tread new paths in research and achieve an unprecedented impact for society. This impact includes not only improved or new types of therapy, but also science communication and even art.

As an Institute, we have more resources and a greater reach among partner organisations than a single research group. As Scientific Director, I’m looking forward to tackling larger joint projects with all four Groups at the Institute and finding out how to best organise our inter- and transdisciplinary collaboration.

How does the cooperation with colleagues from other disciplines at the Institute work?

It’s not that easy. Networks provide us with a common language that people from various backgrounds share and use to develop shared concepts. Networks are relatively easy to understand, even for researchers who have not worked with them previously.

A special aspect of my Research Group and the new Institute is the involvement of artists and other design experts. They are fully integrated into the research process. Artists have been working with virtual reality for decades and are thus experts when it comes to technical and design aspects. They also approach complex problems differently. In a way, they are more audacious. That’s exciting and inspiring for me as a researcher.

What do researchers in other fields think about your research?

I think that network methods have already become a standard tool in biology. In the field of medicine, thinking along these lines is still considered rather avant-garde. I think some fields of medicine understand the molecular mechanisms at the root of diseases better than others. When I give a presentation at a medical congress, I see that the network approach still surprises some people.

When cooperating with experts from other disciplines, it is important to create opportunities for exchange. The University of Vienna is currently building a new research association dedicated to the topic of health in society. The idea is to connect researchers from different faculties. I am part of this new association’s steering committee. I think it is essential to understand that health has several components, not just the medical one, and influences many spheres of life, as we clearly saw during the COVID-19 pandemic.

What are the limits of network medicine?

My goal is to move from a static perspective and unchangeable networks to more dynamic networks whose nodes and edges may change over time. Often this aspect does not receive enough attention.

If the protein-protein interaction network is disrupted, for example, some interactions may stop as a result, while different ones emerge. A drug also constitutes a dynamic intervention in the network of the body. What consequences do such changes have on the network as a whole? How can we model them mathematically and simulate them on a computer?

Another question that still remains unanswered concerns the correct modelling of complex interactions between two or more molecules. To date, many of these more complex relationships were left out of the models so that the latter could focus on the essential aspects. But these complex interactions might play an important role that we have not yet understood.

We want to address these problems in different ways at the new Ludwig Boltzmann Institute for Network Medicine.

What does your work involve?

Hütter: I work with networks consisting of thousands of nodes and edges. It is difficult to illustrate them on a computer in a way that is easy to understand. I want to use virtual reality to make them better accessible and allow for interactions using natural gestures instead of keyboard and mouse.

The algorithms we use allow not only for 3D illustration of the networks, but also for grouping of their nodes according to different criteria. For example, we can look at a disease network in which the nodes represent the different proteins within a cell, while the edges represent interactions between them. We can group the proteins according to diseases so that those associated with the same disease are positioned closer to one another within the network. This way, we can make complex relationships easier to understand.

What is the Multi-Media Kitchen?

The Multi-Media Kitchen is a multi-functional space where we work together in the same virtual space, do professional audio and video recordings and experiment with various technologies.  I use this space to work on my visualisations, which help us to illustrate complex networks and support our external communication with other researchers and the general public.

What’s the role of art in your work?

Art plays an important role in our work. Coming from various backgrounds including architecture, graphic design or virtual and digital art, many members of our Research Group have some experience with visualisations. Again and again, we have great opportunities to engage in art projects that also go beyond our research. Our installation Entangled Realities was exhibited in the Ars Electronica Center’s Deep Space 8k in Linz as part of the 2021 Ars Electronica Festival, for example. This art installation illustrated the diversity of the complex systems that are part of our environment as humans. At the 2022 Festival we contributed the mixed-reality exhibition The Shape of Things to Come, which dealt with the future of our environment. Connected – How the world is more than the sum of its parts is now also part of the Ars Electronica Center’s permanent exhibition. This project vividly illustrates the complex networks that are part of our world.

All members of our Research Group are involved in these projects, performing different tasks such as creating art, giving expert scientific advice, organising exhibitions and supervising installations. You could even say that these projects are a form of team-building. The topic of art resonates in everything the Group does. I’m glad that we take it seriously and also allocate resources to it.

What are your hopes for the future of your research work?

Our goal is to create a common platform as a tool for studying huge data sets and networks using virtual reality and algorithms to analyse them – a tool as useful as the microscope is for studying the physical body. The skills of man and machine can be used to complement each other when it comes to recognising patterns within the data, so that new hypotheses and research questions can be identified.

Together with my colleagues, in particular our digital artists Sebastian Pirch and Martin Chiettini and our data scientist Felix Müller, I work on our “VRNetzer” platform. The VR software we use is freely available online and was also published in Nature Communications in 2021. We are constantly developing it further. We hope we’ll succeed in making this VR platform a widely used standard for data exploration that supports the work of researchers around the world.