European grant for TU Delft research into virus spread

The European Union has awarded TU Delft researcher Piet Van Mieghem an ERC Advanced Grant. This grant, worth 2.5 million euros, was established to encourage groundbreaking, high-risk research and is mainly awarded to academics who have a significant track record of research within their own field. With his research, Van Mieghem hopes to find an answer to the question of why the western world was not successful in preventing the spread of coronavirus. 

A research question that is close to the heart of Van Mieghem, Professor of Network Science at TU Delft: ‘In recent years I have been using simple models to study the spread of epidemics within a network structure. However, with the coronavirus pandemic, these theoretical insights suddenly became reality.’

Shortcomings

According to Van Mieghem, the all-consuming pandemic brought to light a number of shortcomings within traditional epidemiology. He explains: ‘If you look at the spread of an epidemic purely from a physics perspective, two equally important processes present themselves: the viral process – the way in which the virus jumps from A to B; and human behaviour – the way in which A and B make contact. Recently the viral process has been thoroughly mapped; we have been able to measure with a fair degree of accuracy what the chance of transmission is, as a function of the distance A and B. Moreover, we understand how long A and B need to remain in proximity to allow enough virus particles to jump over. Because you become ill when the dose of virus particles in your body is sufficiently large. So the moment when A and B have been close enough for long enough, is when the conditions for a successful viral contact are met.’

Measurement data

Yet Van Mieghem feels that almost all attention has been paid to just one aspect of the physics. ‘This despite the fact that in Network Science, we study precisely the dynamic processes in graph – network topologies (how nodes within a network are arranged and connected). Take for example a time-varying contact graph. You can derive such a graph by looking at where people were located during a short time span. In the case of Covid-19 this remained largely unexamined, because to do so requires lots of measurement data, and these are hard to come by.’

App

According to Van Mieghem, a mobile application could offer a solution for this. ‘These days everyone has a mobile phone. That means that your location, to within an accuracy of around a hundred metres, is always known to your telecom provider. And then of course there are all those customer data that are collected, that advertisers make such clever use of. Hardly anyone objects to such data gathering, yet privacy is regularly used as an argument against apps that can tell us more about where the virus is located. This attitude often only changes when the problem comes very close – when the destructive consequences have become clear. Why is the temporary storage of location data, to within a few metres accuracy, in well-protected data centres and for the sole purpose of calculating coronavirus infections, not an option?’

Markovian

The process by which a virus spreads is usually described using a probability model, explains Van Mieghem. In practise this kind of model is used to simulate and analyse systems or phenomena in which the situation wholly or in part depends on chance, for example the probability of cloud formation or, in the world of finance, when it is advantageous to buy or sell shares. So it sounds only logical to map out coronavirus using probability models. Probability theory often makes use of a Markov chain: a stochastic model (a succession of chance results) in which the current probability depends solely on the previous state in the process. 

Van Mieghem: ‘Besides linearity, a great Markovian property is that the time between two consecutive process states is distributed exponentially. But it turns out that this does not apply to coronavirus. When you look at coronavirus, you see that the degree of infectivity is not constant, and even varies during the entire course of the disease. So it seems that the Markovian theory doesn't apply to coronavirus, and we need a new theory.’

Non-Markovian

Gaining insight into how the virus spreads and how it causes harm already represents a good deal of progress. But what the TU Delft network researcher plans to achieve with the ERC Advanced Grant he has been awarded goes even further than this. Van Mieghem: ‘I have spent the last decade learning how to understand Markovian models and what such a model says about the course of an epidemic within a specific network. Because Covid-19 behaves so differently, I want to develop a new theory: the theory of non-Markovian epidemic processes within networks. Something about which little has been written, even today. Using the virus infection and recovery time, I hope to be able to tell how long a pandemic will last and when a peak will occur. Subsequently I will combine all the available measuring techniques in order to construct the best possible time-varying contact graphs. And finally I hope to be able to say something about how accurately an infection can be predicted.’ 

Ultimately Van Mieghem hopes that all these findings – gathered together under the name ViSioN – can be used to predict, manage and control future epidemics. 

Smart networks

Van Mieghem is not the only scientist at TU Delft to receive an ERC Advanced Grant. Bart de Schutter (Faculty of Mechanical Engineering, Maritime Technology and Materials Science –
 3ME) has also received a grant. De Schutter focuses on the control of complex large-scale networks. De Schutter: ‘A specific example is (future) smart energy grids, where you will be dealing with a complex local combination of energy generation, storage and energy consumption. In such a system, you want the local energy generation to match the local energy consumption as closely as possible. A simple example is a washing machine that only switches on when the energy consumption in the system is low.’ In addition to smart energy networks, we will also increasingly have to deal with smart transport networks in the future. However, the online control of these kinds of large and complex networks is still far from optimal. Researcher Bart De Schutter wants to tackle this problem in the next five years.