Understanding vast amounts of data from experiments is always a challenge, and sometimes we can’t see the wood for the trees – or see what we wish to see rather than what's really there. No matter how powerful computers become, it’s the maths not just the hardware that reveals what is happening under the surface...…
Article by Peter Barr
Understanding vast amounts of data from experiments is always a challenge, and sometimes we can’t see the wood for the trees – or see what we wish to see rather than what's really there. No matter how powerful computers become, it’s the maths not just the hardware that reveals what is happening under the surface...
One of the biggest challenges in computer science today is how to make sense of the fast-growing mountains of data generated by research across a range of different disciplines, including social sciences, life sciences and business – looking at everything from how people move around cities and buildings, the behaviour of insects, biochemical pathways
and nerve cells, to project management, logistics, financial analysis and traffic control.
According to Professor Jane Hillston, SICSA's theme leader for modelling and abstraction, the use of advanced mathematics – particularly appropriate abstractions – makes a critical difference in the analysis of data, sometimes speeding up the process from several hours to a matter of seconds, at the same time as improving understanding, validation of theories and the accuracy of predictions.
Trying to model a system may involve hundreds of interacting variables and millions of states or “possible configurations” of the system. When the number of states gets too large, there can even be what mathematicians describe as “state space explosion,” and modelling becomes extremely difficult. If you tried to analyse every single combination of factors, the process would be highly complex and take lots of time, and the final results may not seem very clear. But if you eliminate the most unlikely outcomes and assume that very large populations will behave in more predictable or “average” ways, the problem becomes much simpler to model and easier to understand. In other words, behaviour can often be less random and more “systematic” than we may at first believe, and understanding this in one application – in nature or human behaviour – can help in many other different spheres by developing appropriate models to map that behaviour.
For example, studying problems in systems biology can inspire the development of a way of modelling which helps design buildings, to make them safer in emergencies such as fires. The two applications may seem very different, but modelling the behaviour of cells in an organism can be regarded as similar to the ways groups of people behave in a designed or engineered environment such as a building, or what Hillston describes, in this case, as “using a dialect of process algebra developed in the context of biology to analyse crowds.”
Hillston initially trained as a mathematician. At that time, using a computer was a novelty – simply a tool which she used overnight to batch-process large calculations. After graduating with degrees in mathematics from the University of York and Lehigh University in the US, she worked for Logica Financial Systems in London, and became increasingly interested in how computers work, ultimately leading her to Edinburgh to do a PhD in computer science. Her specialist subject is the development of languages to sit between computer systems and mathematics, “making it easier to construct and learn from mathematical models about the dynamic behaviour of the system.” In this area of performance modelling, she explains, the focus is on the timeliness of behaviour and efficient use of resources rather than on the correctness of systems with respect to specified behaviour.
Having established herself as a computer scientist, Hillston is now branching out in different directions, with an emphasis on systems biology. Scotland is developing a strong reputation in systems biology, and modelling and abstraction has a large contribution to make, she says, moving beyond its traditional role in computer science, where it is used to check the behaviour of systems, to help marshal the increasingly large datasets generated by advanced experimental techniques.
“Modelling is the bottleneck in systems biology,” Hillston explains, and computer scientists who specialise in modelling and abstraction are in high demand in that sector because they are used to abstracting ideas and enshrining them in languages which help us to understand problems – working with a simple language to give a mechanistic account of what is happening. “We develop formal descriptions,” says Hillston, “to make it easier to describe systems and translate them into formal models which can be used to test and formulate hypotheses to explain the data.”
THE SICSA CHALLENGE
When SICSA decided that modelling and abstraction would be one of its four major themes, Hillston and her colleagues were completely open-minded about which direction to go. The idea was to try to identify the key challenges within the theme, and systems biology simply came top of the list, along with other topics such as ubiquitous systems modelling – for example, modelling how small embedded computers interact with each other.
Another field attracting increasing attention is epidemiological modelling. At the University of Stirling, for example, mathematical biologists are working with computer scientists to use formal modelling methods to understand disease and disease spread. Their research project has looked at the spread of bubonic plague amongst prairie dogs and at bovine tuberculosis, to investigate the effects of different interventions. The Stirling-based research team also recently organised a workshop on multi-scale modelling to disseminate their results and provide a forum for other researchers working on similar topics. At Heriot-Watt University, a different computer science-based approach to modelling has been used to study the progression of HIV infection.
“Computer science is coming of age,” says Hillston, “and playing an important role in many other disciplines such as systems biology, where the interface is a rich one and practical benefits may be achieved.”
One of the projects she is working on now is a study of the Circadian clock mechanism in plants – how cells respond to the light and dark cycle of days. Individual cells behave in individual ways in different seasons and weather conditions, and building a stochastic model (taking account of randomness) helps to analyse what is happening to identify how robust the individual cells are, with a view to developing new breeds of plants which will give better crop yields by taking better advantage of daylight. For this particular project, Hillston's team is led by one of her post-doctoral researchers, Maria Luisa Guerriero, working in collaboration with Professor Andrew Millar and his team at the Centre for Systems Biology at Edinburgh.
PRUNING THE TREE
Research in one area often has surprising implications in other fields. SICSA-funded researchers at the University of Strathclyde, for example, are developing planning techniques to help with project management and scheduling, including new techniques to handle “probabilities of success.” These techniques have a wide range of applications including air traffic control and logistics, and a key area of research is to limit the number of possibilities which must be considered, akin to “pruning a tree” to concentrate growth in the most promising direction. In planning processes such as scheduling planes for landing and take off, choices must be made between alternative orderings. Each choice and the subsequent events can be regarded as one “branch”. Important criteria such as safety and efficiency must be taken into consideration at each branch point, but there may still be many valid branches. The techniques developed at Strathclyde aim to identify an optimal “plan” or sequence of decisions, even in situations when time and probability of success may alter according to the decisions made.
What interests Hillston is identifying the key features of diverse problems, whether it is scheduling the mixing of concrete, delivering parcels or managing airports, to develop modelling approaches which have wide applicability to what may seem at first like very different processes. In her own work, for example, she has used modelling approaches developed for cells and signal processing to analyse human behaviour. One recent study looks at what happens when people evacuate buildings, and Hillston is currently talking to fire engineers to find out how she can help them give advice to designers and owners of buildings to improve their safety. This particular project was inspired by the use of personalised electronic tags which can monitor people in real time as they move around an area such as an airport or hospital, much like electronic tagging for convicted criminals – and so offer personalised advice on navigation.
PSYCHOLOGY VS TECHNOLOGY
Hillston thinks computer scientists have a pivotal role to play in multidisciplinary work – a big change from the days when they were treated as technicians by other researchers. “There was a fear we would be asked to write the software,” she continues, “but computer science can offer not just a service but a true collaboration with other disciplines, solving intellectual problems and stimulating developments in other fields.”
SICSA has encouraged collaboration between different disciplines right from the start. “We always try to think creatively about who talks to whom,” she says. “And this creates a lot of possibilities for cross-fertilisation of ideas within and between different disciplines.”
In Hillston's view, this translates into a “style” of intellectual enquiry rather than simply a collection of methods, and researchers who are exposed to this kind of collaborative environment often find themselves changing direction when they see their work from different perspectives.
Having started off in pure mathematics and gradually become more involved with computers, Hillston is well placed to describe how informatics has evolved through the years. Formal modelling has emerged from its use in computer system testing and design to become a useful tool in other disciplines, but no matter how spectacular, technological progress is not the whole story.
“There has been a change of ethos rather than technology, psychological rather than technological,” she says. “Computer science has gained more confidence as a discipline and realised it has a huge contribution to make to other disciplines.”