New tool key to cancer research

A new research tool, which could help unpick the complex cell interactions that lead to cancer, has been devised by the University’s Department of Statistics and Centre for Complexity Science.

The discovery comes at an important time for researchers and not only does it have long lasting implications for future research into cancer cells, the tool would also allow social scientists to “mime” social networking sites such as Facebook and Bebo for useful insights.

The innovative component of the research is its new “graphical models” approach, which can be used by researchers to understand a wide range of systems with multiple interacting variables. These models use mathematical objects in their application called graphics which can show the probability of relationships between each of the named variables. The flexibility of the tool not only allows it to be used in a variety of science subjects, but also has practical uses in social subjects research. For example, in the field of Economics, researchers could use the tool to understand the relationships between various economic factors and in biology, the same tool could be used to explore influential links between molecules.

Despite the progress made, the research has shown potential points of instability with the new tool. Gaining information is just one weakness, due to the vast range of probability graphs required for even a relatively low number of variables. Investigating just 14 proteins known to be implicated in the formation of a cancer, showed researchers the scale of the challenge they were facing due to the vast number of possible combinations.

A possible solution of this problem was pioneered by Dr Sach Mukherjee of the Department of Statistics and Centre of Complexity Science who lead a team of researchers to produce a method where by current knowledge in mathematical analysis can cut through vast complexity’s of the above type using a mechanism called Informative Priors. Informative Priors can be seen as mimicking how human researchers learn from data, in the way they interpret new information in light of what is already known.

Indeed, using the method, researchers investigated the 14 protein network again, and created a mathematical tool which was able to incorporate all possible interactions, including their limits, complete with a calculation of instance’s probability on these specific proteins and so this allowed a quick and accurate understanding of the probabilities of the interactions between each of the fourteen proteins.

Furthermore, the approach was even able to learn how to cope with misconceptions in current understandings of particular networks, as it was designed to overturn and reject any data included in the Informative Priors that were at odds with any newly observed data.

As a result of the progress made, analysis of these network models was more able to cope with and resolve complex interactions over its simple, correlation-based predecessors. This lead to more accurate results than analysis which incorporated no prior understanding of the network (dubbed flat priors).

The researchers have exciting plans for the future which use their new technique, including the examination of a network of proteins implicated to be behind breast cancer.

However, they are also keen to develop the flexibility of their new tool, especially in the realm of social science where they are working towards mining vast amounts of data from social networking sites.Through this, they hope to use these methods to gain enlightened understandings of social interactions and relationships, within society at large.