NCRI Cancer Conference 2010, 7 November 2010, Liverpool

Professor John Heath – University of Birmingham, UK

Computational modelling of signalling pathways

I will be chairing a session called Computational Modelling and Pathways which is to do with the intersection between computer science, mathematics and cancer pathways and how we can use modern computing methods to analyse how cancer cells behave.

What advances have been made in this area?

From the cancer biology point of view I think we are in exciting times because there are three different things that are beginning to kind of converge. So from the cancer point of view everyone’s now beginning to believe that every tumour has its own story, its own narrative, its own history and its own future. So there isn’t a kind of one size fits all solution, and that poses a dilemma for the clinician because they have to be able to make a therapeutic judgement based on how they think the tumour is going to behave. The second strand is that we now know that mutations in particular pathways seem to be extremely frequent, if not ubiquitous, in tumours but these pathways are incredibly complicated and contain many hundreds, if not thousands of components. And that poses a problem for the biologist because they’ve got to be able to understand how these very complicated pathways function and also how the mutations impact on the behaviour of those pathways. And the third strand is where this particular session comes in, which is in the last three, four, five years, there has been a major series of developments in both the mathematics and computer science communities where tools have been developed which turned out to be extremely appropriate for analysing these cancer pathways. So this session is really about how can we bring these two quite different communities together, on the one hand the computer science community who have developed all these nice tools, and second the cancer biology community who’ve got these very large data sets and this very pressing, therapeutic need to develop predictive tools for looking at how tumours will behave based on their individual characteristics.

How will the advances in computer science help?

I think there are two different ways, one of which is that we have been able to make, at the moment, small scale computer models of cancer, and when I say computer models what they actually are are computer programs that behave exactly like the pathway that we are looking at. What you can then do is a couple of things of interest, one of which is you can run the program on the computer and whereas in the lab it would be very difficult to look at every conceivable experiment you might want to do, on the computer, because they don’t get tired, bored and go home, you can analyse every potential outcome and then you can use that information to look at, or select for, experiments that you might want to do. The more exciting opportunities are offered by a computer technique called model checking, and what that allows you to do is to ask questions or scenarios or hypotheses about how the pathway might work. Say, for example, what are all possible combinations of events that will allow a certain particular kind of outcome to happen? And that means that the computers effectively take us beyond the realm of the individual scientist in terms of their ability to analyse very complicated processes. So that’s the exciting thing in terms of the computer science at the moment.

How could these advances help in the treatment of cancer?

Well the ambitious end game obviously would to be have a sort of program on every clinician’s laptop that would enable them to plug in the particular features of the tumour in the patient that they are studying, run the scenarios, look for drug responses, work out what the best possible combination of drugs or individual drug regimes might be for that particular individual patient, and allow the computer to predict how the tumour is going to respond. Obviously what we have to do is to go from the bench scale, lab scale, analysis of pathways into the, if you like, clinical scale and confront real tumours with these kind of programs rather than the kind of model systems that we’re looking at at the moment.

What issues must be overcome before this can work in practice?

It’s really two things: it’s a problem about software that deals with very, very complicated systems, so where we are with the software at the moment it’s been designed to work with relatively small scale systems, and one of the big challenges is to scale it up, to deal with real scale biological data. And that inevitably means there will be hardware issues in terms of the computing power that you need to make these models run; but the two things work together because as the software becomes more powerful the computing hardware required actually becomes less powerful. So we hope that one day it will run on a laptop.