Professor Jaroslav Koča got the conference off to a colourful start with a talk illustrated by animated molecular models of viruses showing how complex it can be to assess which potential drugs will effectively block a virus’s ability to bind to a cell and begin the process of replicating itself. In the past, computational biology has attempted to match the physical shape of molecules to that of the “docking” locations on the virus: this “key and lock” approach involved testing catalogues of millions of rigid 3D models of drug molecules against the 3D model of the virus. Although each simulation typically requires 1-10 minutes of CPU time, the task can relatively easily be spread across a large number of parallel CPUs provided by an e-science grid.
However it has now been realised that many molecules contain chemical bonds that can rotate, thus changing the shape of the molecule. Rather than just testing each molecule once, many different rotations (twelve per flexible bond even if only rotations by multiples of 30 degrees are considered) of many different bonds may need to be tested. Fortunately the increase in the size of the problem is limited to some extent because some combinations of rotations will create stresses within the molecule that will make it twist back to a less stressed shape. Finding these locally stable combinations of rotations is another complex simulation problem requiring large CPU resources: it can also be parallelised but requires rather more communication between the parallel processes.
While the network requirements of these drug modelling applications remain within the capacity of current research networks (CPU capacity is more likely to be a limit), other areas of biology such as genome sequencing are already generating raw data faster than networks can extract it from the equipment. Thus literally vital work on counter-acting diseases and poisons is challenging the technologies we can provide.