Modelling the insulin receptor
But in order to design a drug to mimic insulin, we first need to know exactly how it binds to its receptor.
Dr Tristan Croll from Queensland University of Technology is using NCI’s supercomputing facilities to simulate the insulin receptor molecule at the atomic level. The project is funded by an ARC Linkage grant.
“I’m interested in figuring out exactly how insulin binds to the receptor, and how the receptor changes shape and activates in response,” he explains.
Early on in the process, Dr Croll discovered something surprising that stuck a spanner in the works.
“When we tried to create a computer simulation, the receptor did some really strange things,” he says.
“We realised that the problem was that the crystal structure we have of the receptor is quite low resolution. In fact, it took quite a heroic effort from CSIRO to get it to crystallise at all.”
So Dr Croll teamed up with Associate Professor Mike Lawrence from the Walter and Eliza Hall Institute to refine the structure and fill in the gaps. The researchers took existing simulation tools and created new functionality.
“The basic technique is called Interactive Molecular Dynamics Flexible Fitting. Like in normal crystallography, you create a map of where the atoms in the protein should be – but then you use the map to define steering forces guiding an interactive molecular dynamics simulation.”
The program can also be used to give a “best fit” to low-resolution, flexible components such as the sugar chains that decorate many mammalian proteins (including the insulin receptor). Such structures are often left out of crystal structures due to lack of atomic-scale resolution, yet are important to include in simulations to maximise their realism.
“The whole thing is interactive,” says Dr Croll. “You can play with the structure in three dimensions. You can grab an atom, pull it, and feel it pulling back at you.”
The ability to move atoms in this way has been around for a while, says Dr Croll, but technical limitations have hindered its usefulness.
“Up until now it has been a somewhat ‘blind’ process, with no easy way to guide it with outside information. The biggest problem was that it was very easy to pull too hard on the atoms, and that would do strange things to your protein structure.
“We’ve applied lots of constraints to ensure the structure retains its natural characteristics.”
Dr Croll says the new technique could be applied to any protein but is particularly useful for those “difficult” cases where the crystal resolution is relatively low.