Skip to main content Skip to footer

Designing better drugs, faster

Chris Skylaris, Reader in Computational Chemistry at the University of Southampton, develops quantum mechanical calculations for chemistry simulation and drug discovery. He has developed a quantum mechanical method that dramatically cuts the computing time needed for large-scale calculations – expanding research possibilities, and attracting interest from drugs companies who are looking for more accurate molecular simulation tools for their research.

The theory of quantum mechanics is necessary if you want to do any simulation of microscopic particles, he says. “You can’t do it with classical mechanics – if you want to describe how atoms and electrons move, they don’t follow the kind of motion we see in everyday objects,” Skylaris says. Therefore, any simulations require supercomputer-level compute resources, and clever theories and coding in order to run fast enough to be useful.

“But, obviously, I’m not simulating everything! My own focus in this example is to simulate how small molecules – drugs, in this case – bind to proteins. I can simulate exactly how they interact at the atomic level, and use that to design better drugs. Traditionally you do this by trial and error, by modifying the molecule in the lab, but the simulations that we’re doing, with the methods I’ve developed, give us the opportunity to include more rational drug design in the process. That can then speed up the actual testing in the lab,” Skylaris says.

Traditionally, the computation cost in quantum mechanical calculations of molecules – the amount of compute power and time needed – increases by a factor of ten when you double the number of atoms being simulated. Skylaris’s team and his collaborators, however, have developed a new method of calculation called ONETEP, which allows a linear increase in computational cost. The compute requirement grows in line with the growth in atom numbers, so twice as many atoms equals twice as much time.

Using the SES facilities is also essential to the work he is doing, Skylaris says. “You can’t run this work even on a small cluster – the proteins that we simulate have typically between 2000 and 5000 atoms, when traditional quantum

chemistry calculations look at, say, 200 atoms. With the linear-scaling we can now work with thousands of atoms, but we still need supercomputing resources to do it,” he says. The simulations have already led to some exciting discoveries. In one project, Skylaris’s team at Southampton has been working alongside colleagues at the University of Cambridge in researching the human RAD51 and BRCA2 proteins, which are implicated in breast cancer.

“The critical point in the replication of cancer cells involves the interaction of these large proteins. And this is a very difficult interaction to target – it’s difficult to develop drugs to disrupt this interaction because of the shape of the proteins,” he says.  Skylaris explains that, conventionally, a protein has a well-defined ‘binding cavity’, an indentation into which a drug can fit, like a key in a lock. But when two proteins interact, they create a smooth surface to which it is very hard to bind anything.

“You have one flat surface, with no very pronounced features, interacting with another flat surface. So chemically it is very hard to find a molecule that will bind with high affinity and high selectivity and disrupt the interaction with the other protein,” he says.

This is an important target for drug development, but it’s very difficult to achieve. However, Skylaris’s simulations have succeeded in identifying sites on this apparently featureless surface.  The ability to study processes at the protein-protein interfaces also opens up new possibilities in probing the chemical and physical processes that happen inside the cell, he says.

“There are a couple of sites – in the literature they’re called ‘energetic hotspots’ – and they actually offer the bulk of the strength when the proteins interact. We’ve been able to measure the strength of the interactions, and the next step is to develop small molecules that bind in these hotspots.”

This work has attracted the attention of drug companies who have avoided research into protein-protein interactions to date. “That’s a new thing. Companies are beginning to look at interactions as potential drug targets, when a few years ago that was just unheard of – they would just have said that the affinity and selectivity is too low. But our work has opened up the possibilities,” Skylaris says. “I have long-standing collaborations with some drugs companies. They are interested in our testing and validating these methods, so that they can

later use them for their own research. So – they want to see how well I can make predictions using these large-scale quantum mechanical applications that I develop.”

The authors would like to acknowledge that the work presented here made use of the Emerald and IRIDIS High Performance Computing facilities provided via the Science and Engineering South Consortium in partnership with STFC Rutherford Appleton Laboratory.


Project contact: Chris-Kriton Skylaris

Professor Chris-Kriton Skylaris

Professor of Computational Chemistry, University of Southampton