Frank von Delft: X-rays for drug discovery

Q: Can you tell us about how X-rays are currently used to develop new drugs?

FVD: The amazing thing about X-ray crystallography is that you get an extremely detailed picture of the protein molecules you want to look at, literally the position of every atom. It has been recognised for decades that there is an opportunity to design molecules that would bind exactly to the (protein) pocket you want to. Because you see every atom, you can also visualise in 3D the position of the atoms of the compound you want to bind, you can model in principal where exactly it sits.

That has been the ambition of crystallography all along, but in practise, it is much more complicated and the computational techniques to do this prediction do not really exist, or aren't mature yet. Nevertheless, the crystallography industry has learned to do this a long time ago, and has pushed very hard to integrate the crystallography observing compounds and structure with the chemists and the biologists who do the tests, to have an ecosystem of developing compounds.

This has worked very well and they've been at the forefront of pushing the field of crystallography. This high-throughput approach (has produced) a very quick turn around on (protein) structures, so they can tell the chemists what molecules to make.

Q: How does your research group's approach differ?

FVD: About two decades ago, people started thinking how we could be much more rational about the process of developing a compound drug. Rather than discovering (drugs) by looking at libraries of millions of compounds, could we take much smaller compounds, which are called fragments, and see how they bind to the protein?

Then when you have seen them binding, you can tell your chemist: there is an opportunity on this side of the pocket, can you build something there? These are called fragment-based approaches.

The problem is that these fragments, because they are so small, bind very weakly, so you have to observe them in a crystal, in a structure; simply testing them by physical techniques wouldn't work. Initially, people tried to do crystal structures of every compound and this turned out to be a lot of work. They then backed off and said, 'Lets pre-screen these fragment libraries (approximately a few thousands) using bio-physical techniques, and then weed them out.'

If you find the ones that bind, you can look at their structure and build on that. The crystallographic experiment is by far the most sensitive way to see binding, so you would really like to do your primary screening in crystallography. We have set out to try and change the experiment, which used to take literally months, to do a few hundred compounds in a matter of days, to do hundreds or thousands of compounds in the crystal structure experiment.

Q: How are you working to achieve these sorts of aims?

FVD: We are talking about big logistical experiments and clearly this is a job that needs geeks and engineers as well. We decided to partner with Diamond Light Source, which is Oxfordshire's own 'Big Science' facility in Didcot, and to combine the medicinal chemistry skills available at the University with our extensive experience at this big facility.

I took charge of one of the experimental stations - we call them beamlines - which serve the whole UK community of crystallographers. This beamline was built with high-throughput in mind, to increase the capacity of crystallography in the UK.

We partnered with them and said: lets develop this experiment at this beamline. Now I can draw on two groups, my group here in Oxford and the group at Diamond, which is a very talented set of individuals. We have looked at all the steps of the process from sample preparation all the way to data collection, just to shrink the process. We have robotic sample mounting, compound collections, and we have configured the beamline so that we can put five hundred crystals into it: it will (then) run overnight and just collect (data) on its own, something that had to be manually supervised and now happens automatically.

Q: What are the most important lines of research that have emerged over the past 5-10 years?

FVD: In this field, I distinguish two. One is the idea that you can use fragment-based approaches, to have this ultra-rational drug design process: that it is feasible to go from something extremely weak (which most chemists would have dismissed in the past) and then build up in steps to come up with a really potent molecule. This has been shown to work and is now a quiet revolution; it's been gradually adopted as a standard product tool kit of all medicinal chemists. I think that it is a big deal and it is not really recognised as such.

Then on the other side, on the crystallography side: this experiment has been transformed in the last two decades. The intensity of X-rays has gone up massively at the synchrotrons. As for the detection technology, we can now contemplate doing an experiment about two orders of magnitude faster than before: something that used to take twenty minutes now takes less than a minute to measure. The software available has become so powerful and so easy to use that we can deploy it for hundreds of analyses rather than for one by one painstakingly.Robotics has changed too.

The fact that we can even contemplate doing five hundred experiments in 24 hours, never mind that we can actually do it, is astonishing on its own.

Q: Why does this line of research matter and why should we fund it?

FVD: Molecules that bind potently to some target, some protein, turn out to be extremely powerful research tools for basic research, and also for clinical or pre-clinical research. In fact, if you look at the citations figures for these molecules, their citation numbers rival that of the top researchers in the world.

The problem is that they are very rare and very expensive to make.

If we look at what happened with the human genome in the 90's, it was really hard to sequence one gene. Then suddenly there was a transformation because of technology and they said, 'Oh hang on, we can actually do the whole genome!'

We need the same thing with molecules. We have a few molecules, and we need to start thinking about how we would get molecules against every one of the twenty thousand proteins in our body. We need to transform the technologies to actually transform the imagination available around the research.

Q: How does your work fit into translational medicine within the department?

FVD: It is well recognised that the big problem with drug discovery is knowing which targets to work on. This is not actually a problem of economics or the pharmaceutical industry, it is a problem for the patients.

The patients are the ones who do not have effective treatments for the diseases. If you want to start validating targets, you will need these molecules, these research tools. If you want to use these research tools, you will have to get much better at developing them, which includes getting much faster at the design.

This is not just about generating molecules: this is about testing them and the whole ecosystem of developing this. So I am working very closely with several colleagues at the SGC, at the TDI, on the computational, the chemical and the assay technologies that are involved - people like Paul Brennan, Brian Marsden, Nicola Burgess-Brown, Oleg Fedorov, to se how we can transform the process. The medium-term goal is to make a 'go to' technology, rather than something that you may aspire to eventually.