Professor Graham Richards is Commander of the Order of the British Empire and former Head of Chemistry at the University of Oxford. He is a pioneer in the field of computational chemistry and computer-aided molecular design, particularly as it applies to the pharmaceutical industry.
To me, Professor Richards is just Graham. He is Graham, my chemistry tutor, my mentor, the person who introduced me to the world of biotech, and the person who helped restart the Women’s Boat Club. I am deeply honored to call Graham a friend and am grateful for his willingness to share both the history of Computer Aided Molecular Design and his views on the future of the field.
Professor Richards’ early work was on ab initio computations and spin-orbit coupling. He then turned his attention to pharmaceutical applications, introducing many techniques that are widely used today such as molecular graphics, where he produced the first colored images, and distributed computing. In fact, with the latter, Professor Richards organized the Screensaver Lifesaver project which involved more than 3.5 million personal computers across 180+ countries. The project screened billions of compounds in search for drugs to treat cancer and to protect against anthrax and smallpox. Currently, Professor Richards is Founder and Chairman of Oxford Drug Design, which focuses on designing novel antibiotics.
Michelle Dipp: Tell us about the beginning of Computer Aided Molecular Design (CAMD) at Oxford. What sparked your interest in the field?
Graham Richards: I am very old, so some of this is almost pre-history. I started research in Oxford in 1961, and my project involved doing some difficult integrals. I was then one of a relatively small number of people who realised that integrals could be done numerically on this newfangled thing, the computer. I did this on the Ferranti Mercury computer, probably the best in the world at the time: a 32,000 valve based machine, the size of a small house yet powerful enough to enable Dorothy Hodgkin to produce the structure of penicillin. I then went to Paris to use the new IBM 704 machine, but my research was on small molecules.
Then, out of the blue, I received a letter from Jim Black (later Nobel Prize winner Sir James Black) who had found the beta blockers for Imperial Chemical Industries. He was then working for Smith Kline & French looking for blockers of histamine H2 effects. He enclosed a paper suggesting that the H1 and H2 effects were due to two conformers of histamine. We showed that this was not the case, but it moved my research into the area of computational pharmacology. Initially, colleagues thought we were mad, but as computational power increased, we became increasingly relevant.
In 1989, I founded Oxford Molecular Group Plc which went from a $500,000 start up to a $900 million public company completing seven takeovers in the US before being sold for rather less.
Michelle Dipp: What types of computers were you using? How has cloud computing changed CAMD?
Graham Richards: The field has progressed as computational power has developed. A big step was the Evans and Sutherland machine and the introduction of computer graphics. We produced the world’s first colour graphics using a black and white screen and photographing through colour filters.
At the end of the last century, I set up a screensaver project which had 3.5 million participants from more than 180 countries. This gave hundreds of thousands of hours of free computing time enabling us to screen literally billions of small molecules as inhibitors of known targets sites on proteins. This need for computational power has been satisfied by cloud computing which is inexpensive and very convenient.
Michelle Dipp: What data did you rely on in the early designs of CAMD? What other types of data are now being used in CAMD?
Graham Richards: Ideally (and it is increasingly the case), we use the crystal structure of the target protein with inhibitors bound in. Failing that, we predict the target structure (and for the potential drug) use data on shape, charge distribution, solubility and transport properties. Increasingly important are genetic data which permit understanding of which patients may be targeted and which other biological responses might be invoked.
Michelle Dipp: Why has it been so hard to apply artificial intelligence technologies for the screening of novel biological targets?
Graham Richards: As long ago as the early 1990s I started using neural networks in the early 1990s, and these methods, now classified as machine learning, work very well for small molecules aimed at specific targets: in many ways just a superior sort of statistics, but without necessarily understanding what is actually happening at the molecular level. As the systems being studied get more complex, bigger molecules, parallel pathways, access problems, etc., the methods are struggling, but I have no doubt they will succeed in time.
Michelle Dipp: What are the key advancements in CAMD that you are most excited about?
Graham Richards: My belief is that in the near future a coalescing of genetics and Drug Design will be a key step: drugs not for entire populations but only for those for whom they will work.
In my own case, our company, Oxford Drug Design, is concentrating on antibacterials, an area not favoured by major pharma companies. There are huge potential problems with antibiotics failing to work and inevitable pandemics exacerbated by the fact that more than 90 percent of antibiotics are used in agriculture and fisheries.