Is there a future for computational chemistry in drug research?

Improvements in computational chemistry methods have had a growing impact on drug research. But will incremental improvements be sufficient to ensure this continues? Almost all existing efforts to discover new drugs depend on the classic one target, one drug paradigm, although the situation is changing slowly. A new paradigm that focuses on a more systems biology approach and takes account of the reality that most drugs exhibit some level of polypharmacology is beginning to emerge. This will bring about dramatic changes that can significantly influence the role that computational methods play in future drug research. But these changes require that current methods be augmented with those from bioinformatics and engineering if the field is to have a significant impact on future drug research.
An accidental dinner-party meeting with Corwin Hansch, the father of the concept of quantitative analysis of the relationships between chemical structures and biological activities, was a catalyst for Yvonne Martin's career in computational chemistry in industry. "In high school, I had recognized that I was interested in solving biological problems using chemistry," she recalls. "But at the time that I met Corwin, my involvement in structureactivity relationships was from my pre-Ph.D. work and through my role as a drug metabolism expert at Abbott. During our conversation, I realized that I had to learn more about quantitative structureactivity relationships and bring it to the company." Since this point, Martin has been involved in computer-assisted drug design at Abbott for more than 30 years. So what has kept her motivated to work in this field for so long? "Simply, it is the continuing challenge to better use the computer to understand structureactivity relationships and to solve problems faced by medicinal chemists in their search for the next new drug," says Martin. "I still remember the day, perhaps 15 years ago, when a chemist came to my office to report on the biological activity of a compound that my models had predicted to be inactive. My heart fell, because he had a big smile on his face and I knew that he was a doubter of the model or he wouldn't have synthesized the molecule. However, his words were: 'You were right. The compound is inactive.' From that time on, he considered the models as he decided which of the possible molecules he should make." This kind of success with models is an aspect of her role that Martin finds particularly rewarding. "Nevertheless, it is frustrating that we don't yet have computational methods that solve most of the problems faced by a medicinal chemist," says Martin. "For example, we wish computational methods could accurately forecast the affinity of a ligand for a protein, or accurately predict the water solubility of compounds." These problems are sure to keep computational chemists occupied in the future, and

there are plenty more ambitious goals if they are addressed, such as predicting how a compound would affect a network of proteins. An important factor in tackling such challenges as a scientist, in Martin's view, is to publish, and to publish work that is helpful and interesting to others. "Yet another paper on some subject has little impact unless it shows all previous work was flawed or that a superior and simpler approach works better it is better to publish something completely different or to write a critical review on the topic," she considers. "I've learned that the power of publishing or presenting at meetings is that it introduces you to other scientists who can provide unpublished insights, helpful criticisms and interesting conversations," Martin says. "A second key lesson is to continually expand your horizons to ancillary disciplines that could affect your primary interest, as happened for me with Corwin Hansch." For Brian Shoichet, the attraction to the field of computational chemistry began during his chemistry degree at the Massachusetts Institute of Technology, USA, in the 1980s. "I realized that the most interesting examples of molecular recognition occurred among biological macromolecules," he says. "Given my training to that point, I thought I'd go into a field where chemistry could play a role in biological molecular recognition. At the time, computational simulations were having their first flowering, and predicting how small molecules interact with proteins through techniques such as molecular docking seemed very exciting." So, Shoichet pursued a Ph.D. in molecular docking in the laboratory of Irwin Kuntz, a pioneer in the field, at the University of California, San Francisco (UCSF), USA. With this theoretical grounding, he then made what he considers to be one of the best moves of his career: a postdoctoral position in a protein crystallography laboratory with Brian Matthews at the Institute of Molecular Biology in Eugene, Oregon, USA. "Protein crystallography was going through its golden age," recalls Shoichet, "and I had come to an experimental lab wanting to be exposed to what I naively thought of as 'real data'. In fact, that was my mantra for several months, but once lab mates were used to the novelty of a theoretician wielding a micropipette, they became tired of it. One day, one of my mentors, Walt Baase, turned to me in exasperation and said: 'Data, Brian, is the weakest part of the scientific enterprise. A really good theory trumps data any day.' What he meant was that good experimentalists don't actually trust their data, not before they've looked at it backwards and forwards and sideways, because data can be very misleading and is often, in fact, simply wrong." This distrust of data has stuck with Shoichet since, and has had a key role in some of the most important discoveries made in his laboratory, including the unexpected finding that many of the false positives that plague industrial high-throughput screening assays are caused by aggregation of the compounds being screened. Odd results like these are one of the aspects that Shoichet finds most exciting in his work. "The challenge comes from distinguishing an observation that has the chance to build into something new and interesting from the vast majority of

times when 'Huh, that's weird' means that you or one of your co-workers has messed things up or got things wrong, or you're just confused," says Shoichet. The field of molecular docking is still some way from realizing its initial promise as a panacea for drug screening, but the excitement that first attracted him to it is still there for Shoichet, who now leads a group at UCSF that mirrors the path he has taken: the research is half theory (molecular docking and inhibitor discovery) and half experiment (enzymology, inhibition mechanisms and crystallography). "We try to work at the interface between the two," says Shoichet, "and occasionally, in moments of bliss, I get into the lab to do experiments myself."

Interview by David Bradley

A Chemist's Thoughts on Computational Power and the Future of 'The Chemical Web'
Computational chemist Steven Bachrach was torn between physics and chemistry. Luckily for the chemical community he found a happy medium in which to explore. Born, August 14, 1959, and graduated from the University of Illinois at Urbana in 1981, he obtained his PhD on Organolithium and Organosilicon Chemistry under Andrew Streitwieser, Jr., from University of California at Berkeley in 1985. He is currently the Dr. D. R. Semmes Distinguished Professor of Chemistry at Trinity University, San Antonio, Texas. Professor Bachrach has more than 100 research papers to his name on everything from potential energy curves of CO and Ne-He, to computational studies of pericyclic reactions between phosphinoboranes and alkenes. Can you tell me briefly about your route into chemistry? My father was a professor of chemistry, first at the University of Illinois at Chicago, and then at Northeastern Illinois University. I do not recall any time he pushed me towards science or chemistry, but undoubtedly he was an influence on my career choice. When I started my collegiate career at the University of Illinois, I was torn between
Steve Bachrach

physics and chemistry. I opted for the latter based mostly on career opportunitiesI figured there were more job opportunities in chemistry than physics. What drove you towards the theoretical/quantum end of the chemical equation? Given my interests in physics, the decision to pursue computational chemistry seemed pretty natural. I took many physics course at Illinois, and in fact my favorite course was quantum mechanics taught by a fantastic professor in the physics department. I was also fortunate that Prof. Cliff Dykstra had recently joined the chemistry department. Cliff, a graduate of Fritz Schaefer's group, was doing computational/theoretical chemistry and I joined his research group in the second semester of my junior year, and continued this for my senior year as well. We examined a new theoretical method called ACCD (an approximated coupled-cluster doubles treatment) and its application to carbon monoxide and the He-Ne cluster. This led to my first publication, a full paper in the Journal of Chemical Physics that appeared shortly before I started graduate school at Berkeley. I chose to go to Berkeley for my doctorate mainly because there were two professors I was interested in studying with, Fritz Schaefer and Andy Streitwieser. I ended up choosing to work with Streitwieser because I wanted to apply quantum mechanics and computational methods to organic reactions. As computational power rises, do you think we will begin to solve more and more complex molecules to the point where Schrodinger can be solved even for proteins? Undoubtedly computers are getting faster and algorithms are becoming more powerful. We are all studying much larger molecules than I did as a graduate student some 20 years ago, let alone even 5 years ago. Nonetheless, tackling proteins is a huge challenge primarily because of the enormous number of possible configurations. I think a different approach will be necessary to deal with the protein folding problem. However, if all we wish to do is compute

the energies and properties of a given structure, certainly that will be amenable at the ab initio level within the near future for proteins of reasonable size. You played a significant role in pioneering the chemical net, as it were, how have things changed since those heady days of the early to mid-90s? Technology has changed the playing field dramatically. Back in the beginning, we had to worry about access to the internet, let alone broadband access. And the development of XML and Java and related web languages has meant that the kinds of ideas we were developing back then have a real framework now. Culturally, things have also changed, but very slowly within the discipline of chemistry. Electronic journals are now the norm, and my guess is that print journals will begin to disappear within the next 10 years. But the current stateof-the-art in electronic publishing is Adobe's PDF, essentially just electronic print. Very little of a revolutionary nature that we were trying to do with the Internet Journal of Chemistry exists even today. The new Prospect initiative from the RSC [see the interview in issue 62 with RSC's Robert Parker] is a real step forward, but this is 10 years after we launched IJC. How have things progressed? Do you feel we are "there" yet? While chemistry publishing is better off today in terms of what is being done with the Internet, we still have a ways to go. Journal articles could be so much richer, with reusable data, and commentary, and live objects, etc. The community also has to come to grips with the cost of information, whether that is subscription-based, Open Access, or some other model.

Can you tell me briefly about your input into the InChI movement? What drivers do we need to push that into mainstream chemistry? I have mainly been a cheerleader, serving on some

committees and championing the adoption of InChI, but I have not participated in its development. The one area that I think can rapidly change the marketplace regarding InChI is if a federal agency or some international body would require the InChI as the identifier. I know that some US agencies face significant costs because they require the use of the CAS number to identify compounds. A move to InChI would save real money and allow all suppliers, agencies, and consumers equal access to the compound identifier without having to pay for what is simply a name from a third-party. How should chemists embrace Wikis, blogs, and socalled Web 2.0 to further our cause? Well, this is really an issue of culture. My personal hesitancy to adopt Web 2.0 technologies is that I don't have the time to read random thoughts by random individuals. I barely have time to keep up with the old-school (i.e., traditional journals) literature in my field. The blogosphere just seemed to me to be filled with the rantings of people who have nothing better to do with their time. Peter Murray-Rust's blog was the first to demonstrate to me that real chemistry content could be had, that interesting and novel ideas could be found and shared and discussed. So I am actually starting up my own blog in the next few weeks. I am having a book called Computational Organic Chemistry published with Wiley, to appear in July 2007. Now, this is a survey of how computational methods have been used in solving organic chemical problems. I sent in the manuscript in October, but science never stops, and so in some sense the book is already out of date. The blog will allow me to write about new articles that have appeared relevant to the topics of the book; it's a way to keep the book up to date, sort of continuous new editions coming out as needed. I will also have an ancillary web site for the book which will have links to all the cited materials that are available in some electronic form on the web and also include the 3-D coordinates of all the molecules discussed in the text so that readers can manipulate these structures in 3-D and even pipe them into their own applications. The books web

site will be http://www.trinity.edu/sbachrac/coc and the blog will be available at http://hackberry.chem.trinity.edu/blog. Simple prototypes are there right now, but the finished versions will appear over the next few weeks. I hope that these web resources will serve as models for how the Internet can really enhance scientific communication. The main hurdle for wider adoption of Web 2.0 communications in chemistry rests largely on changing attitudes in academe. Until blogs and wikis are recognized as meaningful contributions towards tenure and promotion consideration, these tools will be largely ignored.