See a Movie of AutoDock in Action Against HIV Protease [1,6 MB MPEG]
	Movie: © 2003, Garrett M. Morris, All Rights Reserved.

Introduction to Drug Design

Computational scientists conduct scientific investigations using computational methods. They typically write software programs (or use programs developed by others) that encode the physical, chemical and biochemical laws underlying some natural phenomenon. Such work can include learning more about a particular chemical interaction in the human body (such as how molecules dock with each other), or predicting the most likely path of a hurricane, or the effects of man-made carbon dioxide on the earth's climate.

We in the Olson Laboratory are using computational methods to identify candidate drugs that have the right shape and chemical characteristics to block HIV protease. This general approach is called "Structure-Based Drug Design", and according to the NIGMS/NIH, it has already had a dramatic effect on the lives of people living with AIDS.

HIV protease is a key molecular machine that when blocked stops the virus from maturing, and is thus a way of avoiding the onset of AIDS.

Once such drug candidates are identified using computational methods, they can be chemically synthesized in a laboratory, and tested in battery of assays and cell cultures. A few lucky molecules make it past years of testing for toxicology and side effects, and are tested by human volunteers according to rigorous FDA guidelines: Ultimately, the successful compound is released as a prescription drug. Our goal in the Olson Lab is not to manufacture drugs, nor is it to make a profit (The Scripps Research Institute is the largest, private, non-profit research organization in the USA) but to discover new ways of finding drugs, and in particular, to defeat HIV drug resistance.

HIV Protease ("PRO-tee-aze") has a tunnel which normally binds viral substrates, but is filled up and blocked by drugs called "protease inhibitors". It is seen here magnified about 1 000 000 times.
Image: © 2003, David S. Goodsell, All Rights Reserved.

It has been demonstrated repeatedly that the function of a molecule - a substance made up of many atoms - is related to its three-dimensional (3-D) shape. In this context, Art Olson's lab at TSRI is studying ways to design drugs based on the 3-D arrangement of the atoms in the target protein. Data accessible from Internet-accessible databases typically describe the shapes of a protein and a drug separately, but not for the two together. To complicate this problem, not all structural data, in particular candidate drug structures or variants of protein targets, is known.

Our lab is investigating how to use computers to model the forces of nature at atomic and molecular levels to predict how a drug candidate might bind with a protein. This candidate molecule should bind more tightly than the natural substrate of the protein: if it does, it is called an inhibitor. By blocking a certain protein that is part of a biochemical pathway, it is possible to counteract the effect a disease or virus such as AIDS.

Using distributed computing across the Internet, we are running a software program called AutoDock. AutoDock is a suite of tools to predict how small molecules, such as drug candidates, might bind to a receptor of known 3D structure. The very first version of AutoDock was written in 1990 by Dr. David S. Goodsell, while newer versions have been released by Dr. Garrett M. Morris, who has been adding new scientific understanding and strategies to AutoDock, making it computationally more robust, faster, and easier for scientists to use.

	DNA.
	Image: © 2003, David S. Goodsell.

While a number of drugs have been designed against HIV using the structural principles described above, the human immunodeficiency virus ( HIV) virus is a particularly difficult target because of its ability to mutate quickly: eventually some HIV mutants that can reproduce even in the presence of drug begin to dominate the population of virus. This isn't because the virus is "intelligent" but simply because HIV has a "sloppy" way of copying its RNA (the virus's equivalent of DNA) from one generation to the next, leading to a large number of viruses with slightly different characteristics. This process is a perfect example of evolutionary "natural selection", since the viruses that can survive exposure to the drug will pass their genes on to the next generation.

As a result, work in the Olson laboratory has focused on computational ways to model the evolution of drug resistance and design drugs that are robust in the face of such drug resistance. This work places greater demands on computational drug docking since trial drugs must be docked with not one but potentially millions of variants of the viral proteins. The good news, though, is that such computations can lead to the design of new drugs that are better than those that target only one viral variant. Such a process is termed "computational co-evolution" since, in the computer, the drug can "evolve" along with the viral protein.

Watercolour painting of a Human Immunodeficiency Virus particle (pink, in centre) in the blood serum (off yellow surroundings). The deep pink "spikes" sticking out are gp120 coat proteins and the yellow Y-shapes next to them are anti-HIV antibodies (there is one near the centre of this image). All the biomolecules are drawn to scale, roughly .
Image: © 2003, David S. Goodsel, All Rights Reserved.
	.

Such computations require a vast number of trial dockings, testing variations in the target protein and the trial drug molecules. The distributed computing we do across the Internet makes it possible to run such calculations by sending out individual docking calculations to thousands or hundreds of thousands of individual computers. The results of such computations should not only identify the best candidate drugs to keep the AIDS virus at bay, they will also give scientists greater insights more generally into the nature of drug resistance.

For more information, see our Glossary of Terms; you can also find out more information about HIV and AIDS at http://www.aidshivresearch.com/.

New Experiments

The Relaxed Complex method will be used in many of the new experiments performed on the FightAIDS@Home grid. The Relaxed Complex method allows us to evaluate potential drugs by docking fully-flexible versions of known and potential inhibitors against an ensemble of hundreds to thousands of conformations of the target protein (that is, we examine how good a potential drug is at binding to and blocking a massive collection of many of the different shapes that the drug target can sample as it wiggles and jiggles in a warm, watery environment). Proteins are very flexible polymers, and some potential drugs might not bind well to the average conformation that is represented in a particular crystal structure. Thus, by including the many different shapes that the target protein can display when we evaluate potential drugs, the drug design process can become more realistic and more accurate.

Molecular Dynamics simulations (or other methods for sampling the drug target's conformational space) are first used to generate that massive collection of thousands of different snapshots of the shapes that the protein target can sample as it dances around. For descriptions of the Molecular Dynamics technique (MD, for short), the drug-resistance problem, and the particular collections of shapes that will be targeted in these new FA@H experiments, check out the following link: http://legacy.sdsc.edu/Press/03/032504_HIV.html. By AutoDocking fully-flexible inhibitors to the series of snapshots of the conformations that were harvested from these MD simulations, the flexibilities of both the potential drugs and their target can be incorporated into the drug design and evaluation process. Including this flexibility can be especially important when one is trying to inhibit a highly dynamic target such as HIV protease.

The models of the potential inhibitors that we use in our virtual screens are derived from the libraries of ligands that are freely distributed by "ZINC," (which stands for Zinc Is Not Commercial). ZINC is a free database provided by the Shoichet Laboratory in the Department of Pharmaceutical Chemistry at the University of California, San Francisco (UCSF). To learn more about ZINC, see Irwin, J. and Shoichet, B. J. Chem. Inf. Model. 2005; 45(1):177-82. We thank Dr. John Irwin and Prof. Brian Shoichet for creating and maintaining such a wonderfully useful and free site.

The FightAIDS@Home distributed computing network provides unparalleled resources that will enable virtual high-throughput screens of libraries of thousands of different compounds from the NCI against ensembles of hundreds to thousands of different conformations from several of the worst multi-drug-resistant forms of HIV protease. Such mind-bogglingly complex experiments are thought to be impossible by most scientists in the drug design community. We would not be able to perform these experiments without the computational resources that you help provide. Thank you for helping us push the boundaries of what can be accomplished with current technology.

The ensembles of conformations of different multi-drug-resistant (MDR) mutants of HIV protease will be used for two different drug design projects that use the Relaxed Complex method. In one line of research, the NCI diversity set of compounds (as well as other lead compounds developed at TSRI or discussed in the literature) will be docked against the active site of the ensembles of shapes of these MDR mutants of HIV protease. When you are running the FightAIDS@Home project on your BOINC screensaver, these experiments will show the little colored spheres docking to the large cavity in the center of the ribbon diagram of HIV protease. Structural modifications to the best-performing compounds will be performed and tested both in silico and in vitro in a collaborative, iterative process, in order to aid in the development of new drugs that should be effective against these multi-drug-resistant mutants.

In the second line of new experiments, the NCI's library of fragments (i.e., a large collection of the different pieces that are sometimes combined and modified when trying to make a potential drug) will be used in Relaxed Complex experiments that focus on the "exo site" of HIV protease (i.e., on the putative allosteric inhibitor site on the peripheral surface of HIV protease). When you are running the FightAIDS@Home project on your computer, you will be able to recognize these experiments by looking for the little colored spheres that dock to the sides of the protease molecule. To learn about the initial experiments that tested the idea of trying to target this potential new drug binding site and to find some molecular art that helps to explain it, read the following Press Releases: http://www.hhmi.org/news/perryman_mccammon20060310.html and http://www.aumag.org/lifeguide/THApril06.html. A News Article that discusses this exo site and that presents the unbiased comments from a couple scientists who were not involved in these projects can be found at: http://www.biomedicalcomputationreview.org/2/4/4.pdf (read pages 3 and 4).

Thank you very much for your interest in science and for your assistance in the fight against AIDS. We couldn't do this research without your help.