Finding Cures for Tropical Diseases: Is Open Source The Answer?

Computation plays an increasing role in biology. We argue that a new approach, which we call “open source drug discovery,” would significantly reduce the cost of discovering, developing, and manufacturing cures for tropical diseases. First, it would give hundreds of scientists a practical way to donate urgently needed manpower. Second, open source discoveries would not be patented. This would permit sponsors to award development contracts to whichever company offered the lowest bid. Finally, competition from generic drug makers would keep manufacturing prices at or near the cost of production. These cost advantages could significantly accelerate drug development for the 500 million people who suffer from tropical diseases. ARTICLE: Biologists who design new research programs typically spend their time worrying about scientific barriers. Human obstacles – obtaining money and manpower – seem less important. But that assumption is only as good as the institutions that society uses to fund research. Government grants and patent incentives have supported an impressive string of breakthroughs. Nevertheless, nobody says they are perfect. Have important problems fallen though the cracks? The answer is yes. Tropical diseases affect more than 500 million people – one tenth of the world’s population – at any one time. Malaria alone kills 2.7 million people each year. Science tells us that drug development ought to be feasible. In practice, though, only about one percent of new drugs treat tropical disease. [1] The main reason is lack of funding. Patent incentives and commercial pharmaceutical houses have made Western health care the envy of the world. But the commercial model only works if companies can sell enough patented products to cover their R&D costs. This strategy fails badly in the developing world, where most consumers are penniless. Government grants are also too small to get the job done. Designing New Institutions. It’s easy – and correct – to say that Western governments could solve this problem by paying existing institutions to do more work. But that is politically unlikely. Meanwhile, it makes sense to ask whether scientists can make existing resources go further. Academic grants and patent incentives were never designed to find cures for tropical diseases. Can new institutions do better? Most proposals fall into two categories. The first class of proposals argues that charities should fix the patent system. This means using subsidies to prop up drug prices and restore incentives. [2] However, subsidies have an important weakness: Cost containment. The basic dilemma is deciding how large the subsidy should be. In principle, the most cost-effective solution is to set a subsidy that barely covers expected R&D costs. But how large is that? R&D costs are very poorly known, with published estimates ranging from $100 to $500 million per drug. [3] Set the subsidy too low and nothing will happen. Set the subsidy too high and costs skyrocket. To the best of our knowledge, no sponsor has used subsidies to prop up drug prices and restore incentives. The second class of proposals argues that charities should create non-profit venture capital firms. This approach has started to bear fruit: Today, roughly half a dozen “Virtual Pharmas” exist. Like their corporate cousins, Virtual Pharmas look for promising drug candidates and push development through clever contracts with corporate partners. Once again, the problem is guessing private sector R&D costs. As any new car buyer can tell you, you need to understand what a product costs if you expect to negotiate “the best possible price.” Virtual Pharmas are responsible for most drug candidates currently under development. The challenge now is to make them stronger. Virtual Pharmas need more upstream research, particularly in genomics. [4] They also suffer from pinched budgets, which makes cost containment essential. We argue that a new institution – “open source drug discovery” – can fill these gaps. Open Source Drug Discovery. To date, open source methods have made little headway beyond software. [5] However, computing and computational biology are converging. In the same way that programmers find bugs and write patches, biologists look for proteins (“targets”) and select chemicals (“drug candidates”) that bind to them and affect their behavior in desirable ways. In both cases, research consists of finding and fixing tiny problems hidden in an ocean of code. What would open source drug discovery look like? In analogy with current software collaborations, we propose a Web site where volunteers could search and annotate shared databases. Individual pages would host tasks like searching for new targets, finding chemicals to attack known targets, and posting data from related chemistry and biology experiments. There would also be chat rooms and bulletin boards where volunteers could announce discoveries and debate future research directions. Over time, the most dedicated and proficient volunteers would become leaders. Most importantly, all discoveries would be “placed in the public domain” – i.e., made public so that anyone could use them. Just as it does today, Virtual Pharma would choose the best candidates. The difference is that open source drugs could not be patented. This would not stop Virtual Pharma from developing promising discoveries. [6] To the contrary: We argue below that public domain status can make drug development cheaper. Incentives Without Patents. Patents are not the only way to elicit innovation. Consider the biologists who would be asked to work on the project. Economists have shown that software collaborations appeal to a variety of motives including ideology, learning new skills, gaining reputation, and impressing potential employers. [7] These incentives may sound limited. However, open source software would not exist without them. Similar incentives should motivate biologists; in fact, publication is a particularly strong motive for academic biologists. Now consider the universities and corporations who will be asked to supply people and resources. One might expect them to worry about intellectual property rights. However, a sensible manager does not assert rights unless she expects to earn a profit. This explains why sophisticated university licensing offices tolerate most open source software projects. We think that they will be similarly understanding of open source drug discovery. Life sciences companies will probably be equally tolerant. [8] Universities and companies also own important intellectual property. Are they willing to share data, research tools, and other inputs? Once again, the main point is that the commercial value of these inputs depends almost entirely on US and European markets. For this reason, universities and companies have little to lose by sharing their intellectual property with groups that fight tropical diseases. In fact, drug companies already do this. [9] The main challenge will be to show donors that an open source project can keep members from diverting donated information into unauthorized commercial research. Finally, consider the private companies whose facilities will be needed to turn open source discoveries into actual drugs. During the 1950s, the March of Dimes developed polio vaccines without any patents at all. [10] Instead, corporate partners received contract payments to help with development. The arrangement was good business. Contract profits may have been small compared to patents, but so was the risk. Fifty years later, contract research still makes sense. Generic drug companies, developing world drug manufacturers, contract research organizations, and biotech firms have all said that they would consider contracts to develop open source drug candidates. [11] Cost Containment. Open source’s most obvious cost saving is that sponsors do not have to pay for labor. As previously noted, genomics skills are in particularly short supply. Open source can help fill this gap. Open source’s cost advantage does not end when the volunteers go home. We have already explained why patent incentives do a poor job of containing costs. Open source escapes this trap by putting discoveries in the public domain. This means that governments and charities can invite companies to bid against each other for the right to perform further development under contract. Competitive bidding is a powerful method for containing costs. It is also a good way to develop drugs. Virtual Pharma has extensive experience supervising contract research. Finally, the absence of patents would keep prices low once drugs reached the market. Patents, after all, are designed to keep prices high. US drugs frequently fall to about onefourth the original price once patents expire. Getting Physical. So far we have described a shoestring operation that exists mainly on the Web. Except for computer time, budgets would be more or less the same as existing software collaborations. Computing would be expensive but manageable. Today’s biologists routinely scrounge resources from university machines or borrow time on home computers. [12] This Web-centric approach would be a good start, but not a complete solution. Computational biology works best when it can interact with wet chemistry and biology experiments. Nevertheless, a low budget computational approach is probably enough to generate new science, patentable discoveries, and ideas for followup experiments. In practice, an open source drug discovery effort is likely to include modest physical experiments. Many academic scientists control discretionary resources and, in some cases, tropical disease grants. Furthermore, good science generates its own funding. We expect experimentalists to turn the collaboration’s web pages into grant proposals. Truly balanced research would require sponsors. Charities could support open source drug discovery by making wet chemistry and biology experiments a top priority. Corporations can also help by donating funds, laboratory time, or previously unpublished results. One low cost/high value option would be to reveal pre-existing data whenever the collaboration was about to explore a “dry hole.” [13] Conclusion. So far, we have argued that open source is feasible, i.e. that no known scientific or economic barrier bars the way. But what are the risks? Experience with software collaborations highlights the main social and economic challenges. First, the project will have to find and motivate volunteers. Based on existing software collaborations, we estimate a required minimum “critical mass” of a few dozen active members. Second, modest chemistry and biology experiments can increase the chances for success. Resources of several hundred thousand dollars per year – most in the form of in-kind donations such as databases, laboratory access, and computing time – would make open source drug discovery much more powerful. By most standards, such risks are real but acceptable. The largest uncertainties are scientific. Can a volunteer effort based on computational biology and modest experiments produce leads that are promising enough to catch Virtual Pharma’s attention? We have argued that a successful program must (a) make a significant contribution toward supplying the genomic insights that Virtual Pharma needs, and (b) put useful drug candidates in the public domain. Ten years ago, these goals would have been unrealistic. Today, however, researchers frequently use computation to find promising protein targets and lead compounds. [14] Open source drug discovery looks feasible. The only way to be sure is to do the experiment. [1] O. Trouiller & P.L. Olliaro, “Drug Development Output from 1975-1996: What Proportion for Tropical Diseases,” Int. J. Infect. Dis. 3: 61-63 (1999). [2] M. Ganslandt, K. Maskus & E. Wong, “Developing and Distributing Essential Medicines to Poor Countries: The DEFEND Proposal,” The World Economy (Blackwell 2001) 24:785; M. Kremer, “A Purchasing Commitment for New Vaccines Part II: Design Issues” in A. Jaffe, J. Lerner, and S. Stern (eds.), Innovation Policy and the Economy, (MIT 2001); J. Sachs, “Helping the World’s Poorest,” The Economist (14 Aug. 1999). [3] A. Relman & M. Angell, “America’s Other Drug Problem,” The New Republic (Dec. 16 2002), pp. 27-41. [4] S. Nwaka & R. Ridley, “Virtual Drug Discovery and Development for Neglected Diseases Through Public-Private Partnerships,” Nature Reviews: Drug Discovery 2:919 (2003). [5] Current references to “open source biology” invariably mean software development (e.g., Bioperl) or else depositing un-patented data in a community repository (e.g., the SNP Consortium, the Alliance for Cell Signaling). D. Burk, “Open Source Genomics,” Boston Univ. Journal of Science and Technology Law 8:254 (2002); J. Hope, “Open Source Biotechnology?” (2003), unpublished doctoral thesis available at http://rsss.anu.edu.au/~janeth/OSBiotech.html. K. Cukier, “Community Property,” Acumen 1:54 (2003); D. Hamilton, “Open to All,” Wall Street Journal (May 19, 2003). [6] S. Nwaka, Scientific Officer, Medicines for Malaria Venture (personal communication); V. Holt, CEO, OneWorld Health (personal communication). [7] J. Lerner & J. Tirole, “Some Simple Economics of Open Source,” Journal of Industrial Economics 50:197 (2002). [8] A. Rai (unpublished survey); R. Goold, Senior Vice President, Chief Genomics Scientist Incyte Corp. [9] D. Normile, “Syngenta Agrees to Wider Release,” Science, 296:1785 (2002); D. Normile, “Monsanto Donates Its Share of Golden Rice,” Science 289:843 (2000); V. Holt, CEO of One World Health (personal communication). [10] J. Smith, Patenting the Sun (Anchor/Doubleday: 1991). [11] M. Spino, Vice President for Scientific Affairs, Apotex Inc. (personal communication); S. Sharma, Chief Scientific Officer, Nicolas Piramel India Ltd. (personal communication); F. Hijek, Director, Therapeutic Development, Duke Clinical Research Institute (personal communication); D. Francis, President, Vaxgen Corp. (personal communication). [12] Oxford University Centre for Computational Drug Discovery, “Screensaver Lifesaver,” available at http://www.chem.ox.ac.uk/curecancer.html; Stanford University Pande Group, “Genome@Home Distributed Computing,” available at http://www.stanford.edu/group/pandegroup/genome. [13] R. Altman, Stanford University (personal communication). [14] See. e.g., K. Ginalski & L. Rychelewski, “mRNA cap-1 methyltransferase in the SARS genome,” Cell 13: 701 (2003); C. Ring, E. Sun, S. J. McKerrow, G. Lee, P. Rosenthal, I. Kutnz & F. Cohen, “Structure-based inhibitor design by using protein models for the development of antiparasitic agents,” Proc. Natl. Acad. Sci., 90:3583 (2002); B. Shoichet, S. McGovern, B. Wei, J. Irwin, “Lead discovery using molecular docking,” Curr. Opin. Chem. Biol. 4:439 (2002).