The American Journal of Bioethics 2.3 (2002) 15-16
[Access article in PDF]
Open Peer Commentaries
No Patents for Semantic Information
Washington and Lee University
It is too late to prevent patents on individual genes. But I do not think the decision to allow patents on genes was an inevitable extension of prior law—just read the dissenting opinion in Diamond v. Chakrabarty (447 U.S. 303 ), and imagine that one more justice thought similarly. More important, I am not convinced that either citizens or scientists got exactly what they wanted. Though gene patents advance medicine and biology by providing incentives to share, divulge, and develop research, they also distort medicine and biology by altering the incentive structure for research. Rebecca Eisenberg's (2002) article discusses what could be the next step in the aggressive expansion of intellectual property law into biology: patents for genome information.
As Eisenberg points out, patents on individual genes that produce useful products are consistent with patents for chemical compounds and other compositions of matter that would be products of nature had they not been purified and isolated. But what about patents for genomic sequence data? The result of high-throughput technology is scientific information, not a tangible invention. This information has value—considerable money and energy have been expended to produce genomic sequence data for private and public consumption. For example, Genome Therapies Corporation sold the gene sequence for Heliobacter pylori to a Swedish pharmaceutical company for $22 million (Marshall 1997). But can this information be patented?
In an ideal world scientists would map genomic sequence information in direct proportion to its usefulness not only to them but to the scientific community at large. Why would scientists do this? Scientists face numerous incentives for producing and sharing information. These incentives include the desire for credit and priority, direct or indirect altruism, faith in a system based on the free sharing of information, and fear. Scientists fear retaliation (other scientists won't share with me if I don't with them) and competition (if I'm not the first to release this information, someone else will be). Scientists also encounter incentives to not share—restrictive agreements, trade secrecy, retaliating against others who have not shared, protection of poor data from exposure, and the temptation to freeload from a sharing-based system while not contributing to it.
Patent policy can be used to change the incentive structure if necessary. As both Eisenberg and Robertson (2002) note, this does not seem to be necessary in the case of genomic data. For whatever reason, a lot of DNA sequence information has been made freely available. Many bacterial genomes have been submitted to GenBank, a public database run by the National Library of Medicine (Pennisi 1997). Scientists have direct computer access to [End Page 15] both the public and private versions of the human genome, as well as the genome information for several model organisms. 1 Some critics have even argued that the human genome project and related research have led to an overinvestment in sequencing genomes relative to other lines of investigation (Rosenberg 1996). This environment is not an encouraging one for a radical extension of patent rights for information that has already been made available.
Eisenberg also presents a second argument suggesting that rejecting patents on genomic information could endanger the legitimacy of patents on individual genes. If the U.S. Patent and Trade Office grants patents on information in the form of molecules, how can it prohibit patents on information in the form of computer-readable databases? If we reject patents on the latter, wouldn't we have to reject patents on the former? Eisenberg suggests that legal precedent can support treating the two forms of information differently, but if we clarify the different concepts of information in play, there is an even more principled way to separate the two cases.
Broadly speaking, there are two distinct kinds of information, causal and semantic. The concept of causal information dates back to Shannon and Weaver (1949). A sender transmits information to a receiver through a channel when there is...