Albeit a bit anticlimactically, a study published in the journal Genomics assesses the patentability of one of the claims invalidated on Monday by Judge Robert W. Sweet of the Southern District of New York in Association of Molecular Pathology v. U.S. Patent and Trademark Office. The study is interesting in that it illustrates the pitfalls in patents on oligonucleotides related to isolated genes, filed at a time prior to the elucidation of the human genome by the Human Genome Project. Most relevant to the question of whether such claims are patentable are the many revelations from the HGP that overturned settled dogma based on a now-recognized unrealistic view of the randomness of genomic DNA sequences, as illustrated by the results reported in this study.
The research was performed by Thomas B. Kepler, from the Department of Biostatistics and Bioinformatics at Duke University, and the Center for Computational Immunology; Colin Crossman, a lawyer and owner of Memento Mori, LLC; and Robert Cook-Deegan from the Institute for Genome Science and Policy and the Sanford School of Public Policy, Duke University (Kepler et al., "Metastasizing patent claims on BRCA1"). According to these authors, one of the claims of U.S. Patent No. 5,747,282 seemed "exceptionally broad" -- this is claim 5 that depends on claim 1 (which is directed to an isolated DNA molecule encoding a BRCA1 polypeptide):
1. An isolated DNA coding for a BRCA1 polypeptide, said polypeptide having the amino acid sequence set forth in SEQ ID NO:2.
5. An isolated DNA having at least 15 nucleotides of the DNA of claim 1.
The authors estimate that "the human genome contains over one million oligonucleotides covered by this claim," and follow up this statement with the following calculations:
Accounting for bias in the usage of amino acids as reported, for example, in , the usage-weighted geometric mean codon degeneracy per amino acid is 3.107. Therefore, the mean number of 15-mers encoding a polypeptide of length 5 chosen at random from a vertebrate proteome is 3.1075, about 290. There are 5,575 15-mers in BRCA1, so, if we consider all of the nucleotide sequences that encode the BCRA1 protein, there are about 1.6 x 106 15-mers embodied by the claim. There are 415=1.07 x 109 different 15-mers altogether, so the probability that a 15-mer chosen at random will be covered by the claim is p=1.6 x 106 / 1.07 x 109=0.0015 (roughly, 1 in 600 possible 15-mers). A typical human gene (before RNA editing) contains 10,000 bases, so, if human genes were random strings of nucleotides, one would expect a human gene to contain an average of 15 15-mers claimed under the patent.
The predicted results of this analysis were found in the study. The authors analyzed the nucleotide sequence of human chromosome 1 looking for only a subset of 15-mers encompassed by claim 5. ("Computing time" was minimized by excluding two of the six degenerate codons for serine, leucine and arginine amino acids, a reduction said to only "slightly underestimate the degree of redundancy and breadth of claim 5.") The authors reported finding 340,000 "matches" of the claimed sequences in the 250,000,000 nucleotides comprising chromosome 1.
The authors also examined 713 entries in GenBank representing complete coding sequences for human mRNAs deposited in 1994, reporting that 80% (568/713 contained at least one of the claimed 15-mers.
There is nothing incorrect about this analysis; however, it benefits from current knowledge and (in its implied conclusion of intentional overreaching) neglects to consider the state of the genomics art on August 12,1994, the earliest priority date of the '282 patent. The calculation was much more simple (or perhaps naïve) then: any particular 15-mer was expected to occur once in every 1.07 x 109 nucleotides, and thus in a completely random genome the size of the human genome (haploid size of about 3 x 109 nucleotides), to occur about 3 times; a 16-mer would be expected to occur four times less frequently, etc. Thus, the results set forth in this paper were almost completely unexpected when approached from the outlook of the person of skill in the art in 1994.
As it turns out, the human genome (and most other genomes) are much more inhomogeneous than expected, and the effects of evolution and the relatedness of all organisms (as well as the conservation of motifs and functional domains between species) were equally unexpected. Indeed, even the number of genes encoded in the human genome turns out to be much smaller (2- to 3-fold fewer genes) than was expected. As the authors state: "human genes are not random strings [of nucleotides.]" Evident now, not so evident 16 years ago.
No one wants an invalid patent. The in terrorem effect of such a patent is greatly exaggerated, particularly when its invalidity is so easily demonstrated. Here, these claims, as well as similar claims in other patents, turn out to be sufficiently overbroad as to be easily invalidated. It is unlikely that the University of Utah, the National Institutes of Health, or Myriad Genetics (all owners of this patent) wanted this result.
Nevertheless, the remainder of the paper discussed the policy implications of this invalid claim. The authors correctly note that their results indicate that claim 5 was anticipated by the 15-mers known in the prior art as exemplified by the GenBank results. Thus, "[i]f challenged by re-examination or in litigation, claim 5 may be deemed invalid due to readily identifiable prior art covered by the claim." The paper cites U.S. Patent Examiner James Martinell, examining a 1991 expressed sequence tag patent application, for recognizing that certain 15-mers could be found in "many genes" and that it was impossible using then-current computer technology to search all known sequences for the 700,000 15-mers claimed in that patent. Perhaps for that reason, there is no evidence in the prosecution history of the '282 patent that such a search was ever performed.
Despite this strong evidence that claim 5 of the '282 patent is invalid (even before the District Court's decision yesterday), the paper speculates on the scope and reach of the claim (including oligonucleotide primers for performing the polymerase chain reaction). Regarding the effects of this claim (and Myriad's patents) on basic research, the authors recognize what many others have noted: no appreciable effect (which belies the introductory sentence of this section that "[t]he effect of this claim on research is very difficult to assess"):
A PubMed search for the term "BRCA1" returned 7,107 articles. This suggests a large body of research on the gene has been published in the technical literature. Myriad has not enforced its patents against most research, with the exception of laboratories engaged in clinical research that entailed giving test results to individuals beyond their home institutions [11–13]. Any such research that entailed analysis of DNA molecules containing BRCA1 sequences in the United States very likely infringed this claim, however, so enforcement of this claim would have substantial impact on research. (Claims to BRCA1 sequences are somewhat narrower in other English-language jurisdictions such as Canada, Australia and New Zealand, and a fortiori in Europe, where the claims that emerged from opposition proceedings were dramatically narrowed.) There is a very narrow "research exemption" from infringement liability in the United States under common law, and a broader exemption for research that results in data contributed to the government for a regulated medical product or service . Since laboratory-developed tests are not currently subject to Food and Drug Administration approval, however, this exemption may not apply.
The simplest conclusion about the effect of claim 5 and Myriad's other BRCA1 patents on research and clinical testing is that Myriad has only rarely enforced its patents in research, has vigorously enforced its patents against commercial genetic testing, and has selectively enforced its patents in clinical research. It is also apparent that research on BRCA1 for the past 12 years has entailed massive pervasive infringement of this claim, even if the claim's scope were restricted to BRCA1 research. Any such research in the United States was thus undertaken under risk of infringement liability and its associated uncertainty. While Myriad has stated publicly that it has not enforced its patents against basic research [11,12,15], it has not stated it will not do so in the future, and therefore BRCA research in the United States continues only with Myriad's indulgence.
The authors are correct that there is no exemption from infringement liability for basic research, but they admit that Myriad has not enforced (or attempted to enforce) the '282 patent against any basic researchers. The absence of the exemption may be the problem, not claim 5 of the '282 patent (in view of the results of this research). Myriad's track record with regard to basic research is consistent with its public statements that it would not enforce its patents against basic research; there is very little else it can do. And after yesterday's decision, basic or clinical researchers or commercial entities that practice the oligonucleotides recited in claim 5 no longer risk infringement liability.