By Kevin E. Noonan —
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 [7],
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 [14]. 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.
Patent Docs 
Leave a reply to Devesh Srivastava Cancel reply