Guardant Health, Inc.

Patent Trials and Appeals Board

Aug 18, 2020

IPR2019-00652 (P.T.A.B. Aug. 18, 2020)

Trials@uspto.gov Paper 47
571-272-7822 Date: August 18, 2020

UNITED STATES PATENT AND TRADEMARK OFFICE

BEFORE THE PATENT TRIAL AND APPEAL BOARD

FOUNDATION MEDICINE, INC.,
Petitioner,
v.
GUARDANT HEALTH, INC.,
Patent Owner.

IPR2019-00652
Patent 9,834,822 B2

Before SUSAN L. C. MITCHELL, TINA E. HULSE, and KRISTI L. R.
SAWERT, Administrative Patent Judges.
SAWERT, Administrative Patent Judge.

JUDGMENT
Final Written Decision
Determining Some Claims Unpatentable
Dismissing in Part and Denying in Part Petitioner’s Motion to Exclude
Dismissing in Part and Denying in Part Patent Owner’s Motion to Exclude
35 U.S.C. § 318(a)

IPR2019-00652
Patent 9,834,822 B2
2
I. INTRODUCTION
This is a Final Written Decision in an inter partes review challenging
the patentability of claims 1–13 and 17–20 (“the challenged claims”) of
U.S. Patent No. 9,834,822 B2 (Ex. 1001, “the ’822 patent”). We have
jurisdiction under 35 U.S.C. § 6 and enter this Decision pursuant to
35 U.S.C. § 318(a) and 37 C.F.R. § 42.73. For the reasons set forth below,
we determine that Petitioner has shown, by a preponderance of the evidence,
that claims 1–11, 13, and 17–20 are unpatentable. We determine that
Petitioner has not shown, by a preponderance of the evidence, that claim 12
is unpatentable. See 35 U.S.C. § 316(e) (2012).
A. Procedural History
Foundation Medicine, Inc. (“Petitioner”) filed a Petition for an inter
partes review under 35 U.S.C. § 311. Paper 2 (“Pet.”). Petitioner supported
its Petition with the Declaration of Stacey Gabriel, Ph.D. Ex. 1002.
Guardant Health, Inc. (“Patent Owner”) filed a Preliminary Response.
Paper 6. On our authorization (Paper 9), Petitioner filed a Reply to Patent
Owner’s Preliminary Response (Paper 11).
On August 19, 2019, pursuant to 35 U.S.C. § 314(a), we instituted
trial to determine whether any challenged claim of the ’822 patent is
unpatentable based on the grounds raised in the Petition:
IPR2019-00652
Patent 9,834,822 B2
3
Claims Challenged 35 U.S.C. § Reference(s)/Basis
1–13, 17–20 103(a)1 Schmitt,2 Schmitt 2012,3 and Fan4 or
Forshew5
Paper 12, 7, 36 (“Institution Decision” or “Inst. Dec.”).
Patent Owner filed a Response. Paper 26 (“PO Resp.”). Patent
Owner supported its Response with the Declaration of Jay Shendure, M.D.,
Ph.D., Ex. 2023, and the Declaration of John Quackenbush, Ph.D., Ex. 2025.
Petitioner filed a Reply to Patent Owner’s Response. Paper 32 (“Pet.
Reply”). Petitioner supported its Reply with a Reply Declaration of
Dr. Gabriel. Ex. 1104. Patent Owner filed a Sur-Reply. Paper 34 (“PO Sur-
Reply”). Patent Owner supported its Sur-Reply with a Supplemental
Declaration of Dr. Quackenbush. Ex. 2042.

1 The Leahy-Smith America Invents Act (“AIA”), Pub. L. No. 112-29,
125 Stat. 284, 287–88 (2011), amended 35 U.S.C. § 103, effective March 16,
2013. Because the challenged claims have an effective filing date before
this date, the pre-AIA version of § 103 applies.
2 Michael Schmitt et al., U.S. Patent No. 9,752,188 B2, issued Sept. 5,
2017 (Ex. 1011, “Schmitt”).
3 Michael W. Schmitt et al., Detection of Ultra-rare Mutations by
Next-generation Sequencing, 109(36) PROC. NATL. ACAD. SCI. 14508–513
(2012) (Ex. 1047, “Schmitt 2012”).
4 Christina Fan et al., Noninvasive diagnosis of fetal aneuploidy by
shotgun sequencing DNA from maternal blood, 105(42) PROC. NATL. ACAD.
SCI. 16266–271 (2008) (Ex. 1048, “Fan”)
5 Tim Forshew et al., Noninvasive Identification and Monitoring of
Cancer Mutations by Targeted Deep Sequencing of Plasma DNA, 4(136)
SCI. TRANSL. MED. 1–34 (2012) (Ex. 1004, “Forshew”).

IPR2019-00652
Patent 9,834,822 B2
4
Petitioner and Patent Owner each filed respective Motions to Exclude
Evidence. See Paper 38 (“Pet. Mot.”); Paper 39 (“PO Mot.”). Petitioner
filed an Opposition to Patent Owner’s Motion, Paper 40 (“Pet. Opp.”), to
which Patent Owner filed a Reply, Paper 42 (“PO Reply”). Patent Owner
filed an Opposition to Petitioner’s Motion, Paper 41 (“PO Opp.”), to which
Petitioner filed a Reply, Paper 43 (“Pet. Reply Opp.”).
An oral hearing was held on May 13, 2020. A transcript of the
hearing is included in the record. Paper 46 (“Tr.”).
B. Real Parties in Interest
Petitioner identifies Foundation Medicine, Inc., Roche Holdings, Inc.,
Roche Finance Ltd., and Roche Holding Ltd. as the real parties-in-interest.
Pet. 73. Patent Owner identifies Guardant Health, Inc., as the real party-in-
interest. Paper 4, 2.
C. Related Matters
Patent Owner has asserted the ’822 patent against Petitioner in
Guardant Health, Inc. v. Foundation Medicine, Inc., Case No. 17-cv-1616
(D. Del.) (“the co-pending litigation”). Pet. 74; Paper 4, 2. Patent Owner
has also asserted the ’822 patent against Personal Genome Diagnostics, Inc.
(“PGDx”) in Guardant Health, Inc. v. Personal Genome Diagnostics, Inc.,
Case No. 17-cv-1623 (D. Del.). Pet. 74; Paper 4, 2.
Petitioner filed a second petition seeking inter partes review of the
’822 patent, designated IPR2019-00653. Paper 4, 2. A Decision denying
institution in that case was issued on August 19, 2019 (Paper 12), and a
Decision denying Petitioner’s request for rehearing issued on January 22,
2020 (Paper 14).
Petitioner also filed several petitions seeking inter partes review of
patents related to the ’822 patent, including: IPR2017-01170, IPR2017-
IPR2019-00652
Patent 9,834,822 B2
5
01447, and IPR2017-01448 (challenging U.S. Patent No. 9,340,830);
IPR2019-00130 (challenging U.S. Patent No. 9,598,731); IPR2019-00634
(challenging U.S. Patent No. 9,840,743); and IPR2019-00636 and IPR2019-
00637 (challenging U.S. Patent No. 9,902,992). Of these cases, only
IPR2019-00634 is pending.
PGDx also filed petitions seeking post-grant review of the ’822
patent, designated PGR2018-00058, and of U.S. Patent No. 9,840,743,
designated PGR2018-00057. Id. at 2. Both petitions were dismissed before
a decision on institution.
D. Summary of the ’822 Patent
The ’822 patent relates to methods for detecting rare mutations and
copy number variations in cell free polynucleotides. Ex. 1001, code (57).
The ’822 patent states that cell-free DNA (“cfDNA”), found in different
types of bodily fluids, may be used to detect and monitor disease. Id. at
1:29–45. For instance, cfDNA may contain genetic aberrations—like a
change in copy number variations and/or single or multiple sequence
variations associated with a particular disease—that can be used to detect or
monitor such disease. Id. at 1:29–41, 30:8–14. The ’822 patent states that
“there is a need in the art for improved methods and systems for using cell
free DNA to detect and monitor disease.” Id. at 1:41–45.
The ’822 patent states that the disclosed methods generally “comprise
sample preparation, or the extraction and isolation of cell free polynucleotide
sequence[s] from a bodily fluid; subsequent sequencing of cell free
polynucleotides by techniques known in the art; and application of
bioinformatics tools to detect rare mutations and copy number variations as
compared to a reference.” Id. at 30:4–14. The ’822 patent states that
“[s]ample preparation typically involves converting polynucleotides in a
IPR2019-00652
Patent 9,834,822 B2
6
sample into a form compatible with the sequencing platform used.” Id. at
32:58–61. “This conversion [c]an involve tagging polynucleotides” with
“polynucleotide sequence[s].” Id. at 32:61–63. The ’822 patent refers to
these polynucleotide sequences as “identifiers.” Id. at 38:3–6. The
identifier may be a molecular barcode. Id. at 38:6–7.
The ’822 patent explains that “the efficient conversion of individual
polynucleotides in a sample of initial genetic material into sequence-ready
tagged parent polynucleotides” is an important tool “for detecting with high
sensitivity genetic variation in a sample of initial genetic material.” Id. at
32:33–39. According to the ’822 patent, efficient conversion “increase[s]
the probability that individual polynucleotides in a sample of initial genetic
material will be represented in a sequence-ready sample” and “can produce
sequence information about more polynucleotides in the initial sample.” Id.
at 32:39–42.
The ’822 patent states that the parent polynucleotides may be tagged
with either unique or non-unique identifiers. Id. at 37:44–49; see also id. at
3:10–15 (stating that, in some embodiments, the barcodes are unique, but in
other embodiments, the barcodes are not unique); see also id. at 6:26–28
(stating that, in some embodiments, “each tagged parent polynucleotide in
the set is uniquely tagged,” whereas in other embodiments, “the tags are
non-unique”). In the case of non-uniquely tagged parent polynucleotides,
the ’822 patent explains that “the use of non[-]unique barcodes, in
combination with sequence data at the beginning (start) and end (stop)
portions of individual sequencing reads and sequencing read length may
allow for the assignment of a unique identity to individual sequences.” Id. at
37:43–48.
IPR2019-00652
Patent 9,834,822 B2
7
E. Illustrative Claim
Claim 1 is the only independent claim. Claims 2–13 and 17–20
depend directly from claim 1. See Ex. 1001, 62:51–64:22. Claim 1 is
illustrative and reproduced below:
1. A method, comprising:

a) providing a population of cell-free DNA (“cfDNA”)
molecules obtained from a bodily sample from a subject;

b) converting the population of cfDNA molecules into a
population of non-uniquely tagged parent polynucleotides,
wherein each of the non-uniquely tagged parent
polynucleotides comprises (i) a sequence from a cfDNA
molecule of the population of cfDNA molecules, and (ii) an
identifier sequence comprising one or more polynucleotide
barcodes;

c) amplifying the population of non-uniquely tagged parent
polynucleotides to produce a corresponding population of
amplified progeny polynucleotides;

d) sequencing the population of amplified progeny
polynucleotides to produce a set of sequence reads;

e) mapping sequence reads of the set of sequence reads to one
or more reference sequences from a human genome;

f) grouping the sequence reads into families, each of the
families comprising sequence reads comprising the same
identifier sequence and having the same start and stop
positions, whereby each of the families comprises sequence
reads amplified from the same tagged parent
polynucleotide;

g) at each genetic locus of a plurality of genetic loci in the one
or more reference sequences, collapsing sequence reads in
each family to yield a base call for each family at the genetic
locus;
IPR2019-00652
Patent 9,834,822 B2
8

h) determining a frequency of one or more bases called at the
locus from among the families.

Ex. 1001, 62:18–48.
II. PETITIONER’S MOTION TO EXCLUDE
Petitioner moves to exclude Exhibits 2002, 2032, 2036, 2037, 2038,
2039, 2040, and 2041 in their entirety and Exhibits 2023 and 2025 in whole
or in part. Pet. Mot. 1.
We dismiss as moot Petitioner’ Motion to Exclude as it relates to
Exhibits 2002, 2032, 2036, 2037, 2038, 2039, 2040, and 2041, for the
following reasons. First, we do not rely on or cite to Exhibits 2032, 2036,
2037, or 2038 in this Decision. Second, Patent Owner cites to Exhibits
2002, 2039, 2040, and 2041 as support for its arguments about reasonable
expectation of success. To the extent we refer to these exhibits herein, we
determine that the record as a whole supports Petitioner’s position as to
reasonable expectation of success. Infra § IV.E.2.b. Third, Exhibit 2023 is
the Declaration of Dr. Shendure that is limited in scope to the applicability
of Schmitt to cfDNA. Ex. 2023 ¶¶ 16–37. Again, to the extent we refer to
Exhibit 2023, we determine that the record as a whole supports Petitioner’s
position as to motivation to combine the Schmitt references with Fan or
Forshew and as to reasonable expectation of success. Infra § IV.E.2.a–b.
Thus, as to Exhibits 2002, 2023, 2032, 2036, 2037, 2038, 2039, 2040, and
2041, on which we do not rely to support our decision, Petitioner’s Motion
to Exclude is dismissed as moot.
Exhibit 2025 is the Declaration of Dr. Quackenbush. Petitioner
contends that Dr. Quackenbush’s Declaration should be excluded under
Federal Rules of Evidence 702, 703, 705, and 401–403. Pet. Mot. 5–7.
IPR2019-00652
Patent 9,834,822 B2
9
Specifically, Petitioner contends that we should exclude at least certain
paragraphs of Dr. Quackenbush’s Declaration because “they lack a disclosed
basis of sufficient facts or data,” “are not based on sufficient facts or data,
the product of reliable principles and methods, and/or a reliable application
of the principles and methods to the facts,” and “are misleading, confusing,
and/or needlessly cumulative.” Id. at 5.
We have considered Petitioner’s arguments but are not persuaded that
Exhibit 2025 should be excluded in whole or in part. Patent Owner has
shown, and Petitioner does not dispute, that Dr. Quackenbush is qualified to
opine as to the perspective of an ordinarily skilled artisan at the time of the
invention. PO. Opp. 5–93; see also infra § IV.B. As such,
Dr. Quackenbush’s testimony is highly relevant about how an ordinarily
skilled artisan would have interpreted the prior-art references, as well as to
the general knowledge in the field at the time the invention was made. Any
deficiencies in Dr. Quackenbush’s Declaration go to the weight that we
should afford his testimony and do not support a motion to exclude. See,
e.g., Yorkey v. Diab, 601 F.3d 1279, 1284 (Fed. Cir. 2010) (holding that the
Board has discretion to give more weight to one item of evidence over
another “unless no reasonable trier of fact could have done so”); In re Am.
Acad. of Sci. Tech Ctr., 367 F.3d 1359, 1368 (Fed. Cir. 2004) (“[T]he Board
is entitled to weigh the declarations and conclude that the lack of factual
corroboration warrants discounting the opinions expressed in the
declarations.”). Thus, as to Exhibit 2025, Petitioner’s Motion is Exclude is
denied.
III. PATENT OWNER’S MOTION TO EXCLUDE
Patent Owner moves to exclude Exhibits 1002, 1007, 1013, 1016–
1020, 1022, 1036, 1037, 1055, 1080, 1082, 1084–1093, 1100, 1101, 1104,
IPR2019-00652
Patent 9,834,822 B2
10
1110, and 1111. PO Mot. 1. We do not rely on Exhibits 1007, 1013, 1016–
1020, 1022, 1036, 1037, 1055, 1080, 1082, 1084–1093, 1100, 1110, or 1111
in this Decision. Thus, we dismiss as moot Patent Owner’s Motion to
Exclude as it relates to Exhibits 1007, 1013, 1016–1020, 1022, 1036, 1037,
1055, 1080, 1082, 1084–1093, 1100, 1110, and 1111.
Exhibits 1002 and 1104 are the first and second Declarations of
Dr. Gabriel, respectively. PO Mot. 1–8. Patent Owner argues that Exhibit
1002 should be excluded in its entirety because the “entire declaration is
premised on an impossible standard to be met by a person of ordinary skill
in the art” and “Dr. Gabriel uses this impossible perspective to support a
hindsight-based obviousness analysis.” Id. at 1–3. We agree with
Petitioner, however, that Patent Owner’s arguments constitute a
disagreement about a question of fact (i.e., the proper definition of an
ordinarily skilled artisan), and “[a] motion to exclude is not the proper
mechanism to direct [its] attention to differences in the evidence.” Pet. Opp.
2 (quoting Google Inc. v. Visual Real Estate, Inc., IPR2014-01339, Paper 39
at 37 (PTAB Jan. 25, 2016)). Moreover, we note that Dr. Gabriel applies in
her Declarations the same definition of an ordinarily skilled artisan as that
we adopted in our Institution Decision and reaffirm here. Inst. Dec. 7–8;
infra § IV.B. Thus, we are not persuaded that we should exclude
Dr. Gabriel’s Declarations Exhibits 1002 and 1104 for this reason.
Patent Owner also argues that Dr. Gabriel’s second Declaration
(Exhibit 1104) should be excluded because it attempts to “fill[] a gap in
[Petitioner’s] prima facie case,” relies on “new theories regarding the
teachings of Schmitt,” and “presents an entirely different theory of
obviousness with respect to the ‘non-uniquely tagged’ limitation.” PO
Mot. 3–8. We are not persuaded, however, that Exhibits 1002 and 1104
IPR2019-00652
Patent 9,834,822 B2
11
should be excluded. “A motion to exclude is not a mechanism to argue that
a reply contains new arguments or relies on evidence necessary to make out
a prima facie case.” Vibrant Media, Inc. v. General Elec. Co., IPR2013-
00170, Paper 56 at 31 (PTAB June 26, 2014). Thus, Patent Owner’s
arguments in this regard are improper.
In any event, Petitioner has shown, and Patent Owner does not
dispute, that Dr. Gabriel is qualified to opine as to the perspective of an
ordinarily skilled artisan at the time of the invention. Pet. Opp. 2–3; see also
infra § IV.B. As with Dr. Quackenbush’s testimony above, Dr. Gabriel’s
testimony is highly relevant about how an ordinarily skilled artisan would
have interpreted the prior-art references, as well as to the general knowledge
in the field at the time the invention was made. And any deficiencies in
Dr. Gabriel’s Declaration go to the weight that we should afford her
testimony and do not support a motion to exclude.
For these reasons, as to Exhibits 1002 and 1104, Patent Owner’s
Motion to Exclude is denied.
IV. ANALYSIS
We have reviewed the parties’ respective briefs as well as the relevant
evidence discussed in those papers. For the reasons discussed in detail
below, we determine that Petitioner has shown by a preponderance of the
evidence that claims 1–11, 13, and 17–20 of the ’822 patent are unpatentable
under 35 U.S.C. § 103 as having been obvious, but not that claim 12 would
have been unpatentable as obvious.
A. Principles of Law
To prevail in its challenges to the patentability of all claims of the
’822 patent, Petitioner must demonstrate by a preponderance of the evidence
that the claims are unpatentable. 35 U.S.C. § 316(e) (2012); 37 C.F.R.
IPR2019-00652
Patent 9,834,822 B2
12
§ 42.1(d) (2018). “In an [inter partes review], the petitioner has the burden
from the onset to show with particularity why the patent it challenges is
unpatentable.” Harmonic Inc. v. Avid. Tech., Inc., 815 F.3d 1356, 1363
(Fed. Cir. 2016); see also 35 U.S.C. § 312(a)(3) (requiring inter partes
review petitions to identify “with particularity . . . the evidence that supports
the grounds for the challenge to each claim”). That burden of persuasion
never shifts to Patent Owner. Dynamic Drinkware, LLC v. Nat’l Graphics,
Inc., 800 F.3d 1375, 1378 (Fed. Cir. 2015); see also In re Magnum Oil Tools
Int’l, Ltd., 829 F.3d 1364, 1375–78 (Fed. Cir. 2016) (discussing the burden
of proof in inter partes review).
A claim is unpatentable for obviousness if, to one of ordinary skill in
the pertinent art, “the differences between the subject matter sought to be
patented and the prior art are such that the subject matter as a whole would
have been obvious at the time the invention was made.” 35 U.S.C. § 103(a)
(2006); see also KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 406 (2007).
The question of obviousness is resolved on the basis of underlying factual
determinations including the scope and content of the prior art, any
differences between the claimed subject matter and the prior art, the level of
ordinary skill in the art, and objective evidence of nonobviousness. Graham
v. John Deere Co., 383 U.S. 1, 17–18 (1966). A petitioner cannot satisfy its
burden of proving obviousness by employing “mere conclusory statements.”
Magnum Oil, 829 F.3d at 1380. Moreover, a decision on the ground of
obviousness must include “articulated reasoning with some rational
underpinning to support the legal conclusion of obviousness.” KSR, 550
U.S. at 418 (citing In re Kahn, 441 F.3d 977, 988 (Fed. Cir. 2006)).
We analyze Petitioner’s asserted grounds of unpatentability in
accordance with the above-stated principles.
IPR2019-00652
Patent 9,834,822 B2
13
B. Level of Ordinary Skill in the Art
We consider the asserted grounds of unpatentability in view of the
understanding of a person of ordinary skill in the art and, thus, begin with
the level of ordinary skill in the art. The level of ordinary skill in the art is
“a prism or lens through which . . . the Board views the prior art and the
claimed invention” to prevent hindsight bias. Okajima v. Bourdeau, 261
F.3d 1350, 1355 (Fed. Cir. 2001).
Relying on the declaration testimony of its declarant, Dr. Gabriel,
Petitioner contends that a person of ordinary skill in the art for the ’822
patent “would have had a Ph.D. in genetics, molecular biology,
bioinformatics or a related field, and at least five years of research in an
academic or industry setting, including at least two to three years of research
experience in the field of cancer genomics.” Pet. 20 (citing Ex. 1002 ¶ 72).
In response, Patent Owner does not appear to dispute the type of experience
Petitioner proposes for the level of ordinary skill in the art. See generally
PO Resp. We observe, however, that Patent Owner’s declarants,
Dr. Shendure and Dr. Quackenbush, contend that Petitioner’s definition of
an ordinarily skilled artisan relates more to an artisan with “extraordinary,”
rather than “ordinary,” skill. See Ex. 2023 ¶ 15, Ex. 2025 ¶ 23.
In our Institution Decision, we preliminarily adopted Petitioner’s
proposed level of ordinary skill. Inst. Dec. 7–8. We also determined that
the prior art itself was sufficient to demonstrate the level of ordinary skill in
the art at the time of the invention. Inst. Dec. 8. For this Decision, we
maintain that the prior art demonstrates the appropriate level of ordinary
skill in the art. See Okajima, 261 F.3d at 1355 (the prior art, itself, can
reflect appropriate level of ordinary skill in art).
IPR2019-00652
Patent 9,834,822 B2
14
Nevertheless, for further clarity, we set forth the definition of an
ordinarily skilled artisan as follows. As to level of education, because of the
nature of the subject matter of the ’822 patent, we agree with Petitioner that
an ordinarily skilled artisan would have had a doctorate degree (Ph.D.) in
genetics, molecular biology, bioinformatics, or a related field. Ex. 2023
¶¶ 28–29; Ex. 1002 ¶ 25. We also agree with Petitioner that an ordinarily
skilled artisan would have had five years’ research experience in industry or
academia, including at least two to three years of research experience in the
field of cancer genomics.
We acknowledge Petitioner’s contention that an ordinarily skilled
artisan “would have had knowledge of DNA sequencing, including NGS
[next-generation sequencing] and related sequencing methods, and related
sample preparation techniques, bioinformatics methods for grouping and
comparing sequence reads and mapping sequence reads onto genomes, and
methods for identifying genetic variants in a sample.” Pet. 20. But we
conclude that these statements more aptly apply to the scope and content of
the prior art under Graham and, thus, are best addressed in relation to
Petitioner’s asserted grounds of unpatentability based on obviousness.
Finally, we have considered the qualifications of Dr. Gabriel,
Dr. Shendure, and Dr. Quackenbush and find that each is qualified to opine
as to the perspective of an ordinarily skilled artisan at the time of the
invention. See Ex. 1003 (Dr. Gabriel’s curriculum vitae); Ex. 2024
(Dr. Shendure’s curriculum vitae); Ex. 2026 (Dr. Quackenbush’s curriculum
vitae).
C. Claim Construction
The instant Petition was filed on February 2, 2019. Thus, the new
rules amending the claim construction standard apply here because the
IPR2019-00652
Patent 9,834,822 B2
15
Petition was filed after the November 13, 2018, effective date of the
amendment. See Changes to the Claim Construction Standard for
Interpreting Claims in Trial Proceedings Before the Patent Trial and Appeal
Board, 83 Fed. Reg. 51,340, 51,358 (Oct. 11, 2018) (amending 37 C.F.R.
§ 42.100(b) effective November 13, 2018) (now codified at 37 C.F.R.
§ 42.100(b) (2019)). Accordingly, for this inter partes review, the Board
applies the same claim construction standard as that applied in federal
courts.
Under this standard, we construe the claim “in accordance with the
ordinary and customary meaning of such claim as understood by one of
ordinary skill in the art and the prosecution history pertaining to the patent.”
Id.; see also Phillips v. AWH Corp., 415 F.3d 1303, 1312–13 (Fed. Cir.
2005) (en banc) (stating that claim terms “are generally given their ordinary
and customary meaning” as understood by a person of ordinary skill in the
art in question at the time of the invention). Only terms that are in
controversy need to be construed, and then only to the extent necessary to
resolve the controversy. Nidec Motor Corp. v. Zhongshan Broad Ocean
Motor Co., 868 F.3d 1013, 1017 (Fed. Cir. 2017).
1. Overview
At the close of trial, the only claim-construction dispute remaining in
this proceeding concerns the meaning of “non-uniquely tagged” in the
context of a population of parent polynucleotides, as recited in claim 1. See
Ex. 1001, 62:22–24 (reciting “converting the population of cfDNA
molecules into a population of non-uniquely tagged parent
polynucleotides”). Petitioner contends that “non-uniquely tagged” “means
that the number of different identifiers attached to the polynucleotides is at
least 2 and fewer than the number of polynucleotides.” Pet. 21. Patent
IPR2019-00652
Patent 9,834,822 B2
16
Owner argues that “non-uniquely tagged” means “the number of different
identifiers is at least 2 and fewer than the number of polynucleotides that
map to the mappable base position.”6 PO Resp. 16.
As an initial matter, we acknowledge that, in our Institution Decision,
we referred to “non-uniquely tagged” as both “the number of different
identifiers attached to the polynucleotides is at least 2 and fewer than the
number of polynucleotides” and “the number of different identifiers attached
to the polynucleotides is at least 2 and fewer than the number of
polynucleotides that map to the mappable base position.” Compare Inst.
Dec. 9 (“Petitioner contends that ‘[t]he term “non-uniquely tagged” means
that the number of different identifiers attached to the polynucleotides is at
least 2 and fewer than the number of polynucleotides.’ We agree.” (quoting
Pet. 21)), with id. (stating, “In the context of ‘non-unique’ identifiers, the
’822 patent follows the words ‘non-uniquely tagged’ with ‘that is, the
number of different identifiers can be at least 2 and fewer than the number of
polynucleotides that map to the mappable base position.’” (quoting
Ex. 1001, 41:42–46)). As a result, both parties claim that the Board adopted
their respective construction of “non-uniquely tagged.” See PO Resp. 15–16
(stating that “[t]here is no dispute as to the express definition of ‘non-
uniquely tagged’ in the ’822 patent”); Pet. Reply 3 (stating that “[t]he Board
should reject Patent Owner’s arguments and reaffirm its construction, which
is identical to the District Court’s”).

6 The Declarants for both parties agree that an ordinarily skilled
artisan would understand that “mappable base position” means a position in
the reference sequence to which polynucleotide molecules can be
confidently mapped. See Ex. 2027 ¶ 67; Ex. 2022, 23:22–24:13, 30:4–18,
35:5–7; Ex. 2025 ¶ 65. We apply that understanding for this Decision.
IPR2019-00652
Patent 9,834,822 B2
17
To be clear, we have considered the totality of the arguments and
evidence anew at the close of trial, as well as the reasoning set forth in our
Institution Decision, and determine, for the reasons discussed below, that
“non-uniquely tagged” should be construed to mean that “the number of
different identifiers is at least 2 and fewer than the number of
polynucleotides” in a sample, without the additional phrase “that map to a
mappable position.”
At the heart of the parties’ dispute is the following passage from the
’822 patent:
Accordingly, this invention also provides compositions of tagged
polynucleotides. The polynucleotides can comprise fragmented
DNA, e.g. cfDNA. A set of polynucleotides in the composition
that map to a mappable base position in a genome can be non-
uniquely tagged, that is, the number of different identifiers can
be [] at least 2 and fewer than the number of polynucleotides that
map to the mappable base position.
Ex. 1001, 41:42–47 (emphasis added).
The parties agree that “non-uniquely tagged” means that “the number
of different identifiers [i.e., the tag count7] can be at least 2 and fewer than
the number of polynucleotides,” but disagree whether the phrase “that map
to the mappable base position” should be included in the construction of
“non-uniquely tagged.” Put differently, the parties agree that, to be “non-
uniquely tagged,” the lower limit for the tag count is two, but disagree on the
upper limit for the tag count.

7 As relevant here, the ’822 patent refers to the “number of unique
identifiers” as the “tag count.” Ex. 1001, 40:63. For clarity, we use these
terms interchangeably in this Decision.
IPR2019-00652
Patent 9,834,822 B2
18
2. Patent Owner’s Arguments
Patent Owner argues that the upper limit on the tag count is “fewer
than the number of polynucleotides that map to the mappable base position.”
PO Resp. 15; see also PO Sur-Reply 6 (arguing that “the ’822 patent
expressly defines ‘non-uniquely tagged’ as limited by the number of
polynucleotides that map to the mappable base position”). Relying on
Dr. Quackenbush’s Declaration, Patent Owner offers the following
schematic distinguishing between (a) the total polynucleotide fragments in a
sample, (b) the number of polynucleotides that map to a given mappable
base position, and (c) cognates.8 PO Resp. 13.

Schematic submitted by Patent Owner distinguishing between
the polynucleotides in a sample. PO Resp. 13.
According to Patent Owner, the upper limit on the number of different
identifiers does not depend on the total polynucleotide fragments in a sample
(shown by the light blue circle). Id. at 14. Instead, the tag count “depends

8 The ’822 patent defines “cognates” (or “duplicates”) as “more than
one polynucleotide from different genomes [that] have the same start and
stop positions.” Ex. 1001, 40:4–8.
IPR2019-00652
Patent 9,834,822 B2
19
. . . on a subpopulation of fragments that map to a given position in the
reference genome” (shown by the medium blue circle) “and the expected
number of cognates” (shown by the dark blue circle). Id. (citing Ex. 1001,
41:6–13). Patent Owner argues that the ’822 patent expressly defines “non-
uniquely tagged” as limited by the number of polynucleotides that map to
the mappable base position, and that any argument otherwise “strains
credulity.” PO Sur-Reply 5–6; see also PO Reply 15.
3. Petitioner’s Arguments
Petitioner contends that the upper limit on the number of different
identifiers is “fewer than the number of polynucleotides” in a sample.
Pet. 21. Specifically, Petitioner contends that, when “fewer barcodes than
original DNA fragments” are used, “not every fragment has a unique
barcode,” and thus the population of polynucleotides are “non-uniquely
tagged.” Id. at 11–12. As an example, Petitioner contends that “if there are
1000 DNA fragments and only 500 different barcodes, [then] each fragment
does not have its own unique barcode,” and thus the population of
polynucleotides is “non-uniquely tagged” in the context of claim 1. Id. at 12
(citing Ex. 1002 ¶ 51). Conversely, “if there are 1000 original DNA
fragments and at least 1000 different barcodes, [then] each fragment will
have its own unique barcode.” Id. at 11 (citing Ex. 1002 ¶ 50). Petitioner
contends that we should not include “that map to the mappable base
position” in the construction of “non-uniquely tagged” because that phrase
represents “a single embodiment” of the ’822 patent and because that
construction “is inconsistent” with Patent Owner’s arguments in District
Court about the meaning of “non-uniquely tagged.” Pet. Reply 3–10.
IPR2019-00652
Patent 9,834,822 B2
20
4. Analysis
We begin with the words of claim 1. See Amgen Inc. v. Hoechst
Marion Roussel, Inc., 457 F.3d 1293, 1301 (Fed. Cir. 2006) (“claim
construction must begin with the words of the claims themselves”). The first
step of claim 1 recites “providing a population of cell free DNA (‘cfDNA’)
molecules,” and the second step recites “converting the population of
cfDNA molecules into a population of non-uniquely tagged parent
polynucleotides.” Ex. 1001, 62:19–24 (emphases added). We observe that
nothing in the plain language of claim 1 requires the population of cfDNA
molecules that undergoes tagging (i.e., the population that is converted into a
population of “non-uniquely tagged parent polynucleotides”) to be limited to
a population where each polynucleotide maps to a mappable base position.
Instead, claim 1 recites that the population (i.e., not necessarily the mappable
population) is converted into non-uniquely tagged parent polynucleotides.
Id. For these reasons, we determine that the words of claim 1 do not support
Patent Owner’s argument that the tag count is limited by the number of
polynucleotides that map to a mappable base position.
We now turn to the intrinsic record. Starting with the passage
reproduced above that we identified as the heart of the parties’ dispute
(Ex. 1001, 41:42–47), we acknowledge that the use of “that is” (or “i.e.”)
typically “signals an intent to define the word to which it refers.” Edwards
Lifesciences LLC v. Cook Inc., 582 F.3d 1322, 1334 (Fed. Cir. 2009). Here,
however, we agree with Petitioner that the use of “non-uniquely tagged” in
this passage of the written description refers to a specific embodiment of the
’822 patient—that is, an embodiment where the set of polynucleotides map
to a mappable base position. See Pet. 4–5. The ’822 patent describes
various “sets” of polynucleotides and states that, “[i]n some embodiments
IPR2019-00652
Patent 9,834,822 B2
21
each polynucleotide in a set is mappable to a reference sequence.” Ex. 1001,
6:1–2 (emphasis added). The ’822 patent also states that, “[i]n some
embodiments, each set of parent polynucleotides is mappable to a position in
a reference sequence, and the polynucleotides in each set are not uniquely
tagged.” Id. at 18:39–42 (emphasis added). That each polynucleotide in a
set is mappable in some embodiments necessarily implies that, in other
embodiments, not every polynucleotide is mappable. Moreover, that each
parent polynucleotide is mappable and non-uniquely tagged in some
embodiments suggests that mappability and non-unique tagging are not
necessarily concomitant. See WesternGeco LLC v. ION Geophysical Corp.,
889 F.3d 1308, 1323–24 (Fed. Cir. 2018) (“It is well established that claims
are not limited to preferred embodiments, unless the specification clearly
indicates otherwise.”).
Next, we determine that other portions of the ’822 patent do not
support Patent Owner’s argument that the upper limit of the tag count
depends on the number of polynucleotides that map to a mappable base
position. PO Resp. 13–14. The ’822 patent explains that, when
polynucleotides have a tag count equal to 1 (“equivalent to having no unique
tags or not tagging”), “it may not be possible to determine which sequence
reads are derived from which parent molecules.” Id. at 40:34–61. The ’822
patent further explains that “[t]his problem can be diminished by tagging
parent molecules with a sufficient number of unique identifiers.” Id. at
40:61–63. According to the ’822 patent, the methods of the prior art
“uniquely tag[ged] every, or nearly every, different parent molecule in the
sample.” Id. at 40:67–41:2. But, these prior art methods were “cumbersome
and expensive,” because they required “billions of different unique
identifiers.” Id. at 41:2–6.
IPR2019-00652
Patent 9,834,822 B2
22
The ’822 patent explains that “[t]his invention” solves the problem of
distinguishing polynucleotides by tagging a population of DNA “with n
different unique identifiers, wherein n is at least 2 and no more than
100,000*z, wherein z is a measure of the central tendency (e.g., mean,
median, mode) of an expected number of duplicate molecules having the
same start and stop positions” (i.e., cognates). Id. at 41:6–13. The ’822
patent describes embodiments where the lower limit of the tag count is
between 2*z to 20*z, and other embodiments where the upper limit of the
tag count is between 100*z to 100,000*z. Id. at 41:13–18.9 The ’822 patent
then expressly states that “n can range from any combination of these upper
and lower limits.” Id. at 41:18–19.
Thus, the ’822 patent clearly teaches the ordinarily skilled artisan that
the number of different identifiers—that is, the tag count “n”—is based on
the number of expected cognates in a sample (“z”), which in turn “is a
function of the number of haploid genome equivalents in a sample and the
distribution of fragment sizes.” Ex. 1001, 40:8–11; 41:6–39; see also PO
Resp. 12 (“The inventor of the ’822 patent determined empirically that the
number of cognates in a sample is proportional to the size distribution of the
polynucleotide fragments and the number of genomic equivalents—that is,
the number of haploid genome copies present in the sample.”).
Patent Owner does not persuasively point us to any teaching in this
description—which Dr. Quackenbush admits describes a “non-uniquely

9 The ’822 patent’s reference to “1000,000*z” at column 41, line 17,
appears to be a typographical error. See Ex. 2001 (Patent Owner’s claim-
construction brief quoting the same passage from the ’822 patent as “2 and
no more than 100,000*z”). For this Decision, we read that term as
“100,000*z.”
IPR2019-00652
Patent 9,834,822 B2
23
tagged” embodiment of the ’822 patent—that instead correlates the upper
limit of the tag count to a subpopulation of polynucleotides in a sample that
map to a mappable base position. See Ex. 2025 ¶ 58 (citing Ex. 1001, 41:6–
13); Ex. 1103, 36:18–21, 37:7–11, 14–16.10 Rather, according to the ’822
patent—in an embodiment described as “[t]his invention”—the tag count
can encompass up to 100,000*z different identifiers. Ex. 1001, 41:6–13.
The ’822 patent also teaches that, “in a sample of about 10,000 haploid
genome equivalents of human DNA, there are about 3 duplicate
polynucleotides beginning at any given position.” Id. at 40:24–27, 41:10–
13. Thus, for a sample comprising 10,000 haploid copies of
polynucleotides, z=3, the sample can be tagged with up to 100,000*3
(300,000) identifiers and still be considered “non-uniquely tagged.” Id. at
41:6–13. This number of unique identifiers (300,000) is far greater than the
number of polynucleotides that would map to a given mappable base
position (10,000) in a sample of 10,000 haploid genome equivalents under
Patent Owner’s construction. See PO Resp. 17–19 (arguing that for “for
non-uniquely tagged,” a single copy of the genome would produce 1
polynucleotide that maps to a mappable base position, and five haploid
genome equivalents would produce 5 polynucleotides that map to a
mappable base position).

10 Indeed, claim 12 contrasts the expected number of duplicate
molecules with the population of polynucleotides in the sample. See
Ex. 1001, 63:15–19 (Claim 12, reciting “The method of claim 1, wherein the
population of polynucleotides is tagged with n different unique identifiers,
wherein n is no more than 100*z, wherein z is a mean of an expected
number of duplicate molecules having the same start and stop positions in
the sample.” (emphases added)).
IPR2019-00652
Patent 9,834,822 B2
24
Patent Owner appears to acknowledge that this embodiment would be
outside the scope of claim 1, but nevertheless argues that this “is of no
moment” because it “is not required to claim all that it discloses.” PO Sur-
Reply 6–7. Even so, we find that Patent Owner has not provided an
adequately persuasive reason, based on either the claim language or the ’822
patent, for doing so here. See SynQor, Inc. v. Artesyn Techs., Inc., 709 F.3d
1365, 1378–79 (Fed. Cir. 2013) (“A claim construction that excludes the
preferred embodiment is rarely, if ever, correct and would require highly
persuasive evidentiary support.” (quotation marks and citation omitted)); see
also In re Katz Interactive Call Processing Patent Litig., 639 F.3d 1303,
1324 (Fed. Cir. 2011) (“there is a strong presumption against a claim
construction that excludes a disclosed embodiment”).
Turning to the extrinsic evidence, we note that Patent Owner argued
in its Opening Brief on Claim Construction to the District Court in the co-
pending litigation that, as to the construction of “non-uniquely tagged,”
“[t]he dispute between the parties is whether the number of ‘non-unique’
barcodes can approach the total number of polynucleotides.” Ex. 2001, 7.
This accords with the dispute between the parties in this proceeding as well.
But before the District Court, unlike here, Patent Owner argued that “the
total number of barcodes must plainly be less than the total number of
polynucleotides.” Id. at 9 (emphasis added). Patent Owner also argued that
“in describing the embodiments that use ‘non-unique’ barcoding, the
specification explains that they use between ‘2 and no more than 100,000*z’
barcodes ‘wherein z is a measure of central tendency (e.g., mean, median,
mode) of an expected number of duplicate molecules.’” Id. at 8 (quoting
Ex. 1001, 41:10–13). Thus, unlike here, Patent Owner did not rely on the
IPR2019-00652
Patent 9,834,822 B2
25
“number of polynucleotides that map to the mappable base position” as
setting forth the upper limit for the number of different identifiers. Id.
Patent Owner argues that its previous statements to the District Court
were made simply to distinguish the claims over the prior art at issue in the
co-pending litigation and are not inconsistent with its construction of “non-
uniquely tagged” here. PO Sur-Reply 6. We disagree. In its claim-
construction brief, Patent Owner criticized the opposing parties’ proposed
constructions as “uniformly at odds with the intrinsic record,” Ex. 2001, 1
(emphasis added), and argued that its own construction of “non-uniquely
tagged” was most consistent with the specification and the plain meaning of
“non-uniquely tagged,” id. at 7–9. Although we do not consider Patent
Owner’s arguments in the co-pending litigation dispositive, we find that they
nonetheless further support our construction of “non-uniquely tagged” here.
For all the above reasons, we determine that “non-uniquely tagged”
means that “the number of different identifiers can be at least 2 and fewer
than the number of polynucleotides” in a sample. Thus, a population of
parent polynucleotides is “non-uniquely tagged” in the context of claim 1
when the number of different identifiers (the tag count) is at least 2 but
fewer than the number of parent polynucleotides in the sample.
D. Asserted References
Before turning to Petitioner’s asserted grounds of unpatentability, we
provide a brief summary of the asserted references.
1. Schmitt (Ex. 1011)
Schmitt relates to a method for sequencing target DNA fragments
called Duplex Consensus Sequencing (“DCS”). Ex. 1011, code (57).
According to Schmitt, the DCS method greatly reduces sequencing errors by
independently tagging and sequencing each of the two strands of a target
IPR2019-00652
Patent 9,834,822 B2
26
DNA fragment. Id. Because the two strands of DNA are complementary,
true mutations can be found at the same position on both strands, as opposed
to on a single strand if PCR or sequencing errors occur. Id.
Schmitt’s DCS methods utilize “a single molecule identifier (SMI)
adaptor molecule for use in sequencing a double-stranded target nucleic acid
molecule.” Ex. 1011, 2:66–3:1. Schmitt teaches that the SMI adaptor
molecule comprises an SMI sequence or “tag” (which is a degenerate or
semi-degenerate n-mer sequence11) and an SMI ligation adaptor (which
allows the SMI adaptor molecule to be ligated to the target DNA fragment).
Id. at 3:1–9; see also id. at 5:57–59 (providing that the “SMI adaptor
molecule is double stranded, and may include a single molecule identifier
(SMI) sequence, and an SMI ligation adaptor”); id. at 6:46–47 (referring to
the SMI sequence as a “tag”). The DCS method includes the steps of:
ligating a double-stranded target nucleic acid molecule to at least
one SMI adaptor molecule to form a double-stranded SMI-target
nucleic acid complex; amplifying the double-stranded SMI-
target nucleic acid complex, resulting in a set of amplified SMI-
target nucleic acid products; and sequencing the amplified SMI-
target nucleic acid products.
Id. at 3:12–20.
In one embodiment, the SMI sequence is “a unique, double-stranded,
complementary n-mer random tag,” such that every DNA fragment becomes
labeled with two distinct SMI sequences. Id. at 3:47–53. In this
embodiment, the “nucleotide n-mer sequences may be any suitable length to
produce a sufficiently large number of unique tags to label a set of sheared

11 It is well known in the art that an “n-mer” is a short sequence
comprising “n” number of nucleotides. For example, a “4-mer”
polynucleotide is a short sequence of four nucleotides in a DNA sequence.
IPR2019-00652
Patent 9,834,822 B2
27
DNA fragments from a segment of DNA.” Id. at 6:59–63. Schmitt provides
as an example a “nucleotide n-mer sequence which is 12 nucleotides in
length” that, once ligated to each end of the target DNA fragment, “results in
the generation of up to 424 (i.e., 2.8 x 1014) distinct tag sequences.” Id. at
6:66–7:5.
In another embodiment, referred to as the “hybrid method,” the SMI
sequence comprises “a shorter n-mer tag (such as 1 or 2 or 3 or 4 or more
degenerate or semi-degenerate bases).” Id. at 9:9–13. In this embodiment,
the SMI adaptor molecules “serve as unique molecular identifiers” by
combining information from the sheared ends of the target DNA fragment
and the short n-mer tag. Id.
Schmitt teaches that, for error correction through DCS, sequence
reads sharing a unique set of SMI tags are grouped into paired families, each
pair reflecting one double-stranded DNA fragment. Id. at 4:4–10.
Mutations present in only one or a few family members, or mutations
occurring in only one of the two strands, represent sequencing mistakes or
PCR-introduced errors. Id. at 4:10–18. True mutations are present on both
strands and appear in all members of a family pair. Id. at 4:18–20.
2. Schmitt 2012 (Ex. 1047)
Schmitt 2012 also relates to the DCS method and uses the same
library-preparation techniques and sequencing methods disclosed in Schmitt.
Ex. 1047, Abstract, 14509 (Fig. 1). Schmitt 2012 discloses a “data
processing” workflow comprising the steps of filtering sequence reads,
aligning reads to a reference genome, grouping reads containing identical tag
sequences to form single-strand consensus sequence reads (“SSCS”), re-
aligning reads to the reference genome, and identifying and comparing
IPR2019-00652
Patent 9,834,822 B2
28
partner strands among SSCS reads compared to form DCS reads. Id. at
14513, SI1–SI2; see also Ex. 1002 ¶ 91.
3. Fan (Ex. 1048)
Fan relates to a method for diagnosing fetal aneuploidy by directly
sequencing cfDNA from the plasma of pregnant women with high-
throughput shotgun sequencing technology. Ex. 1048, Abstract. In doing
so, Fan was able to measure the over- and under-representation of
chromosomes from an aneuploidy fetus. Id.
4. Forshew (Ex. 1004)
Forshew relates to a method for identifying cancer mutations present
in cfDNA using tagged-amplicon deep sequencing (“TAm-Seq”). Ex. 1004,
Abstract. Tagged-amplicon deep sequencing allows amplification and deep
sequencing of genomic regions spanning thousands of bases. Id. at 1.
Forshew applied the technique to both abundant and rare mutations in
circulating DNA from blood plasma of ovarian and breast cancer patients.
Id. at 1–2.
E. Obviousness over Schmitt in View of Schmitt 2012, and Further in
View of Fan or Forshew
Petitioner contends that claims 1–13 and 17–20 of the ’822 patent are
unpatentable as obvious over Schmitt, Schmitt 2012, and Fan or Forshew.
Pet. 30–70. Patent Owner opposes. PO Resp. 19–62; PO Sur-Reply 3–26.
Having considered the totality of the arguments and evidence, we find that
Petitioner has shown by a preponderance of the evidence that claims 1–11,
13, and 17–20 are unpatentable as having been obvious over Schmitt,
Schmitt 2012, and Fan or Forshew, but has not shown that claim 12 would
have been obvious.
IPR2019-00652
Patent 9,834,822 B2
29
1. Limitations of the challenged claims
Petitioner contends that the combination of Schmitt, Schmitt 2012,
and Fan or Forshew discloses or suggests each element of the challenged
claims. Petitioner presents arguments mapping the language of claims 1–13
and 17–20 to the disclosures of each reference. Pet. 41–71.
Independent Claim 1
Step (a) of claim 1 recites “providing a population of cell-free DNA
(‘cfDNA’) molecules obtained from a bodily sample from a subject.”
Ex. 1001, 62:19–21. We agree with Petitioner that Fan and Forshew teach
this limitation. Pet. 41–42. Fan describes extracting cfDNA from the
plasma of pregnant women to detect fetal aneuploidy. See Ex. 1048, 16266
(“We directly sequenced cell-free DNA with high-throughput shotgun
sequencing technology form plasma of pregnant women . . . .”), 16270
(stating that “DNA was extracted from cell-free plasma”). Specifically, Fan
extracted between 1.2 and 8 ng of cfDNA from plasma samples ranging
from 1.3 to 3.2 ml. Id. at SI7 (Table S1); see also Ex. 1002 ¶ 122.
Similarly, Forshew describes extracting cfDNA from plasma samples to
identify mutations in cancer patients. See Ex. 1004, 1 (stating that “[p]lasma
of cancer patients contains cell-free tumor DNA”), 10 (stating that
“[c]irculating DNA was extracted from between 0.85 and 2.2 ml of
plasma”). As shown in Table S6, Forshew obtained amounts ranging from
0.9 to 19.7 ng of cfDNA from plasma samples of cancer patients. Id. at 32
(Table S6); see also Ex. 1002 ¶ 23.
Turning to step (b) of claim 1, the first portion of this step recites
“converting the population of cfDNA molecules into a population of non-
uniquely tagged parent polynucleotides.” Ex. 1001, 62:22–24. We agree
with Petitioner that Schmitt in combination with Fan or Forshew teaches this
IPR2019-00652
Patent 9,834,822 B2
30
limitation. Pet. 42–46. Schmitt teaches that, in the DCS method, SMI
adaptor molecules comprising an “SMI sequence (or ‘tag’) of nucleotides”
and an SMI ligation adaptor are ligated to the ends of target DNA
polynucleotides. See Ex. 1011, 3:1–9, 5:57–59, 6:46–51. Schmitt refers to
the SMI sequences as a “nucleotide n-mer” sequence. Id. at 6:46–66.
As explained above, we construe “non-uniquely tagged” to mean that
the number of different identifiers attached to the parent polynucleotides is
at least 2 but fewer than the number of polynucleotides in the sample. Supra
§ IV.C.4. Schmitt discloses a DCS “hybrid method” tagging target DNA
polynucleotides that uses “a combination of sheared ends and a shorter n-
mer tag (such as 1 or 2 or 3 or 4 or more degenerate or semi-degenerate
bases) in the adaptor.” Ex. 1011, 9:9–11 (emphasis added); see also Pet. 43–
46; Ex. 1002 ¶¶ 126–127. Schmitt also expressly discloses examples of 4-
mer tags. See Ex. 1011, 4:30–54; id. at 9:9–13. Dr. Gabriel testifies without
rebuttal that using a 4-mer tag yields 44, or 256, different tag sequences—
which when ligated at both ends of a parent polynucleotide—yields 2562, or
65,536 different or “unique” identifiers. See Ex. 1002 ¶ 127 & n.7; Pet. 44–
45; see also Ex. 2025 ¶ 69 (testimony of Dr. Quackenbush as to “[t]he
65,536 possible unique tags from a 4-mer sequence”); PO Resp. 18 (accord).
We agree with Petitioner that the use of n-mer tags in the hybrid DCS
method results in a population of “non-uniquely tagged parent
polynucleotides,” because the n-mer tags are not “unique” for each parent
polynucleotide. That is, the number of different n-mer tags is fewer than the
number of polynucleotides in the sample because Schmitt explicitly
distinguishes the hybrid method from embodiments in which every
polynucleotide in the sample is tagged with a unique SMI identifier. See
Ex. 1011, 9:1–4 (stating that the DCS method “does not strictly require the
IPR2019-00652
Patent 9,834,822 B2
31
use of an SMI tag, as the sheared ends can be used as identifiers to
differentiate unique individuals molecules from PCR duplicates” (emphasis
added)); see also id. at 6:59–63 (stating that, in the main embodiment, the
“nucleotide n-mer sequences may be any suitable length to produce a
sufficiently large number of unique tags to label a set of sheared DNA
fragments from a segment of DNA”). Finally, although Schmitt does not
expressly teach that the target polynucleotide is cfDNA, both Fan and
Forshew teach the extraction and analysis of cfDNA as described
immediately above.
The second portion of step (b) provides that “each of the non-uniquely
tagged parent polynucleotides comprises (i) a sequence from a cfDNA
molecule of the population of cfDNA molecules, and (ii) an identifier
sequence comprising one or more polynucleotide barcodes.” Ex. 1001,
62:24–28. We again agree with Petitioner that Schmitt teaches this
limitation. Pet. 45. Schmitt describes embodiments (a–c) in Figure 3, in
which sequence reads are derived from parent polynucleotides comprising
both a “strand identifier” and a sequence derived from the parent DNA
molecule. Ex. 1011, 4:4–54; see also Ex. 1002 ¶ 129.
Turning to the remaining limitations of claim 1, we agree with
Petitioner’s contentions and supporting evidence about steps (c) through (h)
of the claimed method. See Pet. 47–56. In particular, we agree with
Petitioner’s contentions that Schmitt, either alone or in combination with
Schmitt 2012, discloses:
• amplifying the tagged polynucleotides and sequencing the amplified
progeny (steps (c) and (d) of claim 1), see Pet. 47; Ex. 1011, 3:10–20
(disclosing “amplifying the double-stranded SMI-target nucleic acid
complex, resulting in a set of amplified SMI-target nucleic acid
IPR2019-00652
Patent 9,834,822 B2
32
products; and sequencing the amplified SMI-target nucleic acid
products”); see also id. at 21:55–57; Ex. 1002 ¶¶ 132–133;
• mapping the sequence reads to a reference sequence (step (e) of
claim 1), see Pet. 47–49; Ex. 1011, 20:39–64 (teaching aligning
sequence reads “to the human genome with the Burrows Wheeler
Aligner (BWA)”); id. at 23:10–14, 24:33–37; Ex. 1047, SI1 (“Reads
were then aligned to the reference genome with the Burrows Wheeler
aligner (BWA) and nonmapping reads were discarded. . . . Reads
sharing identical tag sequences were then grouped together and
collapsed to consensus reads.”); see also Ex. 1002 ¶¶ 134–137, 139;
• grouping the sequence reads into families (step (f) of claim1), see
Pet. 40–51; Ex. 1011, 4:9–10, 9:9–14, 17:61–18:2, 20:39–64, 21:55–
61; see also Ex. 1002 ¶¶ 140–144;
• collapsing sequence reads in each family to yield a “base call” (step
(g) of claim 1), see Pet. 51–55; Ex. 1011, 4:20–29, 4:55–66, 18:44–
61, 20:50–56; 22:50–53, 23:19–22, 23:61–24:1, 25:20–30, Figures 3
and 5C; see also Ex. 1002 ¶¶ 145–152; and
• determining the frequency of bases called at the locus from among the
families (step (h) of claim 1), see Pet. 55–56; Ex. 1011, 4:25–29,
Figure 3; see also Ex. 1002 ¶¶ 154–155.
Analysis of Patent Owner’s Arguments
We are not persuaded by Patent Owner’s arguments that the
combination of Schmitt, Schmitt 2012, and Fan or Forshew fails to teach or
suggest all the limitations of claim 1. PO Resp. 19–56; PO Sur-Reply 3–26.
Patent Owner first argues that Schmitt fails to disclose “non-uniquely tagged
parent polynucleotides” as this phrase should be correctly construed, and
that the Petition fails because Petitioner never applied the correct
IPR2019-00652
Patent 9,834,822 B2
33
construction of “non-uniquely tagged” to the prior art in the Petition.
PO Resp. 20–23; PO Sur-Reply 3–4. But, as explained in detail above, we
do not read “non-uniquely tagged” as limiting the tag count to a
subpopulation of polynucleotides that map to a mappable base position.
Supra § IV.C.4. Thus, we disagree with Patent Owner that Petitioner failed
to apply the correct construction of “non-uniquely tagged parent
polynucleotides” in its Petition.
Patent Owner’s next argument, that “Petitioner’s contrived scenario
does not support its assertion that Schmitt teaches non-uniquely tagged
parent polynucleotides,” PO Resp. 23–27, also relies on its unduly narrow
construction of “non-uniquely tagged parent polynucleotides.” In this
regard, Patent Owner argues that “Schmitt is silent regarding a sub-
population of fragments in a sample that map to a given mappable base
position,” id. at 20–21, and that an ordinarily skilled artisan would not have
been able to determine the number of polynucleotides that map to a
mappable base position from Schmitt’s disclosure, id. at 23 (citing Ex. 2025
¶ 72). But, again, we do not read “non-uniquely tagged parent
polynucleotides” as limiting the tag count to a subpopulation of
polynucleotides that map to a mappable base position. Supra § IV.C.4.
Patent Owner’s argument therefore fails, and its follow-on arguments about
Dr. Quackenbush’s testimony and Dr. Gabriel’s cross-examination
testimony as to the tagging and sequencing of one haploid human genome
equivalent, see PO Resp. 24–26, are irrelevant.12

12 Similarly, Patent Owner’s argument that “Schmitt does not disclose
sequencing targets as large as the entire 3 billion base pair haploid human
genome,” PO Resp. 26, is unpersuasive because nothing in claim 1 limits the
size of the sequencing target. See Ex. 1001, 62:18–48; see also In re Hiniker
IPR2019-00652
Patent 9,834,822 B2
34
We also disagree with Patent Owner’s argument that Schmitt fails to
disclose “non-uniquely tagged parent polynucleotides.” PO Resp. 27–33
(arguing that “Schmitt exclusively teaches unique tagging methods”). As
explained above, supra § IV.D.1., Schmitt discloses methods for sequencing
target parent polynucleotides that utilize SMI sequences as tags. Ex. 1011,
2:66–3:1. In the main embodiment, the SMI sequence is “a unique, double-
stranded, complementary n-mer random tag,” and every target
polynucleotide is labeled with two distinct SMI sequences. Id. at 3:47–53
(emphasis added). In this embodiment, Schmitt teaches that the “nucleotide
n-mer sequences may be any suitable length to produce a sufficiently large
number of unique tags to label a set of sheared DNA fragments from a
segment of DNA.” Id. at 6:59–63. Schmitt provides as an example a
“nucleotide n-mer sequence which is 12 nucleotides in length,” that, once
ligated to each end of the target DNA fragment, “results in the generation of
up to 424 (i.e., 2.8 x 1014) distinct tag sequences.” Id. at 6:66–7:5. There is
no dispute that this embodiment of Schmitt is directed to a unique tagging
method because the number of SMI tags exceeds the number of parent
polynucleotides in the sample. See Ex. 1002 ¶ 126 (testimony of Dr. Gabriel
that “[i]n some embodiments, Schmitt describes the use of a large number of
unique barcode sequences such that there is a high probability that a
different barcode sequence is ligated to each parental DNA template”).
But, contrary to Patent Owner’s arguments otherwise, Schmitt also
teaches an alternative “hybrid” method that we conclude results in the
generation of “non-uniquely tagged parent polynucleotides” as claimed.

Co., 150 F.3d 1362, 1369 (Fed. Cir. 1998) (“the name of the game is the
claim”).

IPR2019-00652
Patent 9,834,822 B2
35
Specifically, Schmitt teaches that the DCS method “does not strictly require
the use of an SMI tag” because the sheared ends of the target parent DNA
polynucleotides “can be used as identifiers to differentiate unique individual
molecules from PCR duplicates.” Ex. 1011, 9:1–4. Schmitt acknowledges,
however, that “there are a limited number of shear points flanking any given
genomic position and thus the power to sequence deeply is increased via
inclusion of the SMI tag.” Id. at 9:6–9. Schmitt thus proposes a solution
that utilizes both the sequence information from the sheared ends and an n-
mer: the “hybrid” method. Id. at 9:9–13. Specifically, Schmitt teaches that
the hybrid method uses “a shorter n-mer tag (such as 1 or 2 or 3 or 4 or more
degenerate or semi-degenerate bases)” in combination with the sheared ends
of the target DNA fragment to “serve as unique molecular identifiers.” Id.
We find credible, and supported by preponderant record evidence,
Dr. Gabriel’s testimony that the combination of the sheared ends and a non-
unique short n-mer tag results in a “unique molecular identifier[].” Ex. 1002
¶ 126 (emphasis added); see also Ex. 1011, 9:1–13. Specifically, we are
persuaded by Dr. Gabriel’s testimony that the short n-mer tag used in
Schmitt’s hybrid method is itself not unique for each parent polynucleotide
in the sample. Ex. 1002 ¶ 127. Schmitt expressly teaches that the “shorter
n-mer tag” may comprise 4 or fewer nucleotide bases. Ex. 1011, 9:10–12.
From this disclosure—as Dr. Gabriel persuasively testifies and we agree—
an ordinarily skilled artisan would have readily recognized that a 4-mer, for
example, yields 44, or 256, different tag sequences, which when ligated at
both ends of a parent polynucleotide, would yield 2562, or 65,536 different
or “unique” identifiers. Ex. 1002 ¶ 127 & n.7; see also Pet. 44–45. Indeed,
Dr. Quackenbush acknowledges that “65,536 possible unique tags [result]
from a 4-mer tag sequence.” Ex. 2025 ¶ 69.
IPR2019-00652
Patent 9,834,822 B2
36
An ordinarily skilled artisan would have readily recognized that this
number of unique identifiers is far less than the millions or more parent
polynucleotides that would result from Schmitt’s size selection for DNA
fragments in the range of 200 to 500 base pairs. Ex. 1002 ¶ 127 (citing
Ex. 1011, 22:43–46 (stating that “DNA for sequencing was sheared and end-
repaired by standard methods, with size-selection for fragments in the range
of ~200–500 bp by size-selecting binding”)). Thus, as Dr. Gabriel testifies
and we agree, more than one parent polynucleotide would necessarily share
the same short n-mer tag in Schmitt’s hybrid method, because the number of
unique identifiers (i.e., a tag count of 65,536) is at least 2 but fewer than the
number of parent polynucleotide fragments in the sample (i.e., millions or
more), as we have construed “non-uniquely tagged parent polynucleotides.”
Id.
Patent Owner argues that Petitioner was wrong to “assume a
combination of more fragments in a sample (i.e., millions of fragments) than
SMI tags (i.e., 65,536 tags) in order to conclude that the number of
fragments is greater than the number of tags.” PO Resp. 28. In this regard,
Patent Owner argues that “Schmitt discloses applying DCS to much smaller
targets—‘ranging from ~300 bp to ~20 kb in size’” and that shearing those
targets into 200 bp fragments would result in, at most, between 1 and 100
fragments, “far fewer fragments than ‘millions of fragments’” Petitioner
asserts would be in a sample. Id. at 29 (citing Ex. 1011, 19:65–66, 24:7–11,
Ex. 2025 ¶ 85). Although we have carefully considered Patent Owner’s
argument, we find that it lacks evidentiary support in the disclosure of
Schmitt and, therefore, is not persuasive.
Schmitt clearly discloses shearing 3 micrograms of DNA into
fragments, ligating the SMI adaptor molecules to 750 ng of T-tailed DNA,
IPR2019-00652
Patent 9,834,822 B2
37
and amplifying (by PCR) 375 ng of the resulting adaptor-ligated (or tagged)
DNA. Ex. 1011, 19:58–20:23. Then, for sequence analysis, Schmitt
discloses that “an arbitrary 758 kb region of the genome consisting of both
coding and noncoding sequences” was “targeted” and “capture[d]” using
capture baits. Id. at 20:27–30. The much smaller fragments to which Patent
Owner refers—i.e., DNA molecules ranging from ~300 bp to ~20 kb in
size—represent DNA targets that would have been similarly captured for the
purpose of sequence analysis after the process of shearing and tagging. See
id. at 20:16–30 (describing the process of pre-capture amplification and
DNA capture); 24:7–9 (describing the ~300 bp to ~20 kb DNA molecules as
“vastly smaller target DNA molecules” (emphasis added)); Ex. 1104 ¶ 18
(explaining that “Schmitt discloses tagging DNA fragments before target
capture, not after”). An ordinarily skilled artisan would have understood
from Schmitt’s disclosure, then, that these ~300 bp to ~20 kb DNA
molecules would not have been subjected to shearing and ligation reactions,
because those molecules would have been already tagged with the SMI
sequence identifiers. Ex. 1104 ¶ 18. For these reasons, Patent Owner’s
argument (and Dr. Quackenbush’s corresponding testimony) about the
number of polynucleotide fragments that would result from further shearing
of 300 bp to ~20 kb DNA molecules reflects a misreading of Schmitt’s
disclosure and is not persuasive. See Ex. 2025 ¶ 85 (Dr. Quackenbush’s
testimony “[a]ssuming” shearing of Schmitt’s ~300 bp to ~20 kb DNA
targets).
Moreover, we find that an ordinarily skilled artisan would have
understood that the total number of polynucleotides in Schmitt’s sample of
T-tailed DNA would have far exceeded the 65,536 possible unique tags
resulting from a 4-mer sequence, and thus reads on a “a population of non-
IPR2019-00652
Patent 9,834,822 B2
38
uniquely tagged parent polynucleotides.” Dr. Quackenbush admits that, for
“a sample comprised of five haploid copies of the human nuclear genome
that are fragmented into 200 base pair polynucleotides,” “[t]he entire sample
comprises about 75 million polynucleotide fragments,” which is much more
than “[t]he 65,536 possible unique tags from a 4-mer tag sequence.”
Ex. 2025 ¶ 69 (emphasis added). There can be no dispute that, if a sample
of five haploid copies of DNA produces about 75 million polynucleotide
fragments, then a sample of 750 ng T-tailed DNA in Schmitt would
represent a much greater amount of polynucleotide fragments. Pet. Reply
11–12; Pet. 66; Ex. 1104 ¶ 19 (stating that a sample of 20 ng of DNA
contains approximately 6,060 haploid genome equivalents, correlating to
billions of polynucleotide fragments); Ex. 1001, 40:15–19 (stating that a
sample of 30 ng of DNA contains approximately 10,000 haploid genome
equivalents, while a sample of 100 ng of DNA contains approximately
30,000 haploid genome equivalents).
Patent Owner argues that the DCS method is different from the
claimed subject matter in the ’822 patent because “[t]he method of Schmitt
expressly requires flooding a sample with ‘SMI tags’ or ‘n-mers’ to ensure
that every DNA fragment in the sample is uniquely labeled.” PO Resp. 5;
see also, e.g., id. at 29–30 (arguing that “where Schmitt does relate the
number of barcodes to polynucleotide fragments in a sample it instructs that
the barcodes should be in excess of the sample fragments” (citing Ex. 1011,
3:47–53, 6:59–62; Ex. 2025 ¶ 86)). Although we agree with Patent Owner
that the disclosure of Schmitt is directed, in the main, to the use of an excess
number of unique SMI sequences for tagging, Schmitt’s “hybrid” method
utilizes shorter non-unique n-mer SMI sequences that produce, in substance,
a population of “non-uniquely tagged” DNA fragments. See Pet. 44–45; see
IPR2019-00652
Patent 9,834,822 B2
39
also Ex. 1002 ¶¶ 126–127. Patent Owner’s argument that “[l]imiting n-mer
tags as suggested in the petition materials would be contrary to the express
instruction of Schmitt,” PO Resp. 31, is unpersuasive because it ignores
Schmitt’s express disclosure and examples of 4-mer tags. See Ex. 1011,
4:30–54; id. at 9:9–13. Figure 4a of Schmitt, reproduced below, “illustrates
an example of how a SMI sequence with n-mers of 4 nucleotides in length
(4-mers) are read by Duplex Consensus Sequencing.” Id. at 4:30–32.

Figure 4a of Schmitt provides an example of an SMI sequence
comprising an n-mer of 4 nucleotides in length (i.e., a 4-mer).
Ex. 1011, 4:30–54.
And although Schmitt expressly labels the hybrid approach as
providing a “unique molecular identifier,” Ex. 1011, 9:12–13 (emphasis
added), this is so only because the combination with the sequence
information from the sheared ends with the short n-mer tag produces the
“unique molecular identifier.” The ’822 patent similarly describes a “unique
sequence” identifier that results from the combination of a non-unique
barcode and sequence information at the ligation site:
[A] plurality of barcodes may be used such that barcodes are not
necessarily unique to one another in the plurality. In this
example, the barcodes may be ligated to individual molecules
such that the combination of the bar code and the sequence it
may be ligated to creates a unique sequence that may be
individually tracked. As described herein, detecting of non
unique barcodes in combination with sequence data of beginning
(start) and end (stop) portions of sequence reads may allow
assignment of a unique identity to a particular molecule.
IPR2019-00652
Patent 9,834,822 B2
40
Ex. 1001, 39:13–22 (emphases added); see also Pet. 45–46; Ex. 1002 ¶ 130.
Moreover, although claim 1 recites that the parent polynucleotides are “non-
uniquely tagged,” the parent polynucleotides are nevertheless uniquely
identifiable because their amplified copies are grouped into families,
“whereby each of the families comprises sequence reads amplified from the
same tagged parent polynucleotide.” Ex. 1001, 62:40–42.
Finally, we disagree with Patent Owner that Schmitt’s shearing
process distinguishes over the method recited in claim 1 of the ’822 patent.
PO Resp. 33–35. Specifically, Patent Owner argues that “the hybrid method
critically relies on sheared cellular DNA fragments” and that “shearing is
incompatible with cfDNA samples.” Id. at 33–34 (citing Ex. 1011, 9:9–13;
Ex. 2020 ¶ 196; Ex. 2025 ¶¶ 93–95). Patent Owner’s argument is
unpersuasive because it treats the ordinarily skilled artisan like an
automaton. “A person of ordinary skill is also a person of ordinary
creativity . . . .” KSR, 550 U.S. at 421. And here, the record evidence
supports Petitioner’s contention that an ordinarily skilled artisan would have
understood that cfDNA is already fragmented by nature and, thus, would
have simply omitted the shearing step when tagging cfDNA with Schmitt’s
short 4-mer tags. Ex 1002 ¶¶ 118, 143; Ex. 1048, 16270–271 (“Library
preparation was carried out according to the manufacturer’s protocol with
slight modifications. Because cell-free plasma DNA was fragmented in
nature, no further fragmentation by nebulization or sonication was done on
plasma DNA samples.” (emphasis added)); Ex. 1052,13 94 (describing
plasma DNA library preparation and stating that “plasma DNA molecules

13 Gary J.W. Liao et al., Targeted Massively Parallel Sequencing of
Maternal Plasma DNA Permits Efficient and Unbiased Detection of Fetal
Alleles, 57(1) CLIN. CHEM. 92–101 (2011) (Ex. 1052).
IPR2019-00652
Patent 9,834,822 B2
41
[are] already fragmented in nature, so no additional fragmentation step [is]
required”).
Patent Owner’s insinuation that an ordinarily skilled artisan would
have found the shearing step to be an insurmountable obstacle to the
combination of Schmitt and Fan or Forshew is therefore unpersuasive. See
ClassCo, Inc. v. Apple, Inc., 838 F.3d 1214, 1219 (Fed. Cir. 2016) (“The
rationale of KSR does not support [the] theory that a person of ordinary skill
can only perform combinations of a puzzle element A with a perfectly fitting
puzzle element B.”). Indeed, library generation, shearing (or not shearing),
and ligating adaptors to DNA fragments were well known and routine steps
in the art before the earliest priority date for the ’822 patent. See, e.g.,
Ex. 1011, 19:5–20:15 (Example 1); Ex. 1004, 16–34 (describing technical
methods and design); Ex. 1048, 16270–271 (describing library preparation
and adaptor ligation).
For all the above reasons, we agree with Petitioner and Dr. Gabriel
that claim 1 does not distinguish over Schmitt’s “hybrid” method. Pet. 45;
Ex. 1002 ¶ 130. Although the ’822 patent touts that “a sample comprising
about 10,000 haploid human genome equivalents of cfDNA can be tagged
with about 36 unique identifiers,” Ex. 1001, 41:31–33, the tag count of
claim 1 is not so limited. Indeed, dependent claim 11 recites tagging “with
from 10 to 100,000 different identifiers,” Ex. 1001, 63:11–13, thus
suggesting that the number of non-unique tags in claim 1, from which claim
11 depends, is broader. See AK Steel Corp. v. Sollac & Ugine, 344 F.3d
1234, 1242 (Fed. Cir. 2003) (“Under the doctrine of claim differentiation,
dependent claims are presumed to be of narrower scope than the independent
claims from which they depend.”). Thus, we are satisfied that Petitioner
establishes by a preponderance of the evidence that the combination of
IPR2019-00652
Patent 9,834,822 B2
42
Schmitt, Schmitt 2012, and Fan or Forshew teaches or suggests each and
every limitation of claim 1.
Dependent Claims 3, 4, 6–8, 10, 11, and 17–20
Having decided that the combination of Schmitt, Schmitt 2012, and
Fan or Forshew teaches or suggests each and every limitation of claim 1, we
turn to the remaining challenged claims of the ’822 patent, which all directly
depend from claim 1 (i.e., claims 1–13 and 17–20). Patent Owner presents
separate arguments for dependent claims 2, 5, 9, 12, and 13. See PO Resp.
53–57. We address those claims individually below. See infra § IV.E.1.d–
h.
As to dependent claims 3, 4, 6–8, 10, 11, and 17–20, we find that
Petitioner also shows by a preponderance of the evidence that Schmitt,
Schmitt 2012, and Fan or Forshew account for the limitations in these
claims, or that certain limitations were well-known in the art as admitted in
the ’822 patent. Pet. 59–65, 67–70. We have also reviewed Dr. Gabriel’s
testimony and find that a preponderance of the evidence supports her
contention that the cited references collectively disclose or suggest each and
every limitation of claims 3, 4, 6–8, 10, 11, and 17–20. See Ex. 1002
¶¶ 162–163 (claim 3) (citing Ex. 1011, 3:1–6, 3:44–62, 7:58–62, 15:21–38);
id. ¶¶ 164–166 (claim 4) (citing Ex. 1011, 23:10–14; Ex. 1047, SI1); id.
¶¶ 168–169 (claim 6) (Ex. 1011, 4:25–29, 17:9–11, 29:57–30:10, Figure 3);
id. ¶ 170 (claim 7) (citing Ex. 1011, 21:10–19, 23:10–14); id. ¶¶ 171–172
(claim 8) (citing Ex. 1004, Abstract, 3, SI1, Table S1 (as evidenced by
Ex. 1081, 84, 87)); id. ¶¶ 174–176 (claim 10) (citing Ex. 1001, 32:16–19
(stating that “[m]ethods of reducing noise and/or distortion from a
sequencing process are known” and such methods “include, for example,
filtering sequences, e.g., requiring them to meet a quality threshold, or
IPR2019-00652
Patent 9,834,822 B2
43
reducing GC bias” (evidenced by Ex. 1079, 893; Ex. 1014, 446))); id. ¶ 177
(claim 11) (citing Ex. 1011, 9:1–13); id. ¶¶ 186–188 (claim 17) (citing
Ex. 1011, 20:23–29, 22:50–67); id. ¶¶ 189–190 (claim 18) (citing Ex. 1004,
1, 3, SI1, Table S1); id. ¶¶ 191–192 (claim 19) (citing Ex. 1011, 20:65–
21:19; Ex. 1047, SI2; Ex. 1001, 32:16–19 (as evidenced by Ex. 1079, 893;
Ex. 1014, 446)); id. ¶¶ 193–194 (claim 20) (citing Ex. 1011, 9:1–13).
Again, Patent Owner does not present separate and specific arguments
for any of these dependent claims. See PO Resp. 53–57 (presenting
arguments only as to dependent claims 2, 5, 9, 12, and 13). We, therefore,
adopt the teachings set forth in the Petition and in Dr. Gabriel’s Declaration
as mapped to the limitations of the challenged claims as our own findings.
See In re NuVasive, Inc., 841 F.3d 966, 974 (Fed. Cir. 2016) (explaining that
the Board need not make specific findings about claim limitations that a
patent owner does not dispute are disclosed in the prior art).
Dependent Claim 2
Claim 2 depends from claim 1 and recites “further comprising
detecting, at one or more loci, at least one single nucleotide variant, at least
one gene fusion and at least one copy number variant.” Ex. 1001, 62:49–51.
We agree with Petitioner that Schmitt teaches or suggests detecting each of
these types of genetic aberrations. Pet. 57–59.
Specifically, Schmitt teaches that the DCS method allows for the
confirmation of “the presence of a true mutation (as opposed to a PCR error
or other artifactual mutation) in a target nucleic acid sequence.” Ex. 1011,
16:43–47. Schmitt also teaches that the DCS method has the “ability to
indirectly infer that damage is present on the DNA,” which is a useful
biomarker for, e.g., cancer risk, cancer metabolic state, mutator phenotype
related to defective damage repair, carcinogen exposure, chronic
IPR2019-00652
Patent 9,834,822 B2
44
inflammation exposure, individual-specific aging, and neurodegenerative
diseases. Id. at 15:47–53. As an example, Schmitt teaches using DCS to
detect “single base substitutions” in a series of M13mp2 variants. Ex. 1011,
29:57–30:10; see also id. at 5:3–6 (stating that “DCS analysis removes . . .
strand bias and reveals the true mtDNA mutational spectrum to be
characterized by an excess of transitions”). As another example, Schmitt
teaches using DCS for the “accurate detection of altered genomic copy
number (e.g., for sensitive diagnosis of genetic conditions such as trisomy
21).” Ex. 1011, 18:3–21 (footnote omitted); see also Ex. 1002 ¶¶ 158–159.
Patent Owner argues that Schmitt does not expressly disclose
detecting genomic copy variation (as opposed to genomic copy number).
PO Sur-Reply 23–24. But we agree with Petitioner that an ordinarily skilled
artisan would have understood that Schmitt’s reference to “true mutations”
includes each type of genetic aberration recited in claim 2. Ex. 1002 ¶ 157;
see also Arendi S.A.R.L. v. Apple Inc., 832 F.3d 1355, 1361 (Fed. Cir. 2016)
(stating that we must “consider common sense, common wisdom, and
common knowledge in analyzing obviousness”).
Dependent Claim 5
Claim 5 depends from claim 1 and recites “wherein making the base
call comprises voting, averaging, maximum a posteriori or maximum
likelihood detection, dynamic programming, Bayesian methods, hidden
Markov methods or support vector machine methods.” Ex. 1001, 62:59–63.
As Petitioner points out, the ’822 patent states that “‘voting, averaging,
statistical, maximum a posteriori or maximum likelihood detection, dynamic
programming, Bayesian, hidden Markov or support vector machine
methods’ for generating consensus sequences were ‘known in the art.’”
Pet. 60 (quoting Ex. 1001, 45:64–46:5). Patent Owner argues that Petitioner
IPR2019-00652
Patent 9,834,822 B2
45
fails to sufficiently point to a teaching in Schmitt for these base-calling
methods. PO Resp. 564. But this argument merely attacks the references
individually and ignores what Patent Owner’s own written description
provides as common knowledge in the art. Thus, Patent Owner’s argument
is not persuasive. See Randall Mfg. v. Rea, 733 F.3d 1355, 1362 (Fed. Cir.
2013) (“rejecting a blinkered focus on individual documents” in an
obviousness analysis).
Dependent Claim 9
Claim 9 depends from claim 1 and recites “wherein mapping the
sequence reads comprises using information about a length of each of the
sequence reads.” Ex. 1001, 63:7–9. We agree with Petitioner that the
references teach or suggest mapping sequence reads using information about
the length of each sequence read. Pet. 64. Both Schmitt and Schmitt 2012
teach aligning (or mapping) sequence reads to a reference genome using the
Burrows-Wheeler aligner (BWA). Ex. 1011, 20:57–58; Ex. 1047, SI1.
Dr. Gabriel testifies that an ordinarily skilled artisan would have understood
that the BWA uses information about a length of each of the sequence reads
for mapping. Ex. 1002 ¶ 173. We find Dr. Gabriel’s testimony credible as
supported by preponderant record evidence. Specifically, Li and Durbin14
provide an overview of the BWA and teach that “[d]etermining the allowed
maximum number of differences” is based on “a read of length m.”
Ex. 1077, 1757. Patent Owner’s argument to the contrary lacks specificity
and is unpersuasive. See PO Resp. 55 (arguing that “[t]he petition does not

14 Heng Li and Richard Durbin, Fast and accurate short read
alignment with Burrows–Wheeler transform, 25(14) BIOINFORMATICS 1754–
1760 (2009) (Ex. 1077, “Li and Durbin”).
IPR2019-00652
Patent 9,834,822 B2
46
explain what aspect of the BWA algorithm that allegedly provides this
teaching”).
Dependent Claim 12
Claim 12 depends from claim 1 and recites “wherein the population of
polynucleotides is tagged with n different unique identifiers, wherein n is no
more than 100*z, wherein z is a mean of an expected number of duplicate
molecules having the same start and stop positions in the sample.”
Ex. 1001, 63:15–19. Petitioner contends that Fan and Forshew disclose
cfDNA amounts between 1 and 20 ng. Pet. 66 (citing Ex. 1048, 16270,
Table S1; Ex. 1004, Table S6). Relying on Dr. Gabriel’s testimony,
Petitioner contends that “[a] sample of 20 ng DNA contains about 6,060
haploid genome equivalents, which would have approximately 1–2 duplicate
polynucleotides beginning at any given position.” Id. (citing Ex. 1002
¶ 179). Petitioner contends that, “when there are 6,060 haploid genome
equivalents, z ≈ 1–2 and 100*z ≈ 10–200,” and thus “claim 12 requires
tagging with no more than 100–200 different unique identifiers.” Id. (citing
Ex. 1002 ¶ 179). Petitioner contends that Schmitt teaches “the use of a small
number of different tags,” such as 1-mer tags, and “when considering 1-mer
tags at both ends of a fragment, the number of different combinations is 16.”
Id. at 67 (citing Ex. 1011, 9:1–13; Ex. 1002 ¶ 180, n.12). “Accordingly,”
Petitioner contends, “Schmitt teaches the additional limitation of claim 12.”
Id. (citing Ex. 1002 ¶ 181).
Patent Owner argues that claim 12 is not unpatentable because
“[n]one of the cited references disclose[s] the expected number of duplicate
molecules in a sample much less a tagging scheme based on such a number.”
PO Resp. 55. Patent Owner argues that Petitioner’s analysis of claim 12 is
based on hindsight because Petitioner relies on the ’822 patent’s disclosure
IPR2019-00652
Patent 9,834,822 B2
47
of the empirical relationship between “z” and the number of tags, and that
Petitioner fails to provide any explanation “for selecting a 20 ng sample size
from the 1–20 ng range.” Id. at 55–56 (citing Pet. 66).
In its Reply, Petitioner contends that “[c]laim 12 is directed to a
number of identifiers for non-unique tagging, i.e., no more than 100*z,” and
that the combination of prior-art references teaches this number. Pet.
Reply 16.
Upon review of the parties’ respective arguments and evidence, we
determine that Patent Owner has the better position. As Patent Owner points
out and Petitioner does not dispute, the relationship between “z” and the
number of tags “n” in a sample was unknown prior to the ’822 patent. PO
Resp. 55–56; see also Ex. 1001, 41:6–13 (expressly defining “z” as the
“measure of central tendency (e.g., mean, median, mode) of an expected
number of duplicate molecules having the same start and stop positions”).
This is because the relationship was determined empirically and disclosed in
the ’822 patent. See Ex. 1001, 40:24–27 (“It has been empirically
determined that in a sample of about 10,000 haploid genome equivalents of
human DNA, there are about 3 duplicate polynucleotides beginning at any
given position.”). Petitioner has not explained adequately, nor provided
preponderant evidence, that the relationship between “100*z” and “n,” as
recited in claim 12, was known in the art. See In re Omeprazole Patent
Litigation, 536 F.3d 1361, 1379–80 (Fed. Cir. 2008) (holding that even
where a general method that could have been applied to make the claimed
product was known and within the level of skill of the ordinary artisan, the
claim may nevertheless be nonobvious if the problem which had suggested
use of the method had been previously unknown). For example, Petitioner
IPR2019-00652
Patent 9,834,822 B2
48
does not rely on an inherency theory for a teaching or suggestion of the
claimed subject matter in the prior art. Pet. 66–67; Pet. Reply 16.
We find that Petitioner’s arguments and evidence as to claim 12 are
conclusory and not persuasive. See Magnum Oil, 829 F.3d at 1380 (stating
that a petitioner cannot satisfy its burden of proving obviousness by
employing “mere conclusory statements”). In particular, Petitioner fails to
specifically address the previously unknown relationship between “n” and
“z” recited in the “wherein” clause of claim 12. Thus, we determine that
Petitioner fails to show by a preponderance of the evidence that the prior art
teaches or suggests the subject matter of claim 12.
Dependent Claim 13
Claim 13 depends from claim 1 and recites “wherein no more than
100 nanograms of polynucleotides from the bodily sample are converted in
b).” Ex. 1001, 63:20–22. We agree with Petitioner that both Fan and
Forshew explicitly disclose providing no more than 100 ng of cfDNA
molecules from the plasma of human subjects for sequencing. Pet. 67; see
also Ex. 1002 ¶¶ 182–185. Specifically, Forshew extracted between 0.9 ng
and 19.7 ng of DNA from cell-free plasma samples for sequence library
preparation. Ex. 1004, Table S6. Similarly, Fan extracted between 1.2 ng
and 8 ng of DNA from cell-free plasma for sequencing library construction.
Ex. 1048, 16270, Table S1. These amounts fall within “no more than 100
nanograms of polynucleotides from the bodily sample” as recited in
claim 13. Patent Owner argues that claim 13 “is not unpatentable” because
“Schmitt discloses using 3 μg of human colonic mucosa DNA.”
PO Resp. 56–57. Again, this argument is unpersuasive because it attacks the
references individually. See In re Merck & Co., 800 F.2d 1091, 1097 (Fed.
Cir. 1986) (attacking references individually is improper).
IPR2019-00652
Patent 9,834,822 B2
49
2. Motivation to Combine/Reasonable Expectation of Success
Even “[i]f all elements of the claims are found in a combination of
prior art references,” “the factfinder should further consider whether a
person of ordinary skill in the art would [have been] motivated to combine
those references, and whether in making that combination, a person of
ordinary skill would have [had] a reasonable expectation of success.” Merck
& Cie v. Gnosis S.P.A., 808 F.3d 829, 833 (Fed. Cir. 2015). The
“motivation to combine” and “reasonable expectation of success” factors are
subsidiary requirements for obviousness subsumed within the Graham
factors. Pfizer, Inc. v. Apotex, Inc., 480 F.3d 1348, 1361 (Fed. Cir. 2007).
We address motivation to combine and reasonable expectation of success in
turn below.
Motivation to Combine
Petitioner contends that an ordinarily skilled artisan, at the time of the
’822 patent, would have been prompted to combine “the teachings of
Schmitt and Schmitt 2012 regarding the DCS method” with “the teachings
of either Fan or Forshew regarding sequencing and analysis of cell free
DNA,” with a reasonable expectation of success, based on the teachings of
the art, the knowledge of an ordinarily skilled artisan, and the teachings of
the Schmitt references. Pet. 32–41. In our Institution Decision, we
determined that Petitioner had shown sufficiently for institution that an
ordinarily skilled artisan would have been motivated to use Schmitt’s DCS
method to detect low-frequency genetic mutations in the cfDNA samples
disclosed in Fan or Forshew. Inst. Dec. 24–25. Upon review of the
complete record and the parties’ respective arguments and evidence, we
affirm our initial determination.
IPR2019-00652
Patent 9,834,822 B2
50
Specifically, we find that a preponderance of record evidence supports
Petitioner’s argument that Schmitt “provides explicit motivation to
combine” its teachings with those of Fan, because Schmitt references Fan
when discussing the state of the prior art and limitations thereof. Pet. 32
(citing Ex. 1011, 1:28–55; Ex. 1002 ¶ 99). When discussing the prior art,
Schmitt states that “deep sequencing” technologies have “ushered in a new
era of genomic exploration,” because they “offer the unique ability to detect
minor variants within heterogeneous mixtures.” Ex. 1011, 1:28–37. Schmitt
states that those technologies have been implemented in a variety of fields
and that “[c]linical applications, such [as] prenatal screening for fetal
aneuploidy . . . are rapidly being developed.” Ex. 1011, 1:38–45. As an
example of prenatal screening, Schmitt directly cites to Fan. Id. at 1:42–43
(citing reference “[9]”). Schmitt then discusses the limitations of the prior-
art deep-sequencing techniques, such as limits on sensitivity and accuracy.
See id. at 1:56–2:55 (“Although, in theory, DNA subpopulations of any size
should be detectable when deep sequencing a sufficient number of
molecules, a practical limit of detection is imposed by errors introduced
during sample preparation and sequencing.”). Schmitt states that the method
disclosed in the ’822 patent is “an approach for tag-based error correction,
which reduces or eliminates artifactual mutations arising from DNA
damage, PCR errors, and sequencing errors.” Id. at 2:56–59. Schmitt states
that the disclosed method is an improvement over the prior art because it
“allows rare variants in heterogeneous populations to be detected with
unprecedented sensitivity.” Id. at 2:56–62.
We agree with Petitioner that an ordinarily skilled artisan would have
had a reason to combine the disclosures of the Schmitt references with Fan:
that is, to use Schmitt’s DCS method to accurately detect prenatal genetic
IPR2019-00652
Patent 9,834,822 B2
51
mutations in cfDNA as taught by Fan. Pet. 36–37. Schmitt specifically
refers to “prenatal screening for fetal aneuploidy” and references Fan in
discussing the prior art and limitations thereof. Ex. 1011, 1:42–43; see also
Pet. 36. An ordinarily skilled artisan screening for fetal aneuploidy, as
described by Fan, Ex. 1048, Abstract, 16266, would have been prompted to
look to Schmitt’s improved DCS method to “reduce[] or eliminate[]
artifactual mutations arising from DNA damage, PCR errors, and sequencing
errors.” Ex. 1011, at 2:56–59; see also Pet. 36–37; Ex. 1002 ¶¶ 103–108.
We further agree with Petitioner that an ordinarily skilled artisan
screening cfDNA for cancer mutations, as described by Forshew, Ex. 1004,
Abstract, 1–2, would have also looked to Schmitt’s DCS method for the
same reasons. Pet. 37; see also Ex. 1011, Abstract, 1:28–5; Ex. 1002
¶¶ 109–110. Specifically, in addition to discussing the use of deep-
sequencing techniques for prenatal screening for fetal aneuploidy, Schmitt
also refers to “early detection of cancer.” Ex. 1011, 1:42–45. Thus, we
agree with Petitioner that an ordinarily skilled artisan screening for cancer
mutations in cell-free plasma DNA, as described by Forshew, Ex. 1004,
Abstract, would have been prompted to look to Schmitt’s DCS method to
reduce amplification and sequencing errors and to improve detection of rare
genetic mutations. Ex. 1004, 1 (stating that “[s]ensitive methods for
detecting cancer mutations in plasma may find use in early detection
screening, prognosis, monitoring tumor dynamics over time, or detection of
minimal residual disease”); see also Pet. 36–37; Ex. 1002 ¶ 101.
We also find that the record supports, by a preponderance of the
evidence, Petitioner’s contentions about the knowledge the “ordinarily
skilled artisan would have brought to bear when considering combinations
or modifications” of the prior-art references. Randall Mfg. v. Rea, 733 F.3d
IPR2019-00652
Patent 9,834,822 B2
52
1355, 1362 (Fed. Cir. 2013); see Pet. 32–35. Specifically, the record
evidence shows that the ordinarily skilled artisan would have understood
that genetic mutations in cfDNA occur at low frequencies compared to wild-
type DNA and, thus, represent a small portion of the heterogeneous mixture
of DNA in the plasma. Ex. 1002 ¶ 100; Ex. 1051, Abstract (discussing the
low frequency of mutant adenomatous polyposis coli (APC) DNA molecules
in the plasma of cancer patients compared to normal APC); Ex. 1052, 92
(stating that “the coexistence in maternal plasma of a minor population of
fetal DNA within a major background of maternal DNA” poses challenges
for non-invasive prenatal diagnosis of fetal genetic and chromosomal
diseases); see also Pet. 32–33. Indeed, Fan expressly states that “measuring
aneuploidy remains challenging because of the high background of maternal
DNA; fetal DNA often constitutes <10% of total DNA in maternal cell-free
plasma.” Ex. 1048, 16266.
We agree with Petitioner that an ordinarily skilled artisan, wishing to
detect cfDNA genetic mutations as discussed in Fan or Forshew with high
accuracy and sensitivity, would have been motivated to use Schmitt’s
improved DCS method, because the DCS method “reduces or eliminates
artifactual mutations arising from DNA damage, PCR errors, and sequencing
errors,” and “allows rare variants in heterogeneous populations to be
detected with unprecedented sensitivity.” Ex. 1011, 2:56–60; see also KSR,
550 at 417 (“[I]f a technique has been used to improve one device, and a
person of ordinary skill in the art would recognize that it would improve
similar devices in the same way, using the technique is obvious unless its
actual application is beyond his or her skill.”).
We have carefully considered Patent Owner’s arguments and evidence
to the contrary, but remain persuaded that Petitioner has provided a
IPR2019-00652
Patent 9,834,822 B2
53
sufficient reason with rational underpinning for combining Schmitt with Fan
or Forshew. See PO Resp. 35–38. Citing to the testimonies of Dr. Shendure
and Dr. Quackenbush, Patent Owner argues that an ordinarily skilled artisan
would not have understood “the background section [of Schmitt] to be
contemplating application of DCS to fetal aneuploidy detection.”
PO Resp. 35 (citing Ex. 2025 ¶¶ 102–110; Ex. 2023 ¶ 20). We disagree. As
explained above, Schmitt states that deep-sequencing techniques have been
implemented in a variety of fields, including prenatal screening for fetal
aneuploidy and for early detection and monitoring of cancer. Ex. 1011,
1:28–55. Schmitt then discusses the limitations of the prior-art techniques,
such as limits on sensitivity and accuracy, and states that the method
disclosed is an improvement over the prior art because it “reduces or
eliminates artifactual mutations arising from DNA damage, PCR errors, and
sequencing errors,” and “allows rare variants in heterogeneous populations
to be detected with unprecedented sensitivity.” Id. at 1:56–2:62. Taken
together, we find that an ordinarily skilled artisan would have viewed these
teachings as an express suggestion to use the DCS method for the prior-art
applications described (e.g., prenatal and cancer-mutation screening).
Ex. 1104 ¶¶ 33–34.15

15 We are not persuaded otherwise by the testimonies of Dr. Shendure
and Dr. Quackenbush. Dr. Quackenbush states that “Schmitt is merely
disclosing that next-generation sequencing (‘NGS’) has been used by others
to detect fetal aneuploidy.” Ex. 2025 ¶ 103. Similarly, Dr. Shendure states
that “Schmitt is merely mentioning in some introductory remarks that
different techniques employing DNA sequencing have been used by others
to detect fetal aneuploidy.” Ex. 2023 ¶ 20. These statements fail to read the
entire background section of Schmitt as a whole, and thus lack specificity
and credibility. See Rohm & Haas Co. v. Brotech Corp., 127 F.3d 1089,
IPR2019-00652
Patent 9,834,822 B2
54
Relying on Dr. Quackenbush’s testimony, Patent Owner also argues
that “there would be no good reason to apply Schmitt’s DCS method to
detecting fetal aneuploidy,” because fetal aneuploidy is not a “rare mutation”
and using DCS to detect mutations that are not rare “increase[s] the expense
and complexity of the method.” PO Resp. 35–37 (citing Ex. 2025 ¶¶ 106–
107); PO Sur-Reply 10–11. As an initial matter, we note that, as of the
earliest priority date of the ’822 patent, ordinarily skilled artisans knew that
barcode-based deep-sequencing techniques had been used to screen cfDNA
for genetic defects such as aneuploidy successfully. Ex. 1002 ¶ 113;
Ex. 1058, 72–74 (describing the use of unique molecular identifiers to detect
trisomy 21); Ex. 1059, 1 (describing the use of barcode-based massively
parallel sequencing for the detection of aneuploidy). But even if Patent
Owner is correct that fetal aneuploidy does not qualify as a “rare mutation,”
the lack of economic feasibility due to increased complexity is not a
teaching away from using Schmitt’s DCS method for that purpose. See In re
Farrenkopf, 713 F.2d 714, 718 (Fed. Cir. 1983) (“That a given combination
would not be made by business[wo]men for economic reasons does not
mean that persons skilled in the art would not make the combination because
of some technological incompatibility. Only the latter fact would be
relevant.”).
Finally, Patent Owner argues that Schmitt, Fan, and Forshew are, at
most, analogous art and that their use of next-generation sequencing “is not
evidence that [an ordinarily skilled artisan] would have wanted to—or could
have—combined these methods.” PO Resp. 37–38 (citing Ex. 2025 ¶¶ 108–

1092 (Fed. Cir. 1997) (“Nothing in the rules or in our jurisprudence requires
the fact finder to credit the unsupported assertions of an expert witness.”).

IPR2019-00652
Patent 9,834,822 B2
55
110). We note that analogous art can, in some instances, demonstrate a
reason to combine references, such as where “[o]ne skilled in the art would
naturally look to prior art addressing the same problem as the invention at
hand, and . . . would find an appropriate solution.” In re ICON Health &
Fitness, Inc., 496 F.3d 1374, 1380 (Fed. Cir. 2007). Here, Schmitt teaches
that deep-sequencing techniques have been used “in a variety of fields,”
including in “prenatal screening for fetal aneuploidy” and in “early detection
of cancer.” Ex. 1011, 1:38–45. Schmitt teaches, however, that deep
sequencing “has limitations” in that it produces artifactual mutations
resulting from DNA damage, PCR errors, and sequencing errors. Id. at
1:56–2:62. Schmitt teaches that the disclosed DCS method (which includes
the “hybrid” embodiment discussed above) “reduces or eliminates artifactual
mutations” associated with the prior art. Id. An ordinarily skilled artisan
looking to screen cfDNA for prenatal or cancerous mutations with high
accuracy and sensitivity would have naturally looked to improvements in the
field of next-generation sequencing, such as Schmitt’s DCS method.
Ex. 1001 ¶¶ 99–111; Ex. 1104 ¶¶ 32–39. Thus, although the fact that the
prior-art references are analogous is not dispositive on the issue of
motivation to combine, we maintain that an ordinarily skilled artisan reading
Schmitt would have understood that Schmitt’s DCS method was a
reasonable substitute for the deep-sequencing and error-correction
techniques used in the prior art—such as the next-generation sequencing
techniques Fan and Forshew used for prenatal cfDNA screening and for
tumor cfDNA screening, respectively. Id.
In sum, we find that the facts here constitute a case of “the simple
substitution of one known element for another or the mere application of a
known technique to a piece of prior art ready for the improvement.” KSR,
IPR2019-00652
Patent 9,834,822 B2
56
550 U.S. at 417. Thus, for all the above reasons, we find that a
preponderance of the evidence supports Petitioner’s contention that an
ordinary artisan would have been motivated to combine the teachings of the
Schmitt references with those of Fan or Forshew.
Reasonable Expectation of Success
We next consider whether Petitioner has shown by a preponderance of
the evidence that an ordinarily skilled artisan would have had a reasonable
expectation of success. “The reasonable expectation of success requirement
refers to the likelihood of success in combining references to meet the
limitations of the claimed invention.” Intelligent Bio-Sys., Inc. v. Illumina
Cambridge Ltd., 821 F.3d 1359, 1367 (Fed. Cir. 2016).
In this case, the question before us is whether the ordinarily skilled
artisan would have had a reasonable expectation that Schmitt’s DCS method
could be used successfully with a population of cfDNA molecules.
Ex. 1001, 62:18–48 (claim 1). In making our findings as to “reasonable
expectation of success,” we keep in mind that we cannot demand absolute
certainty. See Medichem, S.A. v. Rolabo, S.L., 437 F.3d 1157, 1165 (Fed.
Cir. 2006) (“While the definition of ‘reasonable expectation’ is somewhat
vague, our case law makes clear that it does not require a certainty of
success.”); see also Pfizer, 480 F.3d at 1364 (“[C]ase law is clear that
obviousness cannot be avoided simply by a showing of some degree of
unpredictability in the art so long as there was a reasonable probability of
success.”).
Upon consideration of the entire record, we are persuaded by and
adopt Petitioner’s contentions (and Dr. Gabriel’s testimony) that an
ordinarily skilled artisan would have had a reasonable expectation of success
in using Schmitt’s DCS method to detect genetic mutations in cfDNA. See
IPR2019-00652
Patent 9,834,822 B2
57
Pet. 37–41; Ex. 1002 ¶¶ 112–120. In particular, we agree with Dr. Gabriel’s
testimony that all the necessary techniques for carrying out the claimed
method steps of the ’822 patent were known and required only ordinary skill
to perform. See Ex. 1002 ¶¶ 116–117. In this regard, the ’822 patent itself
teaches that the techniques needed to implement the claimed method were
well known and routine in the art. See Ex. 1001, 36:1–6 (stating that “cell
free polynucleotides may be isolated and extracted using a variety of
techniques known in the art” including by “commercially available kits”); id.
at 38:20–25 (stating that “assignment of unique or non-unique identifiers, or
molecular barcodes in reactions of this disclosure may follow methods and
systems described by, for example,” prior-art patents and patent
applications); id. at 38:41–48 (stating that “PCR for sequencing may be
performed using any means, including but not limited to use of commercial
kits provided by Nugen (WGA kit), Life Technologies, Affymetrix,
Promega, Qiagen and the like”); id. at 30:8–12 (stating that “subsequent
sequencing of cell free polynucleotides” may be done “by techniques known
in the art”); id. at 35:27–31 (as to mapping and aligning sequences, stating
that “sequences can be interrogated using the genome brow[s]er available”
online); id. at 45:64–46:5 (stating that “[c]onsensus sequences can be
generated from families of sequence reads by any method known in the
art”).
We also agree with Dr. Gabriel’s testimony that Schmitt itself
reasonably suggests that the DCS method can be used successfully for
several different types of applications by
explain[ing] that the DCS method does not require substantive
modification of existing sequencing workflows and can serve as
a general technique for next generation sequencing:
IPR2019-00652
Patent 9,834,822 B2
58
[T]he DCS approach can be generalized to nearly any
sequencing platform because a double-stranded SMI tag
can be incorporated into other existing adaptors, or for
sequencing approaches that do not require adaptors, a
double-stranded SMI tag can be ligated onto duplex DNA
sample prior to sequencing [… and therefore the]
compatibility of DCS with existing sequencing
workflows, the potential for greatly reducing the error rate
of DNA sequencing, and the multitude of applications for
the double-stranded SMI sequences validate DCS as a
technique that may play a general role in next generation
DNA sequencing.
Ex. 1002 ¶ 114 (quoting Ex. 1011, 18:44–61) (alteration in original).
In sum, “[t]his is not a situation where the prior art gave no direction
on how to reach a successful result.” In Re: Copaxone Consol. Cases, 906
F.3d 1013, 1026 (Fed. Cir. 2018). Instead, Schmitt suggests to the skilled
artisan to use the DCS method as a general technique for next generation
sequencing and error correction, and the ’822 patent evinces that the
technical details for isolating, barcoding, amplifying, sequencing, and
analyzing cfDNA molecules were known to be routine in the art. These
factors persuade us that an ordinarily skilled artisan would have had a
reasonable expectation of success.
We are not persuaded by Patent Owner’s arguments and evidence to
the contrary. See PO Resp. 38–53. We observe that many of Patent
Owner’s arguments are, in essence, that the prior art fails to provide proof
that using the DCS method to analyze cfDNA would have been successful.
See, e.g., PO Resp. 38–41 (characterizing Petitioner’s evidence of the
successful prior-art use of barcode-based sequencing and error correction for
cfDNA irrelevant because none “use[d] the Schmitt method,” and arguing
that “[e]ven if other methods using [next-generation sequencing] had been
IPR2019-00652
Patent 9,834,822 B2
59
applied to cfDNA, it does not establish that the different Schmitt method is
applicable to cfDNA”); id. at 41–42 (arguing that “Schmitt does not tag,
much less analyze fragments, similar to the naturally fragmented size of
cfDNA”); id. at 45 (arguing that “Schmitt applies the DCS method
exclusively to samples of cellular DNA” and “does not even mention
applying DCS to cfDNA, let alone provide any corresponding analysis”).
But the test for obviousness does not require “actual success” in the
prior art. See Endo Pharm. Inc. v. Actavis LLC, 922 F.3d 1365, 1379 (Fed.
Cir. 2019) (“As we have repeatedly held, obviousness requires only a
reasonable expectation of success, not proof of actual success.”). “[O]nly a
reasonable expectation of success, not a guarantee, is needed.” Pfizer, 480
F.3d 1364; see also Par Pharm., Inc. v. TWi Pharms., Inc., 773 F.3d 1186,
1198 (Fed. Cir. 2014) (“The reasonable expectation of success requirement
for obviousness does not necessitate an absolute certainty for success.”).
And here, we find that a preponderance of the evidence supports Petitioner’s
contention that an ordinarily skilled artisan would have had a reasonable,
even if not absolute, expectation that the DCS method could have been used
to analyze cfDNA successfully for the reasons discussed above.
Patent Owner also argues an ordinarily skilled artisan would not have
had a reasonable expectation of success in substituting genomic DNA with
cfDNA because the skilled artisan would have considered Schmitt’s DCS
method to be incompatible with cfDNA. See PO Resp. 41–53. In this
regard, Patent Owner argues that “there are several critical aspects of the
Schmitt method (e.g., random shearing, optimal fragment size, randomly
generated tags, numerous processing steps) that would be recognized as
being incompatible with cfDNA sample analysis.” Id. at 46 (citing Ex. 2023
¶¶ 21–22). We address these alleged incompatibilities individually below.
IPR2019-00652
Patent 9,834,822 B2
60
First, Patent Owner argues that DCS is incompatible with cfDNA
because “Schmitt describes the optimal range of fragment sizes for DCS is
200–500 bp”—a size that is greater than that of naturally occurring cfDNA
(i.e., about 140–170 bp). Id. at 41–42, 46–47. This argument is not
persuasive because nothing in Schmitt limits the applicability of DCS to
DNA fragments of approximately 200–500 bp in size. Instead, Schmitt
clearly teaches as part of the DCS method “size-selecting for appropriate
length fragments.” Ex. 1011, 3:53–56 (emphasis added). That the
appropriate length for fragments analyzed in Example 1 of Schmitt was 200–
500 bp does not teach an ordinarily skilled artisan away from the use of
other, differently sized, fragments for different needs. See Ex. 1011, 19:57–
20:15 (describing sequencing library preparation for Example 1); see also
Ex. 1104 ¶¶ 48–50.
Second, Patent Owner argues that Schmitt’s hybrid DCS method is
incompatible with cfDNA because that method relies on a random shearing
step to label the DNA. PO Resp. 42 (citing Ex. 1011, 9:9–13, 25:64–65,
17:41–45). Patent Owner also argues that Dr. Gabriel conceded that
shearing is incompatible with cfDNA. Id. at 50 (citing Ex. 2021, 158:4–12;
Ex. 2022, 105:11–20; Ex. 2023 ¶¶ 33–34). As discussed above, however,
Patent Owner’s arguments about shearing are not persuasive. Supra
§ IV.E.1.b. The record evidence amply supports Petitioner’s contention that
an ordinarily skilled artisan would have understood that cfDNA is already
fragmented by nature and, thus, would have simply omitted the shearing step
when tagging cfDNA with Schmitt’s short 4-mer tags. Ex 1002 ¶¶ 118, 143;
Ex. 1052, 94 (describing plasma DNA library preparation and stating that
“plasma DNA molecules [are] already fragmented in nature, so no additional
fragmentation step [is] required”); Ex. 1104 ¶ 46. Again, “[a] person of
IPR2019-00652
Patent 9,834,822 B2
61
ordinary skill is also a person of ordinary creativity, not an automaton.”
KSR, 550 U.S. at 421.16
Third, Patent Owner argues that Schmitt’s DCS method “requires
large quantities of input DNA that are simply not available when working
with cfDNA due to minute quantities that are reasonably obtained from
patients.” PO Resp. 46, 49 (citing Ex. 2023 ¶ 32). These large amounts of
DNA are necessary, Patent Owner argues, “[i]n view of inefficiencies in the
method and steps requiring discarding significant amounts of sample and
data.” Id. at 49. In particular, Patent Owner argues that “Schmitt instructs
additional steps that discard sample and data during processing and
analysis.” PO Resp. 47 (citing Ex. 1011, 20:42–43, 20:47–50, 20:53–55,
20:65, 21:4–18); see also id. at 47–49 (discussing SMI tags as “an additional
source of data loss). Patent Owner argues that “[t]he difference in the
amount of DNA between what is used in the Schmitt method and what is
practicable to obtain from a patient is not trivial” and “would require
fundamental modification and redesign of the DCS method.” Id. at 49–50.
Although we have carefully considered Patent Owner’s arguments and
evidence on this topic, we remain persuaded that an ordinarily skilled artisan

16 Moreover, although a subtle point, we disagree with Patent Owner
that an ordinarily skilled artisan would have understood the process of
shearing to be a critical aspect of the hybrid method. It is true that the
hybrid method uses a combination of sheared ends and a shorter n-mer tag to
create a “unique” molecular identifier. Ex. 1011, 9:9–13. But the lesson the
skilled artisan would have drawn from Schmitt is that the sequences at the
ends of the DNA molecule plus the non-unique sequences of the 4-mer
create a unique identifier. Id.; see also Ex. 1104 ¶ 47. Again, ligation of
adaptors and tags to DNA sequences (whether the DNA is blunt-ended or
contains overhanging molecules) was a routine, conventional activity in the
art as of the earliest priority date of the ’822 patent. See, e.g., Ex. 1011,
19:5–20:15; Ex. 1004, 16–34; Ex. 1048, 16270–271.
IPR2019-00652
Patent 9,834,822 B2
62
would have had a reasonable expectation of success. In this regard, the
“sample” or “data” to which Patent Owner refers relates to digital
information, not physical molecules of DNA. Ex. 1002 ¶¶ 174–176, 191–
192 (describing the process of filtering out sequence reads that do not meet
certain quality thresholds). Dr. Shendure conceded during cross-
examination that “the actual manipulations are related to the removal of data
rather than molecules.” Ex. 1102, 75:22–79:17. In addition, as discussed
above, Schmitt teaches that “the DCS approach can be generalized to nearly
any sequencing platform” and emphasizes the “compatibility of DCS with
existing sequencing workflows.” Ex. 1011, 18:44–61. Schmitt states that
the compatible sequencing platforms are, for example, “the Illumina
sequencing platform, ABI SOLiD sequencing platform, Pacific Biosciences
sequencing platform, 454 Life Sciences sequencing platform, Ion Torrent
sequencing platform, Helicos sequencing platform, and nanopore sequencing
technology.” Id. at 15:32–38. Petitioner points out that Fan used
approximately 1–8 ng of cfDNA in the Solexa/Illumina sequencing platform
with success, Pet. Reply 24 (citing Ex. 1048, 16270–71), thus suggesting
that any necessary modifications of these well-known sequencing platforms
would have been a routine activity for an ordinarily skilled artisan.
Finally, Patent Owner argues that several articles published after
Schmitt, including one written by the DCS inventors, show that the DCS
method “is inefficient and poorly suited for cfDNA samples.” PO Resp. 43–
45 (citing Ex. 2002; Ex. 2039; Ex. 2040; Ex. 2041; Ex. 1057). But as we
pointed out in the Institution Decision, Inst. Dec. 26, a certain tagging
efficacy is not recited in claim 1 of the ’822 patent, see Ex. 1001, 62:17–48
IPR2019-00652
Patent 9,834,822 B2
63
(claim 1).17 Patent Owner’s argument about efficacy, therefore, is not
persuasive. See Sound View Innovations, LLC v. Hulu, LLC, 2020 WL
3583556, at *3 (Fed. Cir. July 2, 2020) (holding that “[s]ince the claims here
did not require a certain level of practicality, the Board did not err in finding
a reasonable expectation of success in meeting the limitations of the claimed
invention” (quotation marks and alternations omitted)).
3. Objective Indicia of Non-Obviousness
We must consider any evidence of objective indicia of non-
obviousness before reaching our conclusion on obviousness vel non. WBIP,
LLC v. Kohler Co., 829 F.3d 1317, 1328 (Fed. Cir. 2016). Notwithstanding
what the teachings of the prior art would have suggested to one of ordinary
skill in the art at the time of the invention, the totality of the evidence
submitted, including objective evidence of non-obviousness, may lead to a
conclusion that the challenged claims would not have been obvious to one of
ordinary skill. In re Piasecki, 745 F.2d 1468, 1471–72 (Fed. Cir. 1984).
Patent Owner presents evidence of two of these considerations: (1) long-felt
need and (2) commercial success. PO Resp. 57–62; PO Sur-Reply 20–23.
Guardant360
Patent Owner offers the Guardant360 product as evidence for long-
felt need and commercial success of the claimed invention. PO Resp. 57.
According to Patent Owner, “Guardant360 is a ‘liquid biopsy’ assay that
provides clinically actionable sequence information from cfDNA molecules

17 In contrast, claim 1 of U.S. Patent 9,902,992 B2, which was at issue
in related proceedings IPR2019-00636 and IPR2019-00637, expressly
recites “tag[ging] at least 20% of the cfDNA molecules.” Institution in those
proceedings was denied. See IPR2019-00636, Paper 10; IPR2019-00637,
Paper 10.
IPR2019-00652
Patent 9,834,822 B2
64
obtained from a routine blood draw from cancer patients.” Id. at 58 (citing
Ex. 2034, 4). Patent Owner argues that Guardant360 embodies “at least
claim 1 of the ’822 patent,” including “convert[ing] cfDNA obtained from a
patient into non-uniquely tagged parent polynucleotides which are then
amplified and sequenced,” id. (citing Ex. 2029, 5, 17–18; Ex. 2030, 3539,
3540, 3542, 3544); mapping the resulting reads to a reference genome and
grouping into families “based on start and stop position and identifier
sequence,” id. (citing Ex. 2029, 19; Ex. 2030, 3540–3541); and collapsing
sequence reads at each genetic locus “to generate a base call” and
determining “the frequency of one or more bases at each locus,” id. (citing
Ex. 2029, 19; Ex. 2030, 3544).
Nexus
At the outset, we determine that we can give Patent Owner’s
arguments about long-felt need and commercial success no weight in our
obviousness analysis. “For objective evidence of secondary considerations
to be accorded substantial weight, its proponent must establish a nexus
between the evidence and the merits of the claimed invention.” In re Huai-
Hung Kao, 639 F.3d 1057, 1068 (Fed. Cir. 2011) (quotation marks and
emphasis omitted); see also Fox Factory, Inc. v. SRAM, LLC, 944 F.3d
1366, 1373 (Fed. Cir. 2019) (“In order to accord substantial weight to
secondary considerations in an obviousness analysis, the evidence of
secondary considerations must have a nexus to the claims, i.e., there must be
a legally and factually sufficient connection between the evidence and the
patented invention.” (quotation omitted)).
Patent Owner argues that we must presume a nexus. PO Resp. 57–58.
But a nexus is presumed only when “the patentee shows that the asserted
objective evidence is tied to a specific product and that product ‘embodies
IPR2019-00652
Patent 9,834,822 B2
65
the claimed features, and is coextensive with them.’” Fox Factory, 994 F.3d
at 1373 (quoting Polaris Indus., Inc. v. Arctic Cat, Inc., 882 F.3d 1056, 1072
(Fed. Cir. 2018)); see also WBIP, 829 F.3d at 1329 (setting forth
circumstances in which the presumption of nexus applies). And here, Patent
Owner has failed to show with specific and credible evidence that claim 1
and Guardant360 are “coextensive.”
For example, Patent Owner states that “Dr. Quackenbush explains
[that] the Guardant360 test embodies at least claim 1 of the ’822 patent,” but
provides no actual citations to, or discussion of, Dr. Quackenbush’s
testimony. PO Resp. 58. We decline to search through Dr. Quackenbush’s
Declarations for this evidence. See DeSilva v. DiLeonardi, 181 F.3d 865,
866–67 (7th Cir. 1999) (“A brief must make all arguments accessible to the
judges, rather than ask them to play archeologist with the record.”). Patent
Owner provides citations to Exhibit 202918 and Exhibit 203019, but these
documents appear to be research articles and do not provide any legal or
factual analysis of claim 1 in relationship to Guardant360. Thus, we
determine that Patent Owner’s arguments fall well short of a persuasive
showing of nexus.

18 Richard B. Lanman et al., Analytical and Clinical Validation of a
Digital Sequencing Panel for Quantitative, Highly Accurate Evaluation of
Cell-Free Circulating Tumor DNA, PLOS ONE. 1–27 (Oct. 16, 2015)
(Ex. 2029).

19 Justin I. Odegaard et al., Validation of a Plasma-Based
Comprehensive Cancer Genotyping Assay Utilizing Orthogonal Tissue- and
Plasma-Based Methodologies, 24(15) CLIN. CAN. RES. 1–27 (Aug. 1, 2018)
(Ex. 2030).
IPR2019-00652
Patent 9,834,822 B2
66
4. Conclusion as to Obviousness
In sum, we find that the combination of Schmitt, Schmitt 2012, and
Fan or Forshew teaches or suggests each and every element of claims 1–11,
13, and 17–20. We find that an ordinarily skilled artisan would have been
motivated to combine Schmitt, Schmitt 2012, and Fan or Forshew and
would have had a reasonable expectation of success in achieving the claimed
invention. We also find that Patent Owner has failed to persuasively show
secondary considerations of non-obviousness for failure to show
persuasively that the Guardant360 product is coextensive with claim 1. We
are therefore unable to accord Petitioner’s evidence any weight. Fox
Factory, 944 F.3d at 1373.
Thus, after carefully considering the arguments and evidence, we
determine that the record as a whole weighs in favor of a conclusion of
obviousness of claims 1–11, 13, and 17–20, especially given the disclosures
of the art of record in this case and strength of the obviousness case based on
the first three Graham factors. We find that Petitioner fails to prove by a
preponderance of the evidence that claim 12 would have been obvious.

IPR2019-00652
Patent 9,834,822 B2
67
V. CONCLUSION20
Petitioner establishes by a preponderance of the evidence that claims
1–11, 13, and 17–20 of the ’822 patent are unpatentable. Petitioner fails to
establish by a preponderance of the evidence that claim 12 of the ’822 patent
is unpatentable.
VI. ORDER
In consideration of the foregoing, it is hereby:
ORDERED that claims 1–11, 13, and 17–20 of the ’822 patent have
been proven to be unpatentable;
ORDERED that claim 12 of the ’822 patent has not been proven to be
unpatentable; and
FURTHER ORDERED that because this is a Final Written Decision,
parties to the proceeding seeking judicial review of the Decision must
comply with the notice and service requirements of 37 C.F.R. § 90.2.

20 Should Patent Owner wish to pursue amendment of the challenged
claims in a reissue or reexamination proceeding subsequent to the issuance
of this decision, we draw Patent Owner’s attention to the April 2019 Notice
Regarding Options for Amendments by Patent Owner Through Reissue or
Reexamination During a Pending AIA Trial Proceeding, 84 Fed. Reg.
16,654 (Apr. 22, 2019). If Patent Owner chooses to file a reissue application
or a request for reexamination of the challenged patent, we remind Patent
Owner of its continuing obligation to notify the Board of any such related
matters in updated mandatory notices. See 37 C.F.R. § 42.8(a)(3), (b)(2).
Claims

35 U.S.C.
§
Reference(s)/Basis Claims
Shown
Unpatentable
Claims
Not Shown
Unpatentable
1–13,
17–20
103(a) Schmitt, Schmitt
2012, and Fan or
Forshew
1–11, 13, 17–
20
12
Overall
Outcome
1–11, 13, 17–
20
12
IPR2019-00652
Patent 9,834,822 B2
68
For PETITIONER:
Rolando Medina
Eric Marandett
Stephanie Schoenwald
Sophie Wang
CHOATE, HALL & STEWART LLP
Rmedina@choate.com
Emarandett@choate.com
Sschonewald@choate.com
Swang@choate.com

For PATENT OWNER:
Michael Rosato
Steven Parmelee
Sonja Gerrard
WILSON, SONSINI, GOORDICH & ROSATI
Mrosato@wsgr.com
Sparmelee@wsgr.com
Sgerrard@wsgr.com