Julia Hoeng et al.

Patent Trials and Appeals Board

Sep 24, 2020

2020000630 (P.T.A.B. Sep. 24, 2020)

UNITED STATES PATENT AND TRADEMARK OFFICE
UNITED STATES DEPARTMENT OF COMMERCE
United States Patent and Trademark Office
Address: COMMISSIONER FOR PATENTS
P.O. Box 1450
Alexandria, Virginia 22313-1450
www.uspto.gov
APPLICATION NO. FILING DATE FIRST NAMED INVENTOR ATTORNEY DOCKET NO. CONFIRMATION NO.
14/124,826 03/03/2014 Julia Hoeng 4058.0020003 9814
155543 7590 09/24/2020
STERNE, KESSLER, GOLDSTEIN & FOX P.L.L.C.
1100 NEW YORK AVENUE, N.W.
WASHINGTON, DC 20005
EXAMINER
HARWARD, SOREN T
ART UNIT PAPER NUMBER
1631
NOTIFICATION DATE DELIVERY MODE
09/24/2020 ELECTRONIC
Please find below and/or attached an Office communication concerning this application or proceeding.
The time period for reply, if any, is set in the attached communication.
Notice of the Office communication was sent electronically on above-indicated "Notification Date" to the
following e-mail address(es):
e-office@sternekessler.com
PTOL-90A (Rev. 04/07)

UNITED STATES PATENT AND TRADEMARK OFFICE
_________________

BEFORE THE PATENT TRIAL AND APPEAL BOARD
_________________

Ex parte JULIA HOENG, FLORIAN MARTIN,
MANUEL CLAUDE PEITSCH, and ALAIN SEWER

_________________

Appeal 2020-000630
Application 14/124,826
Technology Center 1600
_________________

Before RICHARD M. LEBOVITZ, DEBORAH KATZ, and
JOHN A. EVANS, Administrative Patent Judges.

KATZ, Administrative Patent Judge.

DECISION ON APPEAL
Appellant1 seeks our review,2 under 35 U.S.C. § 134(a), of the
Examiner’s decision to reject claims 1, 4–23, and 26–28. We have
jurisdiction under 35 U.S.C. § 6(b). We AFFIRM.

1 We use the word “Appellant” as defined in 37 C.F.R. § 1.42. Appellant
identifies the real party-in-interest as Phillip Morris Products S.A. (Appeal
Br. 3.)
2 We consider the Final Office Action issued June 28, 2018 (“Final Act.”),
the Appeal Brief filed July 29, 2019 (“Appeal Br.”), the Examiner’s Answer
issued on September 4, 2019 (“Ans.”), the Reply Brief filed November 4,
Appeal 2020-000630
Application 14/124,826

2
Appellant’s Specification is directed to systems for studying the
mechanisms by which biological systems respond to introduced agents.
(Spec. ¶ 5.) Specifically, Appellant claims a computerized method for
quantifying the perturbation of a biological system in response to an agent
by making a model with nodes, representing biological entities, edges,
representing the relationships between these entities, and direction values,
representing the expected direction of change between control and treatment
data and generating a set of hypotheses.
Appellant’s claim 1 recites3:
A computerized method for quantifying a perturbation of a
biological system in response to an agent, comprising:
[a] receiving, at at least one processor, a set of treatment data
corresponding to a response of a biological system to an agent,
wherein the biological system includes or comprises a plurality of
biological entities, each biological entity interacting with at least one
other of the biological entities;
[b] receiving, at the at least one processor, a set of control data
corresponding to the biological system not exposed to the agent;
[c] providing, at the at least one processor, a network model
that represents the biological system and includes or comprises:
nodes representing the biological entities,
edges representing relationships between the biological entities,
and
direction values, for the nodes, representing the expected
direction of change between the control data and the treatment data;
[d] generating, with the at least one processor, a set of
mechanism hypotheses based on the network model, wherein each
mechanism hypothesis in the set of mechanism hypotheses comprises

2019 (“Reply Br.”), and the Specification as the publication of Appellant’s
International Application WO 2012/168483 A1.
3 Bracketed numbers to identify claim elements have been added for
reference.
Appeal 2020-000630
Application 14/124,826

3
a set of causal paths from an upstream node in the network model to a
set of downstream nodes in the network model;
[e] calculating, with the at least one processor, activity
measures for the set of downstream nodes of each mechanism
hypothesis, wherein an activity measure represents a difference
between the treatment data and the control data for each downstream
node;
[f] calculating, with the at least one processor, weight values for
the set of downstream nodes of each mechanism hypothesis, wherein
at least one weight value is different from at least one other weight
value, and wherein the calculating the weight values comprises
calculating a probability that the activity measures represent a
departure from a null hypothesis of a zero difference;
[g] generating, with the at least one processor, a scorable
network model based on the set of mechanism hypotheses; and
[h] generating, with the at least one processor, a score for the
scorable network model representative of the perturbation of the
biological system in response to the agent, wherein the perturbation of
the biological system in response to the agent is unable to be
measured directly, and the score is based on the direction values, the
weight values, and the activity measures.

(Appeal Br. 41–42.) Appellant’s claimed method results in a score for the
network that represents perturbation of the system in response to an agent.
The Examiner rejects all of Appellant’s pending claims as failing to
comply with the written description requirement of 35 U.S.C. § 112, first
paragraph. (See Final Act. 6–7.) The Examiner also rejects all of
Appellant’s pending claims under 35 U.S.C. § 101 because the claimed
inventions are directed to non-statutory subject matter. (See id. at 7–9; see
Ans. 4–7.) In addition, the Examiner makes rejections of Appellant’s claims
as being obvious under 35 U.S.C. § 103(a). (See Final Act. 11–21; see Ans.
7–15.) A rejection of claims 1–23 and 26–28 under 35 U.S.C. § 112, second
Appeal 2020-000630
Application 14/124,826

4
paragraph, was withdrawn in view of Appellants arguments in the Appeal
Brief. (See Ans. 4.)
Appellant does not provide separate arguments for any of the rejected
claims. Accordingly, we focus on claim 1 in our analysis.

35 U.S.C. § 112, first paragraph – written description
The Examiner rejects all of Appellant’s pending claims as lacking
written description support, under 35 U.S.C. § 112, first paragraph, asserting
that the claim phrase “the perturbation of the biological system in response
to the agent is unable to be measured directly” when “generating ... a score
for the computational causal network model representative of the
perturbation of the biological system” recited in independent claims 1, 23,
and 28 are not supported in the Specification. (See Ans. 4.) According to
the Examiner, paragraph 50 of the Specification describes a situation in
which “it is not necessary for the computerized method to receive data for all
such measurable nodes,” but does not describe a situation in which
perturbation cannot be measured directly. (See id.)
Appellant persuades us that the Specification sufficiently describes
generating a score for a network model wherein “the perturbation of the
biological system in response to the agent is unable to be measured directly.”
(See Appeal Br. 14–18.) Appellant argues that the purpose of the claimed
methods is to quantify the perturbation of a biological system in response to
an agent, instead of measuring the perturbation itself. (Id. at 15–16, citing
Spec. ¶¶ 4–5.) Appellant explains that the claimed method involves “the
traversal of the causal networks” causing downstream effects to generate a
Appeal 2020-000630
Application 14/124,826

5
cumulative score, wherein the actual cause of the downstream effect cannot
be measured. (See Appeal Br. 15.)
We find that Appellant’s Specification supports Appellant’s argument.
Appellant’s Specification explains that
[t]he nodes in the graph can also represent relationships
between nodes. Thus, it is possible to represent relationships
between relationships, or relationships between a relationship
and another type of biological entity represented in the graph.
For example a relationship between two nodes that represent
chemicals may represent a reaction. This reaction may be a
node in a relationship between the reaction and a chemical that
inhibits the reaction.

(Spec. ¶ 50; see Appeal Br. 17.) This portion of the Specification describes
interrelationships that go beyond direct measurement of a cause and an
effect. Appellant’s Specification expressly describes “biological entities
[that] are not necessarily limited to those biological entities for which
treatment or control data are received or available” as being nodes of the
claimed method. (Spec. ¶ 50.)
Furthermore, Appellant’s Specification provides an example of the
results of analyzing TNF treatment of NHBE cells to cause downstream
gene activation by NF-κB and downstream gene activation, where TNF does
not directly mediate transcription by NF-κB. (See Spec. ¶¶ 99–105; see
Appeal Br. 15.) We are persuaded by Appellant’s argument that this
example demonstrates effects on an upstream biological entity where only
raw RNA expression resulting from the perturbation, and not the
perturbation itself, was measured directly. Thus, we are persuaded that the
example provides a written description of generation of a score “wherein the
perturbation of the biological system in response to the agent is unable to be
Appeal 2020-000630
Application 14/124,826

6
measured directly, and the score is based on the direction values, the weight
values, and the activity measures,” as recited in claim 1.
Accordingly, we reverse the Examiner’s rejection under 35 U.S.C.
§ 112, first paragraph.

35 U.S.C. § 103
The Examiner rejects Appellant’s claims as being obvious under 35
U.S.C. § 103(a). Specifically, the Examiner rejects claims 1, 4–13, 15–21,
23, and 28 over Ladd,4 Toyoshiba,5 and Strimmer6 (see Final Act. 11–18);
claim 14 over Ladd, Toyoshiba, Strimmer, and Löfroth7 (see id. at 18–19);
claims 14, 26, and 27 over Ladd, Toyoshiba, Strimmer, and Sexton8 (see id.
at 19–20); and claim 22 over Ladd, Toyoshiba, Strimmer, and Friedman9
(see id. at 20–21). Appellant does not raise any separate arguments against
any of the claims or groups of claims in the separate rejections. (See Reply
Br. 40.) Thus, we focus on claim 1. See 37 C.F.R. § 41.37(c)(1)(iv).

4 Ladd and Elliston, U.S. Patent Application Publication 2009/0099784 A1,
published April 16, 2009.
5 Toyoshiba et al., “Gene Interaction Network Suggests Dioxin Induces a
Significant Linkage between Aryl Hydrocarbon Receptor and Retinoic Acid
Receptor Beta,” Environmental Health Perspectives, 112:1217–24 (2004).
6 Strimmer, “A unified approach to false discovery rate estimation,” BMC
BIOINFOrmatics 9:303 (2008).
7 Löfroth and Rannug, “Ah receptor ligands in tobacco smoke,” Toxicology
Letters, 42: 131–36 (1988).
8 Sexton et al., “Genomic biomarkers of pulmonary exposure to tobacco
smoke components,” Pharmacogenetics and Genomics, 18:853–60 (2008).
9 Friedman et al., “Data Analysis with Bayesian Networks: A Bootstrap
Approach,” Uncertaintly in Artificial Intelligence 196–215 (1999).
Appeal 2020-000630
Application 14/124,826

7
Ladd teaches software assisted methods for identifying similarities
and differences between biological states in a causal system model (“CSM”)
using nodes to represent the differences between two biological states and
links between the nodes to indicate a causal directionality between the
nodes. (See Ladd ¶ 25.) Ladd teaches that a researcher can conduct a series
of experiments involving perturbations to the system to see which
perturbations result in that outcome. The data from these experiments is
mapped to a model that sums all of the upstream or downstream causal
hypotheses explaining the outcome. (See id. ¶¶ 110, 111.) Ladd teaches,
further, a method of “pruning” hypotheses by applying logic based criteria to
each member of the set of models to reject paths or portions not likely to be
representative of real biology and leaving a smaller number of remaining
models to constitute new active causative relationships. (See id. ¶ 112.)
The Examiner’s rejection of Appellant’s claim 1 is based on the
finding that Ladd teaches steps [a] through [e], [g], and [h] of claim 1,
including receiving biochemical data, providing a network model with nodes
and mapped relationships between the nodes, generating a set of mechanism
hypotheses, calculating activity measures, and generating a scorable network
model and a score for the model, which correspond to each step of the
recited method, except step [f]. (See Ans. 7–9.)
The Examiner finds that Toyoshiba teaches an example of the ability
to test hypotheses, for example to assess the health risks of compounds that
affect genetic regulatory networks. (See Ans. 10, citing Toyoshiba 1217,
1223.) The Examiner finds further that Toyoshiba teaches a log-linear
expression model that can be used to mathematically compare different
models reflecting different mechanistic hypotheses (e.g., CSM). (See Ans.
Appeal 2020-000630
Application 14/124,826

8
10; see Toyoshiba 1218.) The Examiner bases the rejection of Appellant’s
claims on the finding that, whereas Ladd teaches using CSM to simulate
biological behavior and then uses biological measurements and simulated
behavior to evaluate the resulting mechanistic hypotheses, Toyoshiba
teaches a mathematical procedure that can be used to compare different
CSMs. (See Ans. 10; see Toyoshiba 1218 (“One of the simplest types of
weighting function used to describe a gene expression network is the log-
linear weighting function given by the following form . . . .”).)
The Examiner finds further that Strimmer teaches a method of
estimating the false (non-)discovery rate (“fndr”) for a parameter and
selecting a cutoff based on the fndr. (See Ans. 11, citing Strimmer 9
(“Selection of suitable truncation point using the false non-discovery rate”).)
The Examiner also finds that Strimmer teaches that “false discovery rate
analysis is a key statistical innovation that has found widespread application
in the study of high-dimensional data.” (Ans. 11, citing Strimmer 13.) The
Examiner points to the disclosure in Appellant’s Specification:
One value that may be advantageously used for weighting is the
local false non-discovery rate fndri (i.e., the probability that a
fold-change value βi represents a departure from the underlying
null hypothesis of a zero fold-change, in some cases,
conditionally on the observed p-value) as described by
Strimmer et al. in “A general modular framework for gene set
enrichment analysis,” BMC Bioinformatics 10:47, 2009 and by
Strimmer in “A unified approach to false discovery rate
estimation,” BMC Bioinformatics 9:303, 2008 [cited by the
Examiner], each of which is incorporated by reference herein in
its entirety.

(Spec. ¶ 72; see Ans. 11 (emphasis added).) Thus, According to the
Examiner, the procedure of Strimmer distinguishes values that represent
Appeal 2020-000630
Application 14/124,826

9
statistically significant differences from zero, from values that represent
statistically insignificant differences from zero and therefore provides a
more rigorous way of identifying statistically insignificant β parameters in
the model of Toyoshiba. (See Ans. 11.)
The Examiner finds that at the time of invention one of ordinary skill
in the art would have found it obvious to have combined the teachings of
Ladd and Toyoshiba because Toyoshiba teaches specific details of a
mathematical procedure to evaluate mechanistic hypotheses based on CSMs.
(See Ans. 13–14.) The Examiner finds further that a practitioner would have
been motivated to modify the methods of Ladd and Toyoshiba to identify the
βis describing gene interactions as “substantial” or “insubstantial” using
cutoff value for a parameter based on a local FNDR, as taught by Strimmer,
because Strimmer teaches a more robust procedure for discriminating false
positives and negatives that Toyoshiba. (See id. at 14.)
Appellant argues that the Examiner fails to articulate why Toyoshiba
would be combined with Ladd. (See Appeal Br. 31–32.) According to
Appellant, Ladd teaches a method that compares one causal system to
another by searching for branch points and pruning those that are not likely
representative of real biology. (See id. at 31, citing Ladd ¶ 21.) Appellant
argues further that Ladd teaches “simulation tools [that] are used to probe
the assembly” and teaches “suitable tools” as described in patent application
10/992,973 (“Chandra”). (See Appeal Br. 31, citing Ladd ¶ 103.)
According to Appellant, because Ladd refers to suitable simulation tools,
there is no apparent reason why the mathematical procedure of Toyoshiba
should be used for performing the simulations of Ladd. (See Appeal Br. 31.)
Appellant argues that Chandra, to which Ladd refers, discusses using a
Appeal 2020-000630
Application 14/124,826

10
mathematical procedure for simulations. (See Reply Br. 31, citing Chandra
14:66–15:21.)
We are not persuaded by Appellant’s argument because even if Ladd
refers to a way to simulate biological networks quantitatively, other means
of doing so may still have been obvious. As explained below, Appellant is
mistaken that the Examiner erred by not giving a reason “why the use of
such a model would be preferable over the simulation tools that are already
referenced in Ladd.” (See Appeal Br. 31; see also Reply Br. 35 (“the
Examiner provides no rationale as to why the logical simulation of
Toyoshiba, which is directed to a specific application of TCDD needs to be
applied instead of the logic criteria of Change to Ladd for hypothesis
prunning.”).)
We are persuaded there would have been a reason to use the method
of Toyoshiba because the Examiner finds that Ladd can use simulations
tools other than those taught in Chandra and finds that the quantitative
simulation of Toyoshiba provides more detail about the characteristics of a
biological network, using a Bayesian approach to parameterizing that is
advantageous by giving an estimate of the uncertainty in the quantitative
estimates. (See Ans. 23, 26.) Appellant argues that the Examiner does not
show that an estimate of uncertainty is quantitative estimates is “uniquely
advantageous when provided by a Bayesian approach.” (Reply Br. 36.) But
the method of Toyoshiba need not be preferred or the most desirable for
there to have been a reason to combine the references and render Appellant’s
claimed method obvious. A particular combination of prior art need not be
preferred or the most desirable in order to provide motivation for the current
invention, rather the question is whether there is something to suggest the
Appeal 2020-000630
Application 14/124,826

11
desirability of making the combination. See In re Fulton, 391 F.3d 1195,
1200 (Fed. Cir. 2004) (rejecting Appellant’s argument that because
hexagonal patterns on shoe soles were not found to be preferred over other
alternatives disclosed in the prior art, Appellant’s claims were not obvious);
see also Novartis Pharm. Corp. v. W.-Ward Pharm. Int’l Ltd., 923 F.3d
1051, 1059 (Fed. Cir. 2019) (“After reviewing the prior art, the district court
found that a person of ordinary skill ‘would have been motivated to pursue
everolimus as one of several potential treatment options for advanced solid
tumors, including advanced RCC.’ This finding should have affirmatively
answered whether there would have been a motivation to combine. . . . It is
thus improper to require West-Ward to prove that a person of ordinary skill
would have selected everolimus over other prior art treatment methods.”)
(internal citations omitted).
Appellant does not argue that the method of Toyoshiba would not
have any advantages, rather, Appellant argues that it would not be preferred
over the disclosure of Chandra.
In many fields it may be that there is little discussion of obvious
techniques or combinations, and it often may be the case that
market demand, rather than scientific literature, will drive
design trends. Granting patent protection to advances that
would occur in the ordinary course without real innovation
retards progress and may, in the case of patents combining
previously known elements, deprive prior inventions of their
value or utility.

KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 419 (2007).
Appellant argues further that modification of the method of Ladd with
the method of Toyoshiba would render Ladd inoperable. (See Appeal Br.
32–33; Reply Br. 32–33.) Appellant argues that Toyoshiba is specifically
Appeal 2020-000630
Application 14/124,826

12
directed to a log-linear network for the response of one type of cell to one
agent without any mention of assembling a simulation involving the
“increases or decreases in the quantity or activity of nodes within the
assembly . . . result[ing] in generation of a large number of branching paths”
as required in Ladd. (See Appeal Br. 33, quoting Ladd ¶¶ 76, 103.)
Appellant argues that it is not clear how the log-linear network for “defining
a Bayesian network to derive the posterior distribution for the parameters of
interest” would generate the large number of branching patterns or how
these branches would be pruned or scored as taught in Ladd. (See Appeal
Br. 33, citing Ladd ¶ 133.)
We are persuaded by the Examiner’s explanation that a log-linear
network, as in the simulation of Toyoshiba, does not refer to any particular
network structure and imposes no limits on the structure of the network.
The Examiner finds that a log-linear network can have as many or as few
branches as desired and that scoring is based on the result of the simulation
mathematics. Thus, according to the Examiner, the details of the simulation
have no effect on the scoring. (See Ans. 24–25.)
Appellant argues that the Examiner is wrong about the effects on
scoring because the score directly informs the pruning of hypotheses in
Ladd, which in turn informs the score of Ladd. (See Reply Br. 35–36.)
Although we understand that the ultimate score is directly impacted by the
mathematical details of the simulation, because Appellant does not claim a
method that produces a specific score, we are not persuaded that different
mathematical details cannot be used in the method of Ladd and still fall
within the claimed method. That is, Appellant does not direct us to evidence
that the claimed method is limited to a particular simulation or mathematical
Appeal 2020-000630
Application 14/124,826

13
details to produce a certain result. Appellant’s argument does not persuade
us that the mathematical details taught in Toyoshiba are not encompassed by
the claimed method.
Appellant argues that the combination of the Ladd and Toyoshiba
would fail to operate as recited in the claimed method because the
combination does not provide a step of calculating activity and weight
values for the downstream nodes of each mechanism hypothesis and
subsequently generating a score for the network model based on direction
values of an expected direction of change, weight values, and activity
measures as required in claim 1. (See Appeal Br. 35–38.)
Appellant argues further that the Examiner does not consider the
claimed method as a whole because the Examiner “disregards that the
scoring of the model in Ladd is based on the simulation performed, and the
subsequent pruning that occurs.” (Appeal Br. 34; see Reply Br. 34.)
Appellant focuses first on the claim limitation “providing direction values,
for the nodes, representing the expected direction of change between the
control data and the treatment data.” (See Appeal Br. 35–36.) According to
Appellant, the teaching in Ladd of a causal system model that includes nodes
representing differences in a first biological state and a second biological
state does not include a teaching of an “expected direction of change.” (See
id.) Appellant argues that such expected directions are not taught in
Toyoshiba or Strimmer either. (See id. at 36, 38.)
The Examiner refutes Appellant’s argument, explaining that
Toyoshiba teaches a factor, Iji, to indicate stimulation, inhibition, or no
change, thus indicating whether two nodes are expected to change in the
same direction (stimulation) or different directions (inhibition), as described
Appeal 2020-000630
Application 14/124,826

14
in Appellant’s Specification. (See Ans. 25, citing Toyoshiba 1218 (“Iji is an
indicator variable describing the direction of the change denoted by βji,
where IJi; = l for stimulation, IJi; = - l for inhibition, and IJi; = 0 for no
effect.”); Spec. ¶ 5.) Appellant argues that “the provision of direction
values, as well as the inclusion of said provided direction values in such a
score generation step, has not been shown in any of Ladd, Toyoshiba, or
Strimmer.” (Reply Br. 37–38.) The teaching of Toyoshiba provides
direction values, thus, we are persuaded that determining values for nodes,
representing an expected direction of change, when developing a network
model was known in the art.
Appellant argues further that the Examiner has not shown how
calculating the activity measures and weight values for the set of
downstream nodes of each mechanism hypothesis are met by the prior art.
(See Appeal Br. 36–37.) This argument reiterates Appellant’s argument that
there would have been no motivation to use the mathematical procedures of
Toyoshiba with the method of Ladd because Ladd refers to Chandra for
simulation tools. (See id.) As explained above, we are not persuaded by this
argument.
Appellant argues further that the Examiner inappropriately equates the
teaching in Strimmer of a false non-discovery rate fndr with the claim step
of “calculating the weight values comprises calculating a probability that the
activity measures represent a departure from the null hypothesis of a zero
difference.” (See Appeal Br. 37.) Appellant acknowledges that Strimmer
teaches the fndr probability may be useful for weighting, but argues that
there is no weighting taught in Toyoshiba to which Strimmer may be
applied. (See id.) In particular, Appellant argues that Strimmer teaches that
Appeal 2020-000630
Application 14/124,826

15
fndr probability is conditioned on the observed p-value, but the p-values are
not calculated in the methods of Toyoshiba. (See id.) Appellant argues
further that Strimmer does not teach a calculation of weight values and that
the Examiner does not show why weight values should be part of the overall
network score. (See id. at 38.)
The Examiner responds that Toyoshiba teaches computation of weight
values in its score simulation by teaching inclusion of weight parameters βji
when scoring a network model by computing the likelihood of the model.
(See Ans. 26.) The Examiner explains that Strimmer teaches a specific
procedure for determining when the values of particular βji parameters are
statistically significant and for zeroing statistically insignificant weight
parameters. (See id. at 26–27.) Thus, we are not persuaded that the
Examiner erred in combining Strimmer with Toyoshiba, and Ladd, to teach
computation of weighting values.
Appellant also argues that the Examiner has not shown that the prior
art teaches the claim step of generating a score for the network based on
direction values, weight values, and activity measures, as recited in claim 1.
(See Appeal Br. 38.) Appellant reiterates the argument that the prior art fails
to teach including direction values in a score and argues further that
Strimmer does not teach calculating weight values or why they should be
included as part of an overall network score. (See id.)
Appellant argues there is no teaching in Toyoshiba of calculating a
network score, rather Toyoshiba teaches only calculating a conditional
likelihood of data in the form of an equation for deriving the posterior
distribution of a Bayesian network. (See Appeal Br. 38.) The Examiner
explains that the likelihood function of Toyoshiba incorporates an observed
Appeal 2020-000630
Application 14/124,826

16
activity measure (χj), weight value (βij), and direction value (Iij) in
calculating the conditional density, which is used to determine the likelihood
of a data point or an entire dataset. (See Ans. 27, citing Toyoshiba 1218–
19.) Toyoshiba supports the Examiner’s findings by providing that χj is the
“observed level of expression,” βij is the “magnitude by which a change in
one log unit of gene Xj will affect the level of expression of gene Xij,” and Iji
is “an indicator variable describing the direction of the change denoted by
βij.” (Toyoshiba 1218.)
Appellant refutes this explanation, arguing that Iij, βij, and χj are not
equivalent to the direction value, weight value, and activity measure recited
in claim 1 because there “is not a necessary correspondence of an observed
level of expression of a gene Xj with an activity measure as claimed”
because claim 1 requires the activity measure “represents a difference
between the treatment data and the control data for each downstream node.”
(Reply Br. 39 (emphasis added).) Appellant points to the Specification,
which provides that the activity measure may include “a logarithm of the
difference between the treatment data and the control data.” (See Reply Br.
39, quoting Spec. ¶ 5.) Appellant argues further that
the magnitude by which a change in one log unit of gene Xj will
affect the level of expression of gene Xi is not necessarily a
weight value, where in fact this mode of calculation of weight
values as asserted by the examiner does not “comprise
calculating a probability that the activity measures [for a
downstream node] represents a departure from a null hypothesis
of a zero difference” as required by claim 1, but rather
represents the magnitude by which the change in level of one
gene affects the level of expression of another gene.

Appeal 2020-000630
Application 14/124,826

17
(Reply Br. 39–40 (emphasis added).) These arguments are not persuasive
because although Appellant argues that the Examiner’s findings regarding Iij,
βij, and χj may not necessarily be the elements of claim 1, Appellant fails to
direct us to evidence they are not. Instead, Toyoshiba seems to support the
Examiner’s findings that the network could be scored as taught in Ladd
“based on multiple criteria indicative of how close a given
hypothesis/branching path approaches explanation of the operational data.”
(Ans. 24, quoting Ladd ¶ 23.) We agree with the Examiner that the
likelihood function of Toyoshiba is a score with the characteristics needed
by the method of Ladd because it is a statistical measure “indicative of how
close a given hypothesis/branching path [i.e. genetic network model]
approaches explanation of the operational data.” (Ans. 25.)
Appellant does not persuade us that the Examiner erred in rejecting
claim 1 as being obvious. Appellant fails to present separate arguments for
the rejection of any other of the pending claims. Accordingly, we sustain
each of the rejections under 35 U.S.C. § 103.
35 U.S.C. § 101
The Examiner rejects Appellant’s claim 1 as being drawn to ineligible
subject matter under 35 U.S.C. § 101, finding that the claims do not recite
significantly more than the abstract idea of “quantifying the perturbation of a
biological system in response to an agent” on a generic computer and that
this abstract idea is not integrated into a practical application. (See Ans. 4–
7.)
Although 35 U.S.C. § 101 provides that “[w]hoever invents or
discovers any new and useful process, machine, manufacture, or
composition of matter, or any new and useful improvement thereof, may
Appeal 2020-000630
Application 14/124,826

18
obtain a patent therefor . . . ,” the Supreme Court has determined that there
are exceptions to what is patentable. Specifically, “laws of nature, natural
phenomena, and abstract ideas” are not eligible subject matter. See
Diamond v. Diehr, 450 U.S. 175, 185 (1981). To determine if claimed
subject matter is statutorily eligible in light of these judicial exceptions the
Supreme Court has articulated a two-step framework in Mayo Collaborative
Services v. Prometheus Laboratories, Inc., 566 U.S. 66 (2012), and later
cases. Specifically,
[f]irst, we determine whether the claims at issue are directed to
one of those patent-ineligible concepts. If so, we then ask,
“[w]hat else is there in the claims before us?” To answer that
question, we consider the elements of each claim both
individually and “as an ordered combination” to determine
whether the additional elements “transform the nature of the
claim” into a patent-eligible application.

Alice Corp. v. CLS Bank Int’l, 573 U.S. 208, 217 (2014) (quoting
Mayo, 566 U.S. at 78) (internal citations omitted). Thus, we must determine
whether the claim is directed to a judicially determined patent-ineligible
concept and, if so, then ask if there is anything in the claim that transforms it
into patent-eligible subject matter.
The USPTO issued revised guidance on the application of § 101. See
2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. 50–
57 (2019) (“2019 Guidelines”). After determining that claimed subject
matter falls within one of the four categories of patentable subject matter
identified in 35 U.S.C. § 101, the 2019 Guidelines provides a “revised step
2A.” which corresponds to the first step of the Alice/Mayo test articulated
above, to determine whether a claim is directed to a judicial exception. (See
Appeal 2020-000630
Application 14/124,826

19
2019 Guidelines, 84 Fed. Reg. 50, 53–54.) In a first prong of revised step
2A, the Examiner must determine whether the claim recites a judicial
exception. (See id. at 54.) If a judicial exception is identified, the second
prong requires a determination of whether the judicial exception is
integrated into a practical application. (See id.) If so, the inquiry ends and
the claim is determined to be directed to eligible subject matter under the
2019 Guidelines. (See id. at 54 (“When the exception is so integrated [into a
practical application], then the claim is not directed to a judicial exception
(Step 2A: NO) and is eligible. This concludes the eligibility analysis.”).) If
not, the analysis continues to determine if the claim provides an inventive
concept. (See id. at 56.)
The 2019 Guidelines provide that mental processes are a type of
judicial exception. (See 2019 Guidelines, 84 Fed. Reg. at 52.) Mental
processes are concepts performed in the human mind, including observation,
evaluation, judgment, and opinion. (See id.) The Examiner finds that the
claim steps of “receiving . . . a set of treatment data,” “receiving . . . a set of
control data,” and “generating . . . a set of mechanism hypotheses . . .” are
mental processes recited in Appellant’s claim 1. (See Ans. 5.) We agree
with the Examiner that receiving data, generating a set of hypotheses, and
scoring the hypotheses are processes that can be performed by a human
either mentally or with the aid of pen and paper. Specifically, we agree that
receiving data (steps [a] and [b]) is an activity of observation. The recited
steps of providing a network model representing the biological system (step
[c]) and generating a mechanism hypothesis (step [d]) are activities of
judgment and opinion, whereas calculating activity measures and weight
values for the downstream biological entities (steps [e] and [f]), and
Appeal 2020-000630
Application 14/124,826

20
generating a scorable model and a score for the model (steps [g] and [h]), are
activities of evaluation. We are persuaded that nothing would prevent one of
ordinary skill in the field of the observation and evaluation of biological
systems from performing each of these steps without a computer, for at least
a small scale system.
Appellant argues that some of the data and activity values of the
biological entities represented by nodes in the claimed process cannot be
measured directly, but can be, instead, “inferred based on their interactions
with other entities,” for example by using measured data or activity values of
other entities to infer data. (Appeal Br. 21, citing Spec. ¶¶ 49, 53, 57.)
However, these inferences are based on observations and therefore could be
performed in the human mind, as well. The Specification refers to methods
for such inferences, including by “overrepresentation of functionally-related
genes within the differentially expressed genes, Bayesian network analysis, a
graphical Gaussian model technique or a gene relevance network technique,
to identify a relevant biological network based on a set of experimental data
(e.g., gene expression, metabolite concentrations, cell response, etc.).”
(Spec. ¶ 53.) The claims, however, do not require that any of these specific
techniques be performed. Appellant fails to explain how inferences based
on overrepresentation of related genes could not be performed in the human
mind. Similarly, the Specification explains that “[t]he mechanism
hypothesis can be used to make predictions, such as if the abundance of an
entity represented by an upstream node increases, the downstream nodes
linked by causal increase relationships would be inferred to be increase, and
the downstream nodes linked by causal decrease relationships would be
inferred to decrease.” (Spec. ¶ 57.) Appellant argues that such predictions
Appeal 2020-000630
Application 14/124,826

21
and hypotheses are not based on formulaic approaches, but on causal
relationships. (See Appeal Br. 22.) Appellants have not provide adequate
evidence that such inferences, predictions, and hypotheses could not be
performed in the human with the aid of paper and pen, particularly where
they are not based on formulas.
In addition to agreeing with the Examiner that Appellant’s claimed
method recites a mental process, the steps of the method are also based on
the natural relationship between treatment with an agent and the perturbation
of the biological system in response. In Mayo, the steps of “(a)
administering a drug providing 6–thioguanine to a subject having [an]
immune-mediated gastrointestinal disorder; and (b) determining the level of
6–thioguanine in said subject having said immune-mediated gastrointestinal
disorder” were found to set forth only the relationship between
concentrations of certain metabolites in the blood and the likelihood that a
dosage of a drug would prove ineffective or cause harm. Mayo, 566 U.S. at
74 (internal quotation omitted). This relationship was determined to be a
law of nature. See id.
Similarly, the recitation in claim 1 of receiving data in response to an
agent, generating a score for a model of the data by generating a hypothesis,
and calculating activity measures and weights for downstream biological
entities, merely sets forth the natural law of the effect of the agent on the
downstream biological entities. In Mayo, “[t]he relation [between
concentrations of metabolites in the blood and the likelihood that a dosage of
the drug will be ineffective or cause harm] is a consequence of the ways in
which thiopurine compounds are metabolized by the body—entirely natural
processes. And so a patent that simply describes that relation sets forth a
Appeal 2020-000630
Application 14/124,826

22
natural law.” Mayo, 566 U.S. at 77. In Appellant’s claim 1 the relation
between the effect of the agent and a downstream biological entity is a
consequence of the ways in which the agent acts on the biological system –
also an entirely natural process.
Appellant argues that the steps of claim 1 cannot be performed
practically in the human mind because they require a large-scale treatment
data set to ensure that the variability of a vast amount of data is taken into
account. (See Appeal Br. 23–24; see Reply Br. 12–13.) We are not
persuaded by this argument because claim 1 does not recite a limitation on
the size of the data set from which the treatment and control data is obtained.
Nor does claim 1 recite any degree of certainty or confidence in determining
a score representative of the perturbation of the biological system that would
indicate how big the data set must be. Accordingly, claim 1 could
encompass even small databases and limited biological systems.
We are also unpersuaded by Appellant’s argument regarding the size
of the data sets because simply reciting the use of a computer to perform
complicated calculations or algorithms that could be performed by the
human mind does not necessarily make a method patentable. See
CyberSource Corp. v. Retail Decisions, Inc., 654 F.3d 1366, 1375 (Fed. Cir.
2011) (“That purely mental processes can be unpatentable, even when
performed by a computer, was precisely the holding of the Supreme Court in
Gottschalk v. Benson [409 U.S. 63, 67 (1972)].”); see Versata Dev. Grp.,
Inc. v. SAP Am., Inc., 793 F.3d 1306, 1335 (Fed. Cir. 2015) (“Courts have
examined claims that required the use of a computer and still found that the
underlying, patent-ineligible invention could be performed via pen and paper
or in a person’s mind.”). Therefore, although the claim may recite steps
Appeal 2020-000630
Application 14/124,826

23
which are accomplished on a computer, the computer-implementation does
not confer patent-eligibility because the steps could still be performed by a
human. See Intellectual Ventures I LLC v. Symantec Corp., 838 F.3d 1307,
1318 (Fed. Cir. 2016) (“Furthermore, with the exception of generic
computer-implemented steps, there is nothing in the claims themselves that
foreclose them from being performed by a human, mentally or with pen and
paper.”).
Appellant refers to the example in the Specification of TNF-treated
NHBE cells, where a computational causal network model was formed out
of a database repository with 1.5 million nodes and over 7.5 million edges.
(See Appeal Br. 23, citing Spec. ¶ 100.) Although Appellant argues that
scanning such a vast database could not possibly be performed in the human
mind, we do not limit claim 1 to this example. “Claims, not the specification
embodiments, define the scope of the protection.” American Permahedge,
Inc. v. Barcana, Inc., 105 F.3d 1441, 1444 (Fed. Cir. 1997).
Appellant also argues that the method of claim 1 is not an abstract
idea in light of Example 39 of the Subject Matter Eligibility Examples
regarding abstract issued by the USPTO on January 7, 2019. (See Appeal
Br. 20–24, citing
https://www.uspto.gov/sites/default/files/documents/101_examples_37to42_
20190107.pdf.) Appellant argues that, like the claim directed to a method
for training a neural network in Example 39, Appellant’s claims “do not
recite any mathematical concepts including mathematical relationships,
mathematical formulas or equations, or mathematical calculations.” (Appeal
Br. 21.) Appellant argues that instead, the pending claims are drawn to
“using a computational causal network model to simulate interaction within
Appeal 2020-000630
Application 14/124,826

24
a biological system, taking into account edge and direction values which
represent causal activation relationships between nodes which represent sets
of biological entities . . . .” (Id.)
In Example 39, the steps were found to address the problem of false-
positives encountered with prior methods by training a neural network. The
claims were characterized as not invoking the mental process category of
abstract ideas. The difference between Appellant’s claims and the claim of
the network is that the Examiner found that steps in the claims could be
performed in the human mind, whereas the Example steps could not be
practically performed in the mind. An improvement in the judicial
exception, itself, cannot serve as the technological improvement upon which
eligibility is based. Eligibility Guidelines 55 (fn. 24).
Appellant characterizes the claimed method as not being based on any
mathematical formula or algorithm, but rather on dynamic comparisons and
evaluations with respect to received data. (See Appeal Br. 22; see Reply Br.
7–11.) According to Appellant, “although a network model is based off of a
mathematical concept, the model as recited is not itself a mathematical
concept, wherein the network model itself is not a mathematical relationship,
formula/equation, or calculation, as described in p. 3 of the USPTO update.”
(Reply Br. 19.) Appellant’s Specification that a “network model of a
biological system is a mathematical construct that is representative of a
dynamic biological system and that is built by assembling quantitative
information about various basic properties of the biological system.” (Spec.
¶ 47; see Ans. 18–19.) The Specification also states that a score “is
computed by using any of various mathematical and computational
algorithms known in the art and according to the methods disclosed herein,
Appeal 2020-000630
Application 14/124,826

25
employing one or more datasets obtained from a sample or a subject.”
(Spec. ¶ 30.)
Even if we agree that Appellant’s claims do not recite mathematical
relationships, formulas, or calculations as determined in Example 39, the
claim in Example 39 provides a trained neural network, a machine learning
process that bases input on previously learned training processes to train the
network to detect faces, whereas Appellant’s claims merely generate a score.
The Examiner notes further that Appellant’s claims include “‘calculating’
steps” and steps to generate a score. (See Ans. 19.) Thus, even if
Appellant’s claims do not recite mathematical concepts as equations, we are
not persuaded by Appellant’s argument that they are not based on
mathematical concepts based on at least this step of the claim. (See Ans.
18.)
Because we find that the Examiner did not err in determining
Appellant’s claims to recite the judicial exception of an abstract idea,
including the mental process of receiving data, generating a mechanism
hypotheses, and then calculating activity measures and weight values to
generate a score, we must inquire whether the claim as a whole integrates
the abstract ideas into a practical application under step 2A, prong 2 of the
2019 Guidelines. (See 2019 Guidelines, 84 Fed. Reg. at 54–55.)
We agree with the Examiner that Appellant’s claims do not describe
any specific computational steps or any specific structures for a computer to
carry out the abstract ideas. (See Ans. 23.) Appellant argues that the
claimed method solves the need for improved computing systems and
methods of analyzing system-wide biological data, but fails to point to any
improved computer technology that was developed to carry out the claimed
Appeal 2020-000630
Application 14/124,826

26
methods. (See Appeal Br. 25–28, Reply Br. 14, 24.) We are not persuaded
by this argument because Appellant’s Specification provides that the
methods can be carried out on “a conventional standalone computer.” (Spec.
¶ 108.) As the Examiner finds, Appellant’s claimed methods are not about
an improvement in computing technology, but rather an improvement in the
analysis of biological data. (See Ans. 21.)
According to Appellant, the claim element “generating, with the at
least one processor, a score for the scorable network model” creates a
network model to understand disease mechanisms and the underlying effect
of harmful agents, and can thus directly effect a particular treatment or
prophylaxis for a disease or medical condition as provided in the 2019
Guidelines. (See Appeal Br. 29–30; Reply Br. 25–29.) We are not
persuaded by Appellant’s argument because the claimed methods do not
recite any treatments or prophylaxis. (See 2019 Guidelines, 84 Fed. Reg. at
55; see Ans. 20.) Whereas claims that recite actually treatments or
prophylaxis, such as immunization or administration of a drug, have been
held to be patent eligible, Appellant’s claims recite no similar treatment or
prophylaxis. See Classen Immunotherapies, Inc. v. Biogen IDEC, 659 F.3d
1057, 1066–68 (Fed. Cir. 2011); Vanda Pharm. Inc. v. W.-Ward Pharm.
Int’l Ltd., 887 F.3d 1117, 1135 (Fed. Cir. 2018).
Because Appellant’s claims do not integrate the judicial exception
into a practical application, we continue to evaluate patent eligibility by
considering whether the claim provides an inventive concept that amounts to
significantly more than the exception itself, under step 2B of the 2019
Guidelines. (See 2019 Guidelines, 84 Fed. Reg. at 56.) Appellant argues
that the claimed methods more accurately predict the changes of entities
Appeal 2020-000630
Application 14/124,826

27
underlying biological mechanisms. (See Appeal Br. 28–30.) We disagree
because the claimed method simply appends understood, routine,
conventional activities previously known to the industry, specified at a high
level of generality to the judicial exception of the mental activity of
receiving data, generating a set of hypotheses, and scoring the hypotheses.
Claim 1 recites only generic computer components, “processors,” that are
used for implementing the abstract idea. Furthermore, for the reasons
discussed in the analysis of the Examiner’s rejection of claim 1 as being
obvious under 35 U.S.C. § 103(a), we are not persuaded the claimed
methods provide any inventive concept beyond what was known in the art
and the abstract ideas recited in the claims.
Accordingly, we are not persuaded that the Examiner erred in
rejecting Appellant’s pending claims as being directed to ineligible subject
matter under 35 U.S.C. § 101.

Conclusion
Upon consideration of the record and for the reasons given, we affirm
the Examiner’s rejection.
In summary:
Claims
Rejected
35 U.S.C. § Reference(s)/Basis Affirmed Reversed
1, 4–23, 26–
28
112, first
paragraph
1, 4–23, 26–
28
1, 4–23, 26–
28
101 1, 4–23, 26–
28

1, 4–13, 15–
21, 23, 28
103 Ladd, Toyoshiba,
Strimmer
1, 4–13, 15–
21, 23, 28

Appeal 2020-000630
Application 14/124,826

28
14 103 Ladd, Toyoshiba,
Strimmer, Löfroth
14
14, 26, 27 103 Ladd, Toyoshiba,
Strimmer, Sexton
14, 26, 27
22 103 Ladd, Toyoshiba,
Strimmer,
Friedman
22
Overall
Outcome
1, 4–23, 26–
28

No time period for taking any subsequent action in connection with
this appeal may be extended under 37 C.F.R. § 1.136.

AFFIRMED