DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Applicant’s amendments and arguments, filed 12/8/2025, have been carefully considered, but are not completely persuasive.
Claims 1-5, 7-10, 12-20, 24-29 are pending and under consideration. All other claims have been canceled.
The rejection of claims 1-5, 7-10, and 24-29 under 35 USC 101 is withdrawn, as the claims set forth particular neural network architecture, particular training, and particular details as to how the training affects the neural networks, to provide the cancer classification.
New Grounds of Rejection
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-5, 7-10, 24-29 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
The metes and bounds of the term “shallow neural network” in independent claims 1 and 24 are indefinite. The term “shallow” in claims 1 and 24 is a relative term which renders the claim indefinite. The term “shallow neural network” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. The plain meaning of “shallow” indicates a lack of depth. The plain meaning of “shallow neural network” in the computer processing art refers to a neural network architecture with only one or two hidden layers between input and output layers. At [0121] the specification defines genomic region models implementing a neural network as: “In some embodiments where each genomic region model implements a neural network algorithm, each genomic region comprises no more than one hidden layer, no more than two hidden layers, or no more than three hidden layers… Architectures of genomic regions may differ. For example, a first genomic region model may have a different number of hidden layers as a second genomic region…” In contrast, the specification at [0132] defines a “shallow neural network” as follows: “A shallow neural network can be a neural network with few hidden layers. Such neural network architectures can improve the efficiency of neural network training and conserve computational power due to the reduced number of layers involved in the training. A number of hidden layers in each genomic region model can be between two and five hidden layers, or more than five layers.” [0133] sets forth “A genomic region model (e.g., a shallow neural network) can comprise an input layer that accepts inputs and an output layer that generates an output (e.g., a prediction value). The output can comprise a score (e.g., a probability or a likelihood) that an input (e.g., a fragment and/or a dataset) belongs to one or more predetermined classes (e.g., labels). The output can be determined by the genomic region model using a softmax or logistic regression algorithm.” A search of definitions of the prior art for “shallow neural network” indicate that in general neural networks with one hidden layer are considered shallow, while neural networks with more than one hidden layer represent deep neural networks, with some specific exceptions. It is unclear if Applicant intends [0132-0133] to represent a special definition of the term “shallow neural network” to specifically incorporate neural networks having from one hidden layer to more than five hidden layers. (MPEP 2111, 2173) The BRI of this term consistent with the specification will be interpreted herein as a neural network architecture having from one, to approximately five or more hidden layers. If a different interpretation is intended, language such as “a shallow neural network having fewer than three hidden layers” could be employed.
Rejections Maintained
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 12-20 is/are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea of mental steps, mathematic concepts, organizing human activity, or a natural law without significantly more.
Applicant is directed to MPEP 2106 and the Federal Register notice (FR89, no 137 (7/17/2024) p 58128-58138) for the most current and complete guidelines in the analysis of patent- eligible subject matter. The current MPEP is the primary source for the USPTO’s patent eligibility guidance.
With respect to step (1): YES, the claims are drawn to statutory categories: Processes.
With respect to step (2A) (1): YES, the claims recite an abstract idea, law of nature and/or natural phenomenon. The claims explicitly recite elements that, individually and in combination, constitute one or more judicial exceptions (JE).
Abstract ideas are defined in MPEP 2106 as 1) mathematic concepts, including mathematical relationships, formulas, equations or calculations (MPEP 2106.04(a)(2) section I); 2) certain methods of organizing human activity (MPEP 2106.04(a)(2) section II); and 3) mental processes, including concepts performed in the human mind involving observation, evaluation, judgement and opinion (MPEP 2106.04(a)(2) section III).
Mathematic concepts, Mental Processes or Elements in Addition (EIA) in independent claim(s) 12 include:
12. (Currently Amended) A method comprising:
performing, by a sequencing device, a sequencing assay with a targeted panel targeting a plurality of genomic regions, wherein the targeted panel includes, for each genomic region, one or more hybridization probes for capturing cfDNA fragments originating from the genomic region, wherein the sequencing assays yields sequencing data for a biological sample comprising a plurality of cfDNA fragments, each sequence read overlapping at least one genomic region of a plurality of genomic regions;
(EIA- preamble and: Data Gathering step of performing targeted sequencing, using a routine sequencer, with details of the assay and the results gathered. Targeted sequencing is well known: [0056-0060].)
for each cfDNA fragment of the biological sample, generating, by a computer processor, a methylation embedding by inputting the sequencing data for the cfDNA fragment into a trained embedding model, the trained embedding model configured to generate a methylation embedding based on an input cfDNA fragment, each trained embedding model of a plurality of trained embedding models spanning the plurality of genomic regions trained by:
accessing a first plurality of sequence reads obtained from cancer samples and a second plurality of sequence reads obtained from non-cancer samples, wherein the first plurality of sequence reads and the second plurality of sequence reads overlap the genomic region, and
training the trained embedding model with the first plurality of sequence reads and the second plurality of sequence reads;
(Mathematic concept of calculating a “methylation embedding” score using a trained, unspecified model. The aspect of the data from the assay required to calculate the embedding is not set forth. [0117]: “A methylation embedding can be a mathematical vector that captures the methylation signature of a DNA fragment. The DNA fragment or its methylation state vector can describe at least a methylation status of each CpG site covered by the DNA fragment. Generally, the methylation embedding model 310 can reduce dimensionality of the fragment space into an embedding space…” the embedding is a mathematic process. Training occurs outside the bounds of the claim.)
for each cfDNA fragment of the biological sample, generating, by the computer processor, a region embedding for the genomic region that the cfDNA fragment overlaps, the region embedding for a genomic region determined by inputting the methylation embedding of the cfDNA fragment into a region model trained for the genomic region, the region model configured to generate a region embedding based on an input methylation embedding;
(Mathematic concept of collecting, aggregating or otherwise combining methylation embedding scores for each cfDNA fragment for each region using an unspecified model. How the region embedding is calculated is not set forth. [0123]: “a genomic region model may be trained with a fragment classifier. In such an embodiment, the genomic region model is configured to output a region embedding. Fragments or their methylation embeddings are fed through the genomic region model which outputs a region embedding… The analytics system trains the genomic region model… by adjusting weights of the genomic region model… to minimize a loss function between the known labels of the fragments and the predicted labels of the fragment.”)
for each genomic region, determining, by the computer processor, an aggregate region vector by pooling one or more region embeddings of one or more cfDNA fragments overlapping the genomic region;
(Mathematic concept of aggregating, collecting, combining or pooling data from region embeddings. [0157]: “A first pooling step can determine an aggregate region vector for each genomic region by pooling the region embeddings of DNA fragments in each genomic region... Each pooling step can include performing an average pooling operation, a max pooling operation, another weighted geometric pooling operation, another pooling operation, or some combination thereof.”)
determining, by the computer processor, a feature vector by pooling the aggregate region vectors of the genomic regions; and
(Mathematic concept of aggregating, collecting, combining or pooling data from aggregate region vectors. [0157]: “A second pooling step can determine the feature vector by pooling the aggregate region vectors across the genomic regions. Each pooling step can include performing an average pooling operation, a max pooling operation, another weighted geometric pooling operation, another pooling operation, or some combination thereof.”)
inputting, by the computer processor, the feature vector into a machine learning classification model to generate a cancer prediction for the biological sample, the classification model trained by:
accessing sequencing data for a plurality of cancer samples and a plurality of non-cancer samples,
generating a feature vector for each cancer sample and for each non-cancer sample by applying the trained embedding models to the sequencing data of each sample, and
training the classification model with the feature vectors for the cancer samples and the non-cancer samples.
(Mathematic concept of classification, where the feature vector value is classified by an unspecified ML classification model. [0161]: “the analytics system can train the cancer classifier 340 by inputting sets of training samples with their feature vectors into the cancer classifier 340 and adjusting classification parameters so that a function of the classifier accurately relates the training feature vectors to their corresponding label... The analytics system may train the cancer classifier 340 according to any one of a number of methods. As an example, the binary cancer classifier may be a L2-regularized logistic regression classifier that is trained using a log-loss function… may be a multinomial logistic regression… may be trained using other techniques… including potential use of kernel methods, decision trees, random forest classifier, a mixture model, an autoencoder model, machine learning algorithms such as multilayer neural networks, etc.” Training occurs outside the bounds of the claim.)
Natural law embraced by independent claim(s) 12:
The claims also embrace the natural law describing the naturally occurring correlations between naturally occurring cfDNA fragments in a sample, and a naturally occurring phenotype: cancer. (MPEP 2106.04)
The sequence information of the sample is a natural phenomenon and the relationship of that information to a phenotypic trait is simply put a natural law. Nothing more than this observation is required by the claims. In Mayo, the discovery underlying the claims was that when blood levels were above a certain level harmful effects were more likely and when they were below another level the drug's beneficial effects were lost. Mayo, 566 U.S. at 74--76. The claims provided that particular levels of measured metabolite indicated a need to increase or decrease the amount of drug subsequently administered to the subject. Id. at 75. However, the claims did not require any actual action be taken based on the measured level of metabolite. Id. at 75-76. Thus, the claims which required only the observation of a natural law were deemed patent-ineligible. Similarly, here, the claim requires only the observation of the natural law, and for this reason too are properly deemed patent-ineligible.
These correlate to at least the following examples provided in MPEP 2106.04b:
“iii. a correlation between variations in non-coding regions of DNA and allele presence in coding regions of DNA, Genetic Techs. Ltd. v. Merial LLC, 818 F.3d 1369, 1375, 118 USPQ2d 1541, 1545 (Fed. Cir. 2016);
iv. a correlation that is the consequence of how a certain compound is metabolized by the body, Mayo Collaborative Servs. v. Prometheus Labs., 566 U.S. 66, 75-77, 101 USPQ2d 1961, 1967-68 (2012);
v. a correlation between the presence of myeloperoxidase in a bodily sample (such as blood or plasma) and cardiovascular disease risk, Cleveland Clinic Foundation v. True Health Diagnostics, LLC, 859 F.3d 1352, 1361, 123 USPQ2d 1081, 1087 (Fed. Cir. 2017);
vii. qualities of bacteria such as their ability to create a state of inhibition or non-inhibition in other bacteria, Funk Bros., 333 U.S. at 130, 76 USPQ at 281; and
xi. the natural relationship between a patient’s CYP2D6 metabolizer genotype and the risk that the patient will suffer QTc prolongation after administration of a medication called iloperidone, Vanda Pharmaceuticals Inc. v. West-Ward Pharmaceuticals, 887 F.3d 1117, 1135-36, 126 USPQ2d 1266, 1281 (Fed. Cir. 2018).”
With respect to step 2A (2): NO. The claims were examined further to determine whether they integrated any JE into a practical application (MPEP 2106.04(d)). The claimed additional elements are analyzed alone, or in combination to determine if the JE is integrated into a practical application (MPEP 2106.05(a-c, e, f and h)).
Independent claim(s) 12 recite(s) the additional non-abstract element(s) (EIA) of data gathering or a description of the data gathered.
Data gathering steps are not an abstract idea, they are extra-solution activity, as they collect the data needed to carry out the JE. The data gathering does not impose any meaningful limitation on the JE, or how the JE is performed. The additional limitation (data gathering) must have more than a nominal or insignificant relationship to the identified judicial exception. (MPEP 2106.04/.05, citing Intellectual Ventures LLC v. Symantec Corp, McRO, TLI communications, OIP Techs. Inc. v. Amason.com Inc., Electric Power Group LLC v. Alstrom S.A.).
Dependent claim(s) 13-20 have been analyzed with respect to 2A-2. See MPEP 2106.05(a, b, c, e and h).
Dependent claim(s) 13-20 each recite(s) an abstract limitation to the JE reciting additional mathematic concepts, or mental processes. Additional abstract limitations cannot provide a practical application of the JE as they are a part of that JE.
In combination, the limitations of data gathering, for the purpose of carrying out the JE, fail to integrate the JE into a practical application.
With respect to step 2B: NO. The claims recite a JE, do not integrate that JE into a practical application, and thus are probed for a specific inventive concept. The judicial exception alone cannot provide that inventive concept or practical application (MPEP 2106.05). The additional elements were considered individually and in combination to determine if they provide significantly more than the judicial exception. (MPEP 2106.05.A i-vi).
With respect to independent claim(s) 12: The limitation(s) identified above as non-abstract elements (EIA) related to data gathering do not rise to the level of significantly more than the judicial exception.
With respect to using a sequencing device for the purpose of sequencing, and targeted sequencing assays, on a test sample, or reference healthy/ cancer samples, these are well known in the prior art:
Gross (US 2019/0287652 A1) discloses performing targeted sequencing assays using a sequencing device to generate sequence read data from samples. [0085-0091]
Guo (WO 2019/084559 A1) discloses performing targeted sequencing assays using a sequencing device to generate sequence read data from samples. [0179-0181]
Mamanova (2010) discloses targeted sequencing assays using a sequencing device to generate sequence read data from samples.
Chen (April, 2019) discloses performing targeted sequencing assays using a sequencing device to generate sequence read data from samples containing circulating DNA from tumors (cell free- cfDNA and circulating tumor- ctDNA).
These elements meet the BRI of the identified data gathering limitations. As such, the prior art recognizes that this data gathering element was routine, well understood and conventional in the art (as in Alice Corp., CyberSource v. Retail Decisions, Parker v. Flook).
In the specification at [0046] it is disclosed that the steps identified as data gathering can be met using a variety of prior art known methods. “A sequence read can be obtained in a variety of ways, e.g., using sequencing techniques or using probes, e.g., in hybridization arrays or capture probes, or amplification techniques, such as the polymerase chain reaction (PCR) or linear amplification using a single primer or isothermal amplification.” In the specification, paragraphs [0056- 0064, 0205] set forth the use of widely used and commercially available kits (such as the EZ DNA Methylation™ kit) for the treatment of the sample, and commercial technologies (such as the SOLID platform, SMRT technology, ION TORRENT technology, HiSEQ analyzers et al.) for generating the sequencing reads, which are then received by the computer or user.
This underscores the argument that steps of performing targeted sequencing assays to generate cfDNA sequence reads, and receiving those sequence reads was routine, well understood and conventional in molecular biology.
Activities such as data gathering do not improve the functioning of a computer, or comprise an improvement to any other technical field. The limitations do not require or set forth a particular machine, they do not effect a transformation of matter, nor do they provide an unconventional step (citing McRO and Trading Technologies Int’l v. IBG). Data gathering steps constitute a general link to a technological environment. Simply appending well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception are insufficient to provide significantly more (as discussed in Alice Corp.,).
Dependent claim(s) 13-20 have been analyzed with respect to step 2B. MPEP 2106.06(a-h).
Dependent claim(s) 13-20 each recite a limitation requiring additional mathematic concepts or mental processes. Additional abstract limitations cannot provide significantly more than the JE as they are a part of that JE (MPEP 2106.05).
In combination, the data gathering steps providing the information required to be acted upon by the JE, fail to rise to the level of significantly more than that JE. The data gathering steps provide the data for the JE. No non-routine step or element has clearly been identified.
The claims have all been examined to identify the presence of one or more judicial exceptions. Each additional limitation in the claims has been addressed, alone and in combination, to determine whether the additional limitations integrate the judicial exception into a practical application. Each additional limitation in the claims has been addressed, alone and in combination, to determine whether those additional limitations provide an inventive concept which provides significantly more than those exceptions. For these reasons, the claims, when the limitations are considered individually and as a whole, are rejected under 35 USC § 101 as being directed to non-statutory subject matter.
Applicant’s arguments:
Applicant’s arguments have been carefully considered, but are not completely persuasive.
With respect to claim 12, claim 12 does not recite specific neural network embodiments. Claim 12 recites methylation embedding by a trained embedding model. The specification at [0117-0118] defines methylation embedding as “A methylation embedding can be a mathematical vector
that captures the methylation signature of a DNA fragment.” The methylation embedding model is defined as “Generally, the methylation embedding model 310 can reduce dimensionality
of the fragment space into an embedding space... Some approaches can include Principal Component Analysis (PCA), t-distributed stochastic neighbor embedding, autoencoder, linear discriminant analysis, other dimensionality reduction techniques, or other embedding techniques. The methylation
embedding model may implement machine-learning algorithms, such as a neural network
algorithm, a support vector machine algorithm, a decision tree algorithm, a multinomial logistic
regression algorithm, a linear regression algorithm, or some other machine learning algorithm.
The methylation embedding model 310 may be trained independently or concurrently with other
components.” These clearly set forth mathematic concepts, relationships, and calculations as defined in MPEP 2106. The examiner suggests considering limiting the methylation embedding model to a data element with inherent structure, such as an autoencoder, as described in [0118].
The examiner acknowledges Applicant’s arguments which set forth that the claims lead to an improvement in prediction of cancer. According to the guidance set forth in MPEP 2106, this is an improvement to the judicial exception itself, and is not reflected back into a specific technological environment or practically applied process.
An improvement in the judicial exception itself is not an improvement in the technology. For example, in In re Board of Trustees of Leland Stanford Junior University, 989 F.3d 1367, 1370, 1373 (Fed. Cir. 2021) (Stanford I), Applicant argued that the claimed process was an improvement over prior processes because it ‘‘yields a greater number of haplotype phase predictions,’’ but the Court found it was not ‘‘an improved technological process’’ and instead was an improved ‘‘mathematical process.’’ The court explained that such claims were directed to an abstract idea because they describe ‘‘mathematically calculating alleles’ haplotype phase,’’ like the ‘‘mathematical algorithms for performing calculations’’ in prior cases. Notably, the Federal Circuit found that the claims did not reflect an improvement to a technological process, which would render the claims eligible (FR89 no.137, p58137, 7/17/2024).
Here, Applicant has provided an improved mathematical process of generating a cancer prediction.
The improvement in generation of a cancer prediction (carried out by the judicial exception) does not provide an improvement in the technology of sequencing a sample, or receiving sequencing data. The sequencing of the sample, and collection of sequencing data is carried out, unchanged, whether or not the judicial exception is applied. (Cleveland Clinic Foundation: using well-known or standard laboratory techniques is not sufficient to show an improvement (MPEP2106.05(a)).
The improvement in the generation of a cancer prediction (achieved by the judicial exception) does not require a non-conventional interaction with a specific element of a computer as was required in Enfish. The disputed claims in Enfish were patent-eligible because they were "directed to a specific improvement to the way computers operate, embodied in [a] self-referential table." Enfish, 822 F.3d at 1336. The court found that the "plain focus of the claims" there was on an improvement to computer functionality itself-a self-referential table for a computer database, designed to improve the way a computer carries out its basic functions of storing and retrieving data- not on a task for which a computer is used in its ordinary capacity. Id. at 1335-36. The court noted that the specification identified additional benefits conferred by the self-referential table (e.g., increased flexibility, faster search times, and smaller memory requirements), which further supported the court's conclusion that the claims were directed to an improvement of an existing technology. Id. at 1337 (citation omitted).
The improvement in the generation of a cancer prediction (carried out by the judicial exception) does not improve the functionality of the computer itself as in Finjan, Visual Memory, or SRI Int’l. In Finjan, claims to virus scanning were found to be an improvement in computer technology. In Visual Memory, claims to an enhanced computer memory system were found to be directed to an improvement in computer capabilities. In SRI Int'l, claims to detecting suspicious activity by using network monitors and analyzing network packets were found to be an improvement in computer network technology.
The process of generating a cancer prediction using embedding and machine learning is not a technological process; it is information evaluation.
The addition of a routine sequencing device to claim 12, where the sequencing device is used according to its original purpose: for sequencing, fails to integrate the JE into a practical application, and fails to provide significantly more than the JE. The sequencing device does not carry out the elements of the JE, nor is it changed by carrying out the JE. The sequencing occurs, whether or not the JE is applied. The sequencing has no unusual or non-routine steps or elements. These data gathering steps do not apply, rely on, or use the steps identified as making up the JE. Rather, the steps identified as reciting the JE avail themselves of the data gathered. The data gathering in the claims constitutes insignificant pre-solution activity. See MPEP § 2106.05(g):
MPEP2106.05(g). “The term "extra-solution activity” can be understood as activities incidental to the primary process or product that are merely a nominal or tangential addition to the claim...”
“An example of pre-solution activity is a step of gathering data for use in a claimed process, e.g., a step of obtaining information about credit card transactions, which is recited as part of a claimed process of analyzing and manipulating the gathered information by a series of steps in order to detect whether the transactions were fraudulent.”
See also CyberSource Corp. v. Retail Decisions, Inc., 654 F.3d 1366, 1372 (Fed. Cir. 2011) ("[E]ven if some physical steps are required to obtain information from the database ... such data-gathering steps cannot alone confer patentability.").
The addition of a routine computer processor, for carrying out computer processing, fails to integrate the JE into a practical application, and fails to provide significantly more than the JE. The computing system limitations are recited at such a high level of generality, they can be met by a general-purpose computer system and are not considered a particular machine or manufacture integral to the claim (MPEP 2106.05(b)). Routine computer elements acting upon the data in a manner consistent to and according to their design are not considered to be sufficient to provide eligibility. (see, for example MPEP 2106.04(d): Gottschalk v. Benson “‘held that simply implementing a mathematical principle on a physical machine, namely a computer, was not a patentable application of that principle.”)
Further, with respect to the arguments regarding the alleged improvement, it is unclear that the independent claims recite all the necessary and sufficient steps required to achieve that improvement. MPEP 2106.05(a): “An important consideration in determining whether a claim improves technology is the extent to which the claim covers a particular solution to a problem or a particular way to achieve a desired outcome, as opposed to merely claiming the idea of a solution or outcome. McRO, 837 F.3d at 1314-15, 120 USPQ2d at 1102- 03; DDR Holdings, 773F.3d at 1259, 113 USPQ2d at 1107.”
The MPEP sets forth that “if the examiner concludes the disclosed invention does not improve technology, the burden shifts to applicant to provide persuasive arguments supported by any necessary evidence to demonstrate that one of ordinary skill in the art would understand that the disclosed invention improves technology. Any such evidence submitted under 37 CFR 1.132 must establish what the specification would convey to one of ordinary skill in the art and cannot be used to supplement the specification.” Applicant’s arguments cannot take the place of evidence.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claim(s) 1, 3, 5, 7-10, 24, 26-29 is/are rejected under 35 U.S.C. 102a1 as being anticipated by Khwaja (2018).
Applicant’s earliest priority date is 3/31/2020.
Khwaja, M. et al. (5 Oct 2018) A deep autoencoder system for differentiation of cancer types based on DNA methylation state. arXiv: 1810.01243v2, 8 pages.
Claim 1 has been amended to recite a sequencing device, a processor, and the use of “shallow neural networks”. As set forth above, at [0121] the specification defines genomic region models implementing a neural network as: “In some embodiments where each genomic region model implements a neural network algorithm, each genomic region comprises no more than one hidden layer, no more than two hidden layers, or no more than three hidden layers… Architectures of genomic regions may differ. For example, a first genomic region model may have a different number of hidden layers as a second genomic region…” In contrast, the specification at [0132] defines a “shallow neural network” as follows: “A shallow neural network can be a neural network with few hidden layers. Such neural network architectures can improve the efficiency of neural network training and conserve computational power due to the reduced number of layers involved in the training. A number of hidden layers in each genomic region model can be between two and five hidden layers, or more than five layers.” The BRI of this term consistent with the specification will be interpreted herein as a neural network architecture having from one, to approximately five or more hidden layers.
With respect to claim 1, a plurality of samples containing cfDNA fragments, both cancer and non-cancer, were processed using a targeted sequencing assay (Illumina Human450 BeadChip platform) to generate sequence reads by a sequencing device. The sequence reads each overlapped a known genomic region. These were scored, to provide a first score, based on a trained deep belief neural network. The trained deep belief neural network has three hidden layers, as shown in Fig 3, meeting the BRI of a “shallow neural network.”
The trained deep belief neural network predicts the methylation state of each sequence read, and a likelihood it is from a cancer cfDNA fragment. Fig 2-3. Fig 3 is the specific DBN, which is the first box of Fig 2.
A feature vector of each sample is created by counting. (p3-4)
The sample feature vector is input into a trained deep autoencoder network, trained on sequencing data from healthy and known cancer samples, to make a cancer prediction and to classify the type of cancer. (see three boxes on right of Fig 2) See pages 4-5 “feature processing and reduction” and “deep autoencoder network”. Figure 5 discloses more of this second model: the deep autoencoder. This meets the limitations of claim 1, and the trained model meets the limitations of claim 24.
With respect to claim 3, 26 the hidden layers have any number of nodes.
With respect to claim 5, 27 genomic regions with differing sites are all analyzed.
With respect to claim 7, 28 up to four different types of cancer are discriminated, by generating scores for each type of cancer.
With respect to claim 8, 29, normalization is part of the pre-processing.
With respect to claim 9, the fragment sequence data can be filtered for anomalous fragments.
With respect to claim 10, Khaja utilizes neural network algorithms that are deep belief networks, and deep autoencoder networks.
Applicant’s arguments:
Applicant’s arguments with respect to the “deep” or “shallow” neural networks have been considered, but are not persuasive, as Applicant’s definition of a shallow neural network encompasses the structures provided by Khwaja. Limiting the shallow neural network to having fewer than three hidden layers would appear to exclude Khwaja.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claim(s) 12-16, 18-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Khaja as set forth over claims 1, 3, 5, 7-10, 24, 26-29 above, in view of Li.
Claim 12 is a variation of the method of claim 1.
Khwaja, M. et al. (5 Oct 2018) A deep autoencoder system for differentiation of cancer types based on DNA methylation state. arXiv: 1810.01243v2, 8 pages.
Li, W. et al. (12 June, 2018) CancerDetector: ultrasensitive and non-invasive cancer detection at the resolution of individual reads using cell-free DNA methylation sequencing data. NAR vol 46, no 15, e89, 15 pages and some supplemental material.
Claim 12 recites methylation embedding, and methylation embedding models, which are defined in the specification at [0117] as: “A methylation embedding can be a mathematical vector
that captures the methylation signature of a DNA fragment. The DNA fragment or its methylation state vector can describe at least a methylation status of each CpG site covered by the DNA fragment. Generally, the methylation embedding model 310 can reduce dimensionality of the fragment space into an embedding space… Some approaches can include Principal Component Analysis (PCA), t-distributed stochastic neighbor embedding, autoencoder, linear discriminant analysis, other dimensionality reduction techniques, or other embedding techniques. The methylation embedding model may implement machine-learning algorithms, such as a neural network algorithm, a support vector machine algorithm, a decision tree algorithm, a multinomial logistic regression algorithm, a linear regression algorithm, or some other machine learning algorithm. The methylation embedding model 310 may be trained independently or concurrently with other components.”
With respect to claim 12, Khwaja teaches a plurality of samples containing cfDNA fragments, both cancer and non-cancer, were processed using a targeted sequencing assay (Illumina Human450 BeadChip platform) to generate sequence reads, by a sequencing device. The sequence reads each overlapped a known genomic region. These were scored, to provide a first score, based on a trained deep belief neural network. The trained deep belief neural network predicts the methylation state of each sequence read, and a likelihood it is from a cancer cfDNA fragment. Fig 2-3. Fig 3 is the specific DBN, which is the first box of Fig 2. This meets the BRI of generating methylation embedding for each cfDNA fragment overlapping a known region, using a neural network for the methylation embedding model.
A feature vector of each sample is created by counting. (p3-4)
The sample feature vector is input into a trained deep autoencoder network, trained on sequencing data from healthy and known cancer samples, to make a cancer prediction and to classify the type of cancer. (see three boxes on right of Fig 2) See pages 4-5 “feature processing and reduction” and “deep autoencoder network”. Figure 5 discloses more of this second model: the deep autoencoder.
Khaja does not teach that each genomic region is analyzed to provide region embedding vectors.
In the same field of research, Li provides methods of detecting cancer, using cfDNA fragments and methylation patterns, where after each cfDNA is analyzed, all cfDNA overlapping a given region are pooled to generate a region embedding, including pooling or aggregation. Li provides using these embeddings to generate a feature vector, which is then submitted to a classification model to provide the prediction.
In KSR Int 'l v. Teleflex, the Supreme Court, in rejecting the rigid application of the teaching, suggestion, and motivation test by the Federal Circuit, indicated that “The principles underlying [earlier] cases are instructive when the question is whether a patent claiming the combination of elements of prior art is obvious. When a work is available in one field of endeavor, design incentives and other market forces can prompt variations of it, either in the same field or a different one. If a person of ordinary skill can implement a predictable variation, § 103 likely bars its patentability.” KSR Int'l v. Teleflex lnc., 127 S. Ct. 1727, 1740 (2007).
Applying the KSR standard of obviousness to Khaja and Li, the Examiner concluded that the combination of the two trained neural networks, the first generating a first methylation embedding as disclosed by Khaja with the pooled region data as disclosed by Li represents the use of a known technique to improve similar methods. The nature of the problem to be solved may lead inventors to look at references relating to possible solutions to that problem. Li provided a technical solution which would have improved upon the methods of Khaja, because it eliminates cfDNA fragment information which is irrelevant to the analysis, and reduces the number of feature sets to be tested. Therefore, it would have been obvious to use the region modeling of Li to improve the selectivity and sensitivity of the classification algorithm, to determine the presence or absence of cancer. As such the invention as claimed would have been prima facie obvious to one of ordinary skill in the art at the time of filing, absent evidence to the contrary.
With respect to claim 13, Li discloses a multiplicity of genomic regions with certain numbers of CpG sites. Khaja uses CpG islands with limited numbers of CpG sites.
With respect to claim 14, both Li and Khaja set forth filtering out reads that are not anomalous, or that do not meet a threshold.
With respect to claims 15-16, the poolings are average poolings in Li.
With respect to claim 18, Khaja teaches neural network algorithms, Li provides SVM and random forest.
With respect to claims 19-20, both Li and Khaja teach binary classification. Khaja teaches multi-class predictions.
Applicant’s arguments:
Applicant’s arguments with respect to Khwaja and Li have been considered but are not persuasive. The elements of Khwaja meet the requirements for the embedding models, which are used to generate vectors, to be applied to a classification model. Li provides the required regional aspect.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MARY K ZEMAN whose telephone number is 5712720723. The examiner can normally be reached on 8am-2pm M-F. Email may be sent to mary.zeman@uspto.gov if the appropriate permissions have been filed.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Larry Riggs can be reached on 571 270-3062. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/MARY K ZEMAN/ Primary Examiner, Art Unit 1686