DETAILED ACTION
Applicant’s response, filed 01 Sept. 2026 has been fully considered. The following rejections and/or objections are either reiterated or newly applied. They constitute the complete set presently being applied to the instant application.
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 .
Status of Claims
Claims 1-29, 36-103, 105-106, and 117-118 are cancelled.
Claim 120 is newly added.
Claims 30-35, 104, 107-116, and 119-120 are pending.
Claims 30-35, 104, 107-116, and 119-120 are rejected.
Claim 120 is objected to.
Priority
Applicant’s claim for the benefit of a prior-filed application, U.S. Provisional App. Nos. 62/878,095 filed 24 July 2019 and 62/949,025 filed 17 Dec. 2019 under 35 U.S.C. 119(e) or under 35 U.S.C. 120, 121, 365(c), or 386(c) is acknowledged.
Accordingly, the effective filing date of the claimed invention is 24 July 2019.
Claim Objections
Claim 120 is objected to because of the following informalities:
Claim 120 recites “determining, by a trained machine learning model…”, which should be amended to recite “determining, by the trained machine-learning model” to use consistent language with the first step of the claim and increase clarity.
Appropriate correction is required.
Claim Interpretation
Claim 30 recites “determining, by a trained machine-learning model…, wherein the trained machine-learning model is trained using gene-expression data from brain tumors”. The wherein clause regarding how the trained machine-learning model is trained is interpreted to be a product-by-process limitation defining the process in which the machine learning model was previously trained. However, a step of training the machine learning model using gene-expression data from brain tumors is not required by the claim. "[E]ven though product-by-process claims are limited by and defined by the process, determination of patentability is based on the product itself. The patentability of a product does not depend on its method of production. If the product in the product-by-process claim is the same as or obvious from a product of the prior art, the claim is unpatentable even though the prior product was made by a different process." In re Thorpe, 777 F.2d 695, 698, 227 USPQ 964, 966 (Fed. Cir. 1985) (citations omitted). See MPEP 2113 I.
Claim 119 recites “receiving gene-expression data comprising an expression metric for each of a set of genes measured in a sample collected from the subject, wherein the set of genes consists of two or more of SV2A…”. MPEP 2111.03 II. states when the phrase "consisting of" appears in a clause of the body of a claim, rather than immediately following the preamble, there is an "exceptionally strong presumption that a claim term set off with ‘consisting of’ is closed to unrecited elements." Multilayer Stretch Cling Film Holdings, Inc. v. Berry Plastics Corp., 831 F.3d 1350, 1359, 119 USPQ2d 1773, 1781 (Fed. Cir. 2016). However, the "consisting of" phrase limits only the element set forth in that clause; other elements are not excluded from the claim as a whole. See also In re Crish, 393 F.3d 1253, 73 USPQ2d 1364 (Fed. Cir. 2004) (The claims at issue "related to purified DNA molecules having promoter activity for the human involucrin gene (hINV)." Id., 73 USPQ2d at 1365. In determining the scope of applicant’s claims directed to "a purified oligonucleotide comprising at least a portion of the nucleotide sequence of SEQ ID NO:1 wherein said portion consists of the nucleotide sequence from … to 2473 of SEQ ID NO:1, and wherein said portion of the nucleotide sequence of SEQ ID NO:1 has promoter activity," the court stated that the use of "consists" in the body of the claims did not limit the open-ended "comprising" language in the claims (emphases added). Id. at 1257, 73 USPQ2d at 1367. Therefore, in the instant case, the “gene expression data comprising an expression metric for each of a set of genes” is not closed to unrecited elements and may include other sets of genes (analogous to the oligonucleotide comprising at least a portion.., wherein the portion consists of…).
Claim Rejections - 35 USC § 112(b)
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.
Claim 30-35, 104, and 107-116 are rejected under 35 U.S.C. 112(b) as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor regards as the invention. This rejection is newly recited and necessitated by claim amendment.
Claim 30, and claims dependent therefrom, are indefinite for recitation of “determining, by a trained machine-learning model, that the tumor that is not located in the brain is non-neuronal and non-neuroendocrine based on determining that the gen-expression data does not correspond to a neuronal genetic signature, wherein the machine learning model is trained…such that the model comprises a plurality of learned parameters defining a hyperplane separating a brain tumor class and a non-neuronal, non-neuroendocrine tumor class”. Given the tumor of the subject is already known to be “not located in a brain of the subject”, the metes and bounds of what classes are intended by a “brain tumor class” are not clear. The use of “brain tumor” class, suggests the classification is for a tumor located in a brain. However, given the machine learning model that predicts whether a tumor is a brain tumor or is a non-neuronal, non-neuroendocrine tumor class is making a prediction on a tumor already known to not be a brain tumor, this suggests the “brain tumor” class is not intended to be restricted to tumors located in the brain. This is further evidenced by claim 33, which requires that the training gene-expression data from the plurality of reference brain tumors include a “first label of being indicative of a tumor having a neuronal property”. This further raises confusion on whether the “tumor brain class” is intended to include any tumor having a neuronal property, or if the tumor brain class is restricted to tumors located in the brain (which is in conflict with claim 33). A review of Applicant’s specification at para. [0047] and [0064] suggests that the classes being predicted are a “neurally related” tumor, which is the label/class used for brain tumors, and a “non-neurally related tumor”, and Applicant’s specification does not appear to provide support for using labels for a “brain tumor” (i.e. a tumor located in the brain). Therefore, for purpose of examination, the “brain tumor class” is interpreted to be a “tumor having a neuronal property” class, consistent with Applicant’s specification and dependent claim 33.
Claim Rejections - 35 USC § 103
The rejection of claims 105-106 under 35 U.S.C. 103 in the Office action mailed 02 June 2026 has been withdrawn in view of the cancellation of these claims received 01 Sept. 2026.
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.
Claims 30-35, 104, 107-116, and 119-120 are rejected under 35 U.S.C. 103 as being unpatentable over Graeber (2019) in view of Spetzler (2018) and Taube (2018), as evidenced by Abdelhafez (2025). Any newly recited portion is necessitated by claim amendment.
Graeber et al., US 2022/0244263 A1, effectively filed 28 May 2019 (previously cited);
Spetzler et al., US 2018/0045727 A1 (previously cited);
Taube et al., Implications of the tumor immune microenvironment for staging and therapeutics, 2018, Modern Pathology, 31, pg. 214-234 (previously cited); and
Abdelhafez, Why Logistic Regression is a Linear Model despite the Non-Linear Sigmoid, 2025, Medium, pg. 1-9 (newly cited).
Regarding independent claims 30 and 119-120:
Regarding claim 30¸ Graeber discloses a method for identifying a neuroendocrine tumor in a patient (Abstract) by identifying neuroendocrine features in tumors ([0055]) comprising the following steps:
Graeber discloses accessing expression levels comprising an expression level (i.e. metric) for each of a plurality of biomarker genes, measured from a biological sample of the patient ([0011]-[0013]), wherein the set of genes include SV2A, NCAM1, ITGB6, SH2D3A, and TACSTD2 ([0163], e.g. genes in SCN gene signature; [0008]; [0017], e.g. biomarkers in Table 1 measured; claim 8). These genes are part of a small cel neuroendocrine signature (SCN) and thus associated with neuroendocrine identify.
Graeber determining the gene expression levels’ are different than a to a small cell neuroendocrine (SCN) signature (i.e. a neuronal genetic signature) using a trained machine learning model, including a logistic regression model ([0020], e.g. level of measured markers determined to statistically different compared to control expression levels of biomarkers in a SCN cancer; [0023]-[0024]; [0055]; [0053]-[0054] and [0263-[0265], e.g. prediction of samples with SCN using logistic regression model). Graeber discloses the model distinguishes between cancer types, including lung cancer (LUAD), small cell lung cancer (SCLC), ovary, breast, bladder, and pancreas cancer (i.e. non-brain tumors), that are positive or negative to SCN (i.e. determining that the non-brain tumor is non-neuronal and non-endocrine) ([0016]; [0265]; FIG. 15E-F). Given the trained machine learning model is a logistic regression model, the parameters of the model inherently define a hyperplane separating the negative SCN (i.e. non neuronal and non-neuroendocrine) to positive SCN classes (i.e. a tumor having a neuronal property), as evidenced by Abdelhafez, which discloses logistic regression classifiers generate a linear hyperplane separating the two classes (pg. 5, para. 2-4).
Regarding the process in which the machine learning model was previously trained, as discussed above in claim interpretation, this limitation is a product by process limitation defining the process in which the machine learning model was previously trained. In the instant case, the trained machine learning model of Graeber is configured to determine a non-brain tumor is non-neuronal and non-neuroendocrine based on the gene expression data not corresponding to a neuronal genetic signature, and Graeber further discloses the machine learning model is trained using expression data of genes of small cell neuroendocrine (SCN) (i.e. neurally related) positive tumors to predict SCN status of a sample ([0265]), such that the machine learning model comprises learned parameters for distinguishing neuronal from non-neuronal tumors. The machine learning model of Graeber trained on gene expression data of SCN positive tumors is considered the same product as the machine learning model trained on gene expression data of brain tumors, as evidenced by Applicant’s specification which at para. [0054], defines “neurally related” tumors have a molecular that is more similar to molecular profiles of tumor cells of a neural embryonic origin (e.g., cell lineages traceable back to the neural crest or the neural tube, including both central nervous system and neuroendocrine cell types) relative to molecular profiles of tumor cells not having a neural embryonic origin. Accordingly, the machine learning model and learned parameters of Graeber is the same product as the machine-learning model recited in the claims, even if trained by a different process, and thus reads on the limitation. See MPEP 2113 I. Furthermore, it is noted that training using expression data from brain tumors would have been obvious as applied to claim 120 below.
Graeber discloses selecting a cancer treatment including an immunotherapy of checkpoint inhibitors ([0063]-[0064]). Graeber discloses the patient has not been previously treated (i.e. the therapy is an initial use) ([0098]).
Regarding claim 119, Graeber discloses a method for identifying a neuroendocrine tumor in a patient (Abstract) by identifying neuroendocrine features in tumors ([0055]) comprising the following steps:
Graeber discloses accessing expression levels comprising an expression level (i.e. metric) for each of a plurality of biomarker genes, measured from a biological sample of the patient ([0011]-[0013]), wherein the set of genes include SV2A, NCAM1, ITGB6, SH2D3A, and TACSTD2 ([0163], e.g. genes in SCN gene signature; [0008]; [0017], e.g. biomarkers in Table 1 measured; claim 8). Graeber discloses that any one or more biomarkers from Table 1 may be measured ([0008]), which encompasses using all the biomarkers in Table 1 for classification (i.e. gene expression data comprising an expression metric for each of a set of genes consisting of SV2A, NCAM1, ITGB6, SH2D3A, and TACSTD2).
Graeber determining the gene expression levels’ are different than a to a small cell neuroendocrine (SCN) signature (i.e. a neuronal genetic signature) using a trained machine learning model ([0020], e.g. level of measured markers determined to statistically different compared to control expression levels of biomarkers in a SCN cancer; [0023]-[0024]; [0055]; [0053]-[0054] and [0263-[0265], e.g. prediction of samples with SCN using logistic regression model). Graeber discloses the model distinguishes between cancer types, including lung cancer (LUAD), small cell lung cancer (SCLC), ovary, breast, bladder, and pancreas cancer, that are positive or negative to SCN (i.e. determining that the tumor does not correspond to a neuronal genetic signature) ([0016]; [0265]; FIG. 15E-F).
Graeber discloses selecting a cancer treatment including an immunotherapy of checkpoint inhibitors ([0063]-[0064]). Graeber discloses the patient has not been previously treated (i.e. the therapy is an initial use) ([0098]).
Regarding claim 120, Graeber discloses a method for identifying a neuroendocrine tumor in a patient (Abstract) by identifying neuroendocrine features in tumors ([0055]) comprising the following steps:
Graeber discloses training a machine learning model using gene expression levels of biomarkers in SCN and non-SCN cancers ([0020], e.g. level of measured markers determined to statistically different compared to control expression levels of biomarkers in a SCN cancer; [0023]-[0024]; [0055]; [0053]-[0054] and [0263-[0265], e.g. prediction of samples with SCN using logistic regression model after training).
Graeber discloses accessing expression levels comprising an expression level (i.e. metric) for each of a plurality of biomarker genes, measured from a biological sample of the patient ([0011]-[0013]), wherein the set of genes include SV2A, NCAM1, ITGB6, SH2D3A, and TACSTD2 ([0163], e.g. genes in SCN gene signature; [0008]; [0017], e.g. biomarkers in Table 1 measured; claim 8). Graeber discloses that any one or more biomarkers from Table 1 may be measured ([0008]), such that Graber discloses only measuring any one or more of SV2A, NCAM1, ITGB6, SH2D3A, and TACSTD2 above (i.e. the set of genes consists of one or more of SV2A, NCAM1, etc.).
Graeber determining the gene expression levels’ are different than a to a small cell neuroendocrine (SCN) signature (i.e. a neuronal genetic signature) using a trained machine learning model ([0020], e.g. level of measured markers determined to statistically different compared to control expression levels of biomarkers in a SCN cancer; [0023]-[0024]; [0055]; [0053]-[0054] and [0263-[0265], e.g. prediction of samples with SCN using logistic regression model). Graeber discloses the model distinguishes between cancer types, including lung cancer (LUAD), small cell lung cancer (SCLC), ovary, breast, bladder, and pancreas cancer, that are positive or negative to SCN (i.e. determining that the tumor does not correspond to a neuronal genetic signature) ([0016]; [0265]; FIG. 15E-F).
Graeber discloses selecting a cancer treatment including an immunotherapy of checkpoint inhibitors ([0063]-[0064]). Graeber discloses the patient has not been previously treated (i.e. the therapy is an initial use) ([0098]).
Further regarding claims 30 and 119-120, Graeber does not disclose the following limitations:
Regarding claims 30 and 119-120, Graeber does not explicitly disclose the identification of the therapy approach is based on the determination of the tumor as non-neuronal and non-neuroendocrine, and then administering an effective amount of the checkpoint blockade therapy to the subject.
However, Graeber does disclose that, in addition to distinct cancer types like SRBCTs, blood cancers, or small cell and non-small cell histologies seen in biopsies, the SCN phenotype exists along an expression signature-defined spectrum which influences therapeutic vulnerabilities in individual cancers, and thus screening for and targeting the SCN phenotype has clinical benefit ([0204]-[0205]), generally disclosing that SCN status informs treatment decisions.
Furthermore, Taube overviews the implications of the immune microenvironment in cancer therapeutics, including checkpoint inhibitors such as anti-PD-/PD-L1, and discloses characterizing the tumor immune microenvironment enables the identification of therapeutic strategies to guide first-line treatment algorithms (Abstract). Taube discloses various cancers in which checkpoint inhibitors are a first line of treatment either alone or in combination therapies (pg. 218, col. 2, para. 2; Figure 3). Taube discloses prognostic indicators of a response to checkpoint inhibitors involving immunoscores which quantify immune infiltration within the central region and invasive margin of a broad range of tumors, wherein patients with higher immunoscores (e.g. immune infiltration) are recommended for checkpoint inhibitor therapy (pg. 218, col. 1, para. 2-3; FIG. 4). Taube then explains that many solid tumor types, such as neuroendocrine tumors (e.g. such as SCN in Graeber) are not classically recognized as immune infiltrated, and thus if an immune checkpoint blockade were to be effective, it could be used in a combination treatment after treatment of some other inhibitor that first incites an immune response prior to treatment with the checkpoint inhibitor (pg. 227, col. 2, para. 2), rather than using a checkpoint inhibitor as an initial first-line of treatment. This demonstrates that neuroendocrine tumors, such as SCN, generally have low immune infiltration, or immunoscores and do not respond well to immune checkpoint inhibitors as an initial treatment. Last, Taube discloses the administration of checkpoint inhibitors to cancer patients results in responses to treatment between 2 and 12 weeks and show durable responses (pg. 215, col. 2, para. 2 to pg. 216, col. 1, para. 1)
It would have been prima facie obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the method of Graeber to have further identified the therapy approach based on the determination of the tumor as non-neuronal and non-neuroendocrine (e.g. not having a low immunoscore), and then administering an effective amount of the checkpoint blockade therapy to the subject, as shown by Taube, discussed above. One of ordinary skill in the art would have been motivated to combine the methods of Graeber and Taube to provide durable treatment responses to cancer patients who do not have tumors with low immune infiltration, as shown by Taube (pg. 215, col. 2, para. 2 to pg. 216, col. 1, para. 1). This modification would have had a reasonable expectation of success given Graeber discloses determining various cancer types do not have the SCN signature, including bladder cancer, head and neck squamous cell cancer, and lung cancer, ([0016]), which Taube discloses can be treated with checkpoint inhibitors (Table 4), and thus the method of Taube is applicable to Graeber.
Regarding claims 30 and 119-120, Graeber further does not disclose outputting an indication that the subject is amenable to the therapy approach.
However, Spetzler discloses methods and systems, including a computer system with a processor coupled to memory (claim 83) and non-transitory computer readable medium ([0310]) for molecular profiling of cancer via a clinical decision support system for personalized medicine (Abstract). Spetzler discloses the method involves generating a report that identifies a molecular profile for the patient that includes a list describing the expected benefit of the plurality of treatment options based on the assessed characteristics (i.e. outputting an indication the subject is amenable to a therapy approach) ([0006]; [0350]; [0360] FIG. 6). Spetzler additionally discloses clinical responses to immune checkpoint inhibitor therapy ranges by tumor type, and provides an example in which expression data is used to identify a subset of cancer cases as candidates for immune checkpoint inhibitor therapy ([0531]).
It would have been prima facie obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the method of Graeber to have further utilized the gene expression data to output an indication that the subject is amenable to the therapy approach, as shown by Spetzler, discussed above. One of ordinary skill in the art would have been motivated to combine the methods of Graeber and Spetzler in order to provide more informed and effective treatment options, resulting in improved patient care and enhanced treatment outcomes, as shown by Spetzler ([0006]). This modification would have had a reasonable expectation of success because both Graeber and Spetzler analyze gene expression data of a patient to identify cancer and cancer treatments, and thus the method of Spetzler is applicable to Graeber.
Further regarding claim 120, Graeber does not explicitly disclose the gene expression data used for training the machine learning model is from a plurality of reference brain tumors.
However, Graeber discloses the machine learning model may be a pan-cancer predictor of small cell neuroendocrine cancers which is trained on data from a plurality of cancer types ([0183]-[0185]; [0265], e.g. prediction of samples with SCN features using logistic regression, and samples included in the model were LUAD, SCLC, NEPC, CRPC, and the pan-cancer SCN positive cases confirmed by pathology (BRCA, CESC, COADREAD, ESCA, HNSC, LUAD, LUSC, PRAD, STAD, THCA; Table S5)). Graeber further discloses the cancer of a subject being determined to be a SCN cancer or not could be glioma (i.e. a brain tumor) ([0016]; [0146]) and scoring glioma cells according to their SCN-gene expression score ([0196]; FIG. 5A).
While Graeber does not explicitly disclose an embodiment in which the pan-cancer model was trained with SCN positive and negative cases for glioma (i.e. a brain tumor) in addition to the other tumor types, it would have been prima facie obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the pan-cancer model of Graeber to have further been trained on expression data from Glioma tumors based on the simple substitution of the known gene expression data from glioma samples of Graeber ([0196]; FIG. 5A) with the known gene expression data of a cancer type of the training expression data of multiple cancer types of the pan-cancer model of Graeber ([0263]), thus resulting in a pan-cancer model for predicting an SCN status of a tumor trained on multiple cancer types including glioma. Given Graeber discloses scoring glioma cells according to their SCN-gene expression score ([0196]; FIG. 5A), one of ordinary skill in the art would recognize the gene expression data of the glioma cells functions as an indicator of SCN status, which is the same function as the expression data of the cancer types in the pan-cancer training data. Furthermore, one of ordinary skill in the art would have recognized the results of the substitution would have been predictable given Graeber discloses training a machine learning model on expression data of multiple cancer types, as discussed above.
Regarding the dependent claims:
Regarding claim 31, Graeber discloses the cancer treatment may be used alone ([0063]), which demonstrates the therapy approach does not include the use of chemotherapy.
Regarding claim 32¸ Graeber discloses determining the gene expression levels’ are different than a to a small cell neuroendocrine (SCN) signature ([0020]; [0023]-[0024]; [0055]) by classifying the expression data of the patient using a classifier trained on a training set of SCN positive cases and a second subset of small cell lung cancer (SCLC) samples to assign the patient to the SCN class or non-SCN class, wherein the SCN and non-SCN classes are differentiated by expression levels of one or more genes ( [0263], e.g. gene expression profiles used)
Regarding claim 33, Graeber further discloses determining the SCN signature by training a logistic regression classifier using training data from multiple cancers ([0263]-[0265]), wherein the training data includes the following:
Graeber discloses the training data includes gene expression values of for each gene in the SCN signature ([0263]).
Graeber discloses the training data includes a set of samples as part of a SCN class (i.e. a first subset with a first label being indicative of a tumor having a neuronal property) and a set of samples as part of a non-SCN class (i.e. a second subset with a second label being indicative of a tumor not having the neuronal property) ([0265]).
While Graeber does not explicitly disclose the training gene expression data includes expression data from reference brain tumors, it would have been obvious to have included expression data from reference brain tumors for the same reasons applied above to claim 120.
Regarding claims 34-35, Graeber discloses the set of genes includes SV2A, NCAM1, ITGB6, SH2D3A, and TACSTD2 ([0163], e.g. genes in SCN gene signature; [0008]; [0017], e.g. biomarkers in Table 1 measured; claim 8).
Regarding claim 104, Graeber discloses the neuronal genetic signature comprises genes that are used to determine of a sample has a tumor with a SCN or neuroendocrine features (i.e. the first class of tumors ) or non-neuronal (the second class of tumors) [0055]. It is noted that the limitation regarding how the set of genes was previously identified is a product by process limitation, and step of identifying the set of genes as informative of assignment is not required by the claims. See MPEP 2113.
Regarding claim 107, Graeber discloses training the model includes identifying the set of genes [0053]-[0054], comprising the following.
Graeber discloses using a gene expression profile for each of a first tumor class with a label indicative of the tumor having a neuronal property (i.e. having the neuronal genetic signature) and a second non-SCN class (i.e. tumors not having the neuronal genetic signature) ([0263]; [0265]), wherein the gene expression profile includes, for each gene, an expression level (i.e. an expression-metric statistic) indicating a level (i.e. a degree) to which the gene is expressed for the given class ([0020]).
Graeber discloses each gene in the determined set of genes has a signature weight greater than a predefined threshold ([0013]-[0015]), wherein the signature weight reflects a difference between the expression levels of the respective gene between the two classes (i.e. between the first and second expression-metric statistic ([0184], e.g. weights determined by PCA between SCN cases and non-SCN tumors).
Regarding claims 108 Graeber discloses measuring the set of genes including SV2A, NCAM1, ITGB6, SH2D3A, TACSTD2, C19orf33, SFN, RND2, PHLDA3, OTX2, and TBC1D2 ([0163], e.g. genes in SCN gene signature; [0008]; [0017], e.g. biomarkers in Table 1 measured; claim 8), and that the genes are measured in a tumor sample ([0025]; claims 1 and 15)
Graeber further discloses using the measured expression levels are used to determine whether the subject has the SCN cancer ([0020]) and ultimately determine whether the patient would benefit from an immune checkpoint inhibitor ([0027]; [0065]), as shown by Graeber in view of Taube and Spetzler as applied to claim 30 above.
Regarding claims 109-116, Graeber discloses the set of genes measured includes SV2A, NCAM1, ITGB6, SH2D3A, TACSTD2, C19orf33, SFN, RND2, PHLDA3, OTX2, and TBC1D2 (i.e. ten or more genes in Tables 2, 3, and 4) ([0163], e.g. genes in SCN gene signature; [0008]; [0017], e.g. biomarkers in Table 1 measured; claim 8).
Therefore, the invention is prima facie obvious.
Response to Arguments
Applicant's arguments filed 01 Sept. 2026 regarding 35 U.S.C. 103 have been fully considered but they are not persuasive.
Applicant remarks, regarding claim 30, that Graeber, Spetzler, and Taube, alone and in combination, fail to teach or suggest “that the model comprises a plurality of learned parameters defining a hyperplane separating a brain tumor class and a non-neuronal, non-neuroendocrine tumor class”, and Graeber, Spetzler, and Taube are all silent to hyperplanes let alone separating the recited classes (Applicant’s remarks at pg. 9, para. 3-4). Applicant further remarks that Graebar’s relevant models are derived from data distinguishing LUAD from SCLC, and not from gene-expression data comprising reference brain tumors, and thus the hyperplane boundary would reflect Graeber’s disclosed SCN/non-SCN framework rather than a brain tumor class and non-neuronal, non neuroendocrine tumor class (Applicant’s remarks at pg. 9, para. 5 to pg. 10, para. 2).
This argument is not persuasive. While Graeber does not explicitly mention the word “hyperplane”, the machine learning model of Graeber is a logistic regression classifier, as explained in the above rejection, which inherently constructs a hyperplane as claimed. This is evidenced by Abdelhafez, which discloses logistic regression classifiers generate a linear hyperplane separating the two classes (pg. 5, para. 2-4).
Regarding the specifically recited two classes separated by the hyperplane, Graeber discloses the two classes are positive to SCN (i.e. neurally related) and negative to SCN (i.e. non-neuronal and non-neuroendocrine), which read on the claimed classes (see claim interpretation under 35 U.S.C. 112(b)). Further regarding the use of reference brain tumors in the training set, as discussed above in claim interpretation and the above rejection, the limitation only serves to define the process in which the machine-learning model of claim 30 was previously trained. MPEP 2113 I. states if the product in the product-by-process claim is the same as or obvious from a product of the prior art, the claim is unpatentable even though the prior product was made by a different process." In re Thorpe, 777 F.2d 695, 698, 227 USPQ 964, 966 (Fed. Cir. 1985) (citations omitted). Because the trained machine learning model of Graeber also comprises parameters learned to define a hyperplane between a tumor having a neuronal property class (i.e. interpreted to be a brain tumor class as discussed in the 112(b) rejection) and a non-neuronal, non-neuroendocrine tumor class, the machine learning model of Graeber is the same as in the instant claims, even if trained by data other than reference brain tumors.
Regardless, as applied to claim 120 above, it is further noted that it would have been prima facie obvious to have trained the model of Graeber on expression data from brain tumors, given the machine learning model of Graeber is a pan-cancer model trained on data from a plurality of cancer types, as explained in the above rejection.
Applicant remarks, regarding claim 119, that Graeber does not teach or suggest a set of genes consisting of two or more of the genes recited in claim 119 because Table 1 of Graeber includes hundreds of genes, and Graeber provides no teaching to select specifically two or more genes from the particular subset of genes in claim 119, and therefore, a person of ordinary skill in the art would not “at once envisage” the particular combinations recited in claim 119 (Applicant’s remarks at pg. 10, para. 3 to pg. 11, para. 1). Applicant remarks Sptezler and Taube do not cure this deficiency (Applicant’s remarks at pg. 11, para. 1).
This argument is not persuasive. Claim 119 recites “receiving gene-expression data comprising an expression metric for each of a set of genes measured in a sample collected from the subject, wherein the set of genes consists of two or more of SV2A…”. As explained above in claim interpretation, the “gene expression data” used by the trained machine learning model for classification is not closed to unrecited elements, and therefore may include additional gene sets other than the gene set consisting of two or more of the recited genes. See MPEP 2111.03 II. Because Graeber discloses any one or more biomarkers, including all biomarkers, from Table 1 may be measured and used for classification ([0008]; [0117]), this necessarily includes the gene set of SV2A, NCAM1, ITGB6, SH2D3A, and TACSTD2 ([0163]), which reads on the claim. In other words, the “gene-expression data” is not required to consist of an expression metric for each of a set of genes, as argued by Applicant.
Applicant remarks that Graeber, Spetzler, and Taube are silent to training a machine-learning model using gene-expression data from a plurality of reference brain tumors because Graeber teaches training a model on lung and prostate cancers (Graeber, [0172]) (pg. 11, para. 2-3). Applicant remarks that Graeber’s disclosure that its lung and prostate derived SCN signature may be applied to other tumor types does not suggest retraining a model using samples from each tumor type to which it is applied, and Graeber explicitly describes “pan-cancer metastatic SCNs” as being projected onto “the SCN space defined by lung and prostate” (Graeber [0049]), and thus Graeber teaches applying the lung- and prostate-derived framework to additional cancers rather than retraining the framework (Applicant’s remarks at pg. 110, para. 2-4). Applicant remarks Spetzler and Taube fail to cure the deficiencies of Graeber (pg. 12, para. 1).
This argument is not persuasive. First, the citations to Graeber at [0049] and [0172] is merely a principal component analysis performed on RNA-seq datasets from LUAD, CRPC-adeno, SCLC, and NEPC cancers, which is not a disclosure of training the machine learning model. Nor were these paragraphs relied upon for the training step.
As explained in the above rejection, Graeber discloses the machine learning model is a pan-cancer model that predicts samples with SCN features, which was trained on samples from a variety of cancers, including LUAD, SCLC, NEPC, CRPC, and the pan-cancer SCN positive cases confirmed by pathology (BRCA, CESC, COADREAD, ESCA, HNSC, LUAD, LUSC, PRAD, STAD, THCA ([0265]). Accordingly, the SCN signature is not specific to lung and prostate, and is instead is a pan-cancer SCN signature intended to predict SCN status in various cancer types. Given Graeber discloses the cancer of a subject being determined to be a SCN cancer or not could be glioma (i.e. a brain tumor) ([0016]; [0146]) and scoring glioma cells according to their SCN-gene expression score ([0196]; FIG. 5A), it would have been obvious to have included gene expression data from a brain tumor as one of the cancers used for training the pan-cancer model for the reasons based on the simple substation of expression data for a glioma sample in the training set of Graeber. Given Graeber discloses scoring glioma cells according to their SCN-gene expression score ([0196]; FIG. 5A), one of ordinary skill in the art would recognize the gene expression data of the glioma cells functions as an indicator of SCN status, which is the same function as the expression data of the cancer types in the pan-cancer training data. Furthermore, one of ordinary skill in the art would have recognized the results of the substitution would have been predictable given Graeber discloses training a machine learning model on expression data of multiple cancer types, as discussed above.
Citation of Pertinent Prior Art
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Jaegil et al., US 2020/0370133 A1, effectively filed 09 Feb. 2018.
Jaegil discloses a method for characterizing bladder cancer (Abstract), which includes characterizing a neuronal subtype of a sample of a subject using neuroendocrine and neuronal markers, wherein the neuronal markers are MSI1, PLEKHG4B, GNG4, PEG10, RND2, APLP1, SOX2, TUBB2B, TP53, RB1 and E2F3 and the neuroendocrine markers are CHGA, CHGB, SCG2, ENO2, SYP, NCAM1 (claim 1; [0031]; [0272]).
Conclusion
No claims are allowed.
Claims 30-35, 104, 107-116, and 119-120 are patent eligible for the reasons discussed in the Office action mailed 01 Dec. 2025 at para. [018]-[019].
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.
Inquiries
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KAITLYN L MINCHELLA whose telephone number is (571)272-6485. The examiner can normally be reached 7:00 - 4:00 M-Th.
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, Olivia Wise can be reached at (571) 272-2249. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/KAITLYN L MINCHELLA/Primary Examiner, Art Unit 1685
Prosecution Insights