DETAILED ACTION
Applicant’s response, filed 12 June 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 .
Claim Status
Claims 1-17 and 19-30 are pending and examined herein.
Claims 1-17 and 19-30 are rejected.
Priority
Claims 1-17 and 19-30 are granted the claim to the benefit of priority to U.S. Provisional applications 63/067748 and 63/203804 filed 19 August 2020 and 30 July 2021 respectively. Thus, the effective filling date of claims 1-17 and 19-30 is 19 August 2020.
Claim Rejections - 35 USC § 112
The rejection on the ground of 112/a of claims 1-6, 9-11, 14-17, 19-30 for claim 1 reciting “a predicted classification of the new subject’s event occurrence” in Office action mailed 23 February 2026 is withdrawn in view of the amendment of “wherein the new subjects event occurrence is a cancer recurrence or a response to treatment” received 12 June 2026.
The rejection on the ground of 112/b of claims 1-17 and 19-31 for claim 1 reciting “wherein the second dataset is inaccessible by the first entity” in Office action mailed 23 February 2026 is withdrawn in view of the amendment which removes this limitation received 12 June 2026.
The rejection on the ground of 112/b of claims 1-17 and 19-31 for claim 1 reciting “the new subject” in line 17 of the claim in Office action mailed 23 February 2026 is withdrawn in view of the amendment of “a new subject” received 12 June 2026.
The rejection on the ground of 112/b of claims 1-17 and 19-31 for claim 1 reciting “correcting the first dataset based on the adaptation factors” in Office action mailed 23 February 2026 is withdrawn in view of the amendment of “correcting the first dataset based on the first and second sets of adaptation factors” received 12 June 2026.
The rejection on the ground of 112/b of claims 21-25 for claims 21 and 25 reciting “the RNA expression data” in Office action mailed 23 February 2026 is withdrawn in view of the amendment of “wherein the first RNA expression data” received 12 June 2026.
The rejection on the ground of 112/b of claims 27-29 for claims 27 reciting “a training dataset of… both a subset of the first set of RNA expression data and a subset of the second set of RNA expression data…” and claim 28 reciting “wherein a training dataset of … a subset of a third set of RNA expression data…” in Office action mailed 23 February 2026 is withdrawn in view of the amendment of “a training dataset of… both a subset of the transformed first set of RNA expression data and a subset of the transformed second set of RNA expression data…” received 12 June 2026.
112/a New Matter
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
The rejection below was previously recited.
Claims 8 is rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
Claim 8 recites “wherein the classification of the new subject’s response to treatment comprises predicting a high-risk of a cancer type”. There is an inadequate written description for this limitation. The instant disclosure provides “an artificial intelligence engine which predicts a patient’s outcome to treatments” (instant disclosure [0232]), “the first entity may then apply their adapted dataset to the trained engine to predict patient outcomes” (instant disclosure [0232]), “survival curves provide estimates on the duration of time until an event of interest occurs for a patient, which is cancer recurrence in this experimental setup” (instant disclosure [0249]). However, there is no disclosure that the classification of the new subject’s response to treatment comprises a prediction of a high-risk of a cancer type. Thus, this limitation constitutes as new matter.
Response to Arguments
Applicant's arguments filed 12 June 2026 have been fully considered but they are not persuasive.
Applicant argues that the limitation of “wherein the classification of the new subject’s response to treatment comprises predicting a high-risk of a cancer type” is supported by the instant disclosure and points to [0231], [0232], and [0304]-[0305] (Reply p. 7-8).
This argument has been considered but found to be not persuasive. It is noted that “comprises predicting a high-risk of a cancer type” is interpreted as a diagnostic classification (i.e., the risk of the patient of having a particular cancer type”) while the paragraphs provided are in the context of a prognostic classification where particular cancer datasets are used to determine risks in prognostic classifications such as response to treatment. It is noted that the recited diagnostic classification is not supported by the instant disclosure. If applicant intends to have the claim limited to a prognostic classification supported by the instant disclosure an amendment of “wherein the classification of the new subject’s response to treatment comprises predicting a high-risk of the new subject’s response to treatment based on a cancer type”.
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.
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.
The rejection below was previously recited.
Claims 19 is 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 term “high-expression” in claim 19 is a relative term which renders the claim indefinite. The term “high-expression” 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 genes which are being referred to in the claims are rendered indefinite by the use of the term “high-expression” because it is unclear to what degree is a gene considered as a high-expression gene. The specification does not provide a clear and precise definition of the limitation, nor would one skilled in the art recognize the metes and bounds of said limitation. For the sake of furthering examination, the limitation of “high-expression genes” will be interpreted as “genes”.
Response to Arguments
Applicant's arguments filed 12 June 2026 have been fully considered but they are not persuasive.
Applicant states MPEP 2173.05(b) indicates relative terminology is definite if its meaning can be understood with “reasonable certainty” and argues that the term “high-expression” or notion of a highly expressed gene is one that is commonly used and whose meaning is commonly understood by those familiar with RNA transcription, thereby satisfying this requirement for reasonable certainty (Reply p. 8).
This argument has been fully considered but found to be not persuasive. It is noted that although the term “high-expression genes” are utilized in the art, the standard for distinguishing a high-expression gene may vary due to the identification high-expression genes being dependent on datasets used with no explicit standard. Thus, it is unclear to what standard is used for ascertaining the requisite degree genes are considered high-expression genes (such as a threshold value applied to a particular gene expression value). The specification does not provide a clear and precise definition of the limitation, nor would one skilled in the art recognize the metes and bounds of said limitation (e.g., what standard is encompassed for distinguishing a high-expression gene). Thus, the metes and bounds of the claims are indefinite by the use of this term.
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.
The rejection below has been modified necessitated by amendments.
Claims 1-17 and 19-30 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
(Step 1)
Claims 1-17 and 19-30 fall under the statutory category of a process.
(Step 2A Prong 1)
Under the BRI, the instant claims recite judicial exceptions that are abstract ideas of the type that is in the grouping of a “mathematical concept”, such as mathematical relationships and mathematical equations. The instant claims also recite judicial exceptions that is in the grouping of a “law of nature” such as a naturally occurring correlation.
Independent claim 1 recites mathematical concepts and mental processes of transforming the first set of RNA expression data based at least in part on characteristics of a second dataset having a second set of RNA expression data, the transforming comprising reducing a dimensionality of the first and second datasets to generate first and second reduced datasets, generating a first set of adaptation factors related to the first reduced dataset, generating a second set of adaptation factors related to the second reduced dataset, and correcting the first dataset based on the first and second sets of adaptation factors, wherein at least one of the first set of sequencing equipment or methods used to generate the first set of RNA expression data from the first set of sequencing equipment introduce a dataset specific nature into the first set of RNA expression data that is not represented in the second dataset.
Independent claim 1 recites a mental process and law of nature (as a naturally occurring correlation) of generating based on the RNA expression data of the record a predicted classification of the new subject’s event occurrence, wherein the new subject’s event occurrence is a cancer recurrence or a response to treatment.
Dependent claim 7 recites mental processes and laws of nature of predicting a subject’s outcome to treatments based at least in part on the prediction of the new subject’s risk of cancer recurrence or predicting the new subject’s risk of cancer recurrence based at least in part on the classification of the new subject’s response to treatment. Dependent claim 10 recites mathematical concepts and mental processes of generating a transform between the first set of adaptation factors and the second set of adaptation factors, encoding two or more corresponding eigenvalues of the second dataset with the generated transform, and providing the encoded two or more corresponding eigenvalues of the second dataset as the adapted second data set. Dependent claim 12 recites mathematical concepts and mental processes of transforming RNA expression data of the second record based at least in part on characteristics of the first set of RNA expression data and generating a predicted classification of the second subject’s event occurrence.
The claims recite mental processes of “transforming the first set of RNA expression data based at least in part on characteristics of a second dataset… the transforming comprising reducing a dimensionality of the first and second datasets to generate first and second reduced datasets, generating a first set of adaptation factors related to the first reduced dataset, generating a second set of adaptation factors related to the second reduced dataset, and correcting the first dataset based on the first and second sets of adaptation factors, wherein at least one of…” (which may encompasses a judgment that filters out data based on a criteria such as expression data known to not be informative for the classification of a subjects cancer recurrence or response to treatment, calculating eigenvectors, and performing a linear transformation on numerical values), “generating based on the RNA expression data of the record a predicted classification of the new subject’s event occurrence, wherein the new subject’s event occurrence is…” (which encompasses analyzing RNA expression data of a patient to determine a classification of a subjects event occurrence), “predicting a subject’s outcome to treatments based at least in part on…or predicting the new subject’s risk of cancer recurrence based at least in part on the classification of…” (which encompasses making a determination based on classification results), generating a transform between the first set of adaptation factors and the second set of adaptation factors (which encompasses generating a linear transformation which maps numerical values between vector spaces defined by eigenvectors), encoding two or more corresponding eigenvalues of the second dataset with the generated transform, and providing the encoded two or more corresponding eigenvalues of the second dataset as the adapted second data set (which encompasses performing a linear transformation which is a mathematical calculation). The human mind is capable of performing the recited steps above because the human mind can analyze numerical data utilizing mathematical calculations and make judgments based on expression data.
The claims recite mathematical concepts that are mathematical calculations. The MPEP states at 2106.04(a)(2)(I)(C) that “There is no particular word or set of words that indicates a claim recites a mathematical calculation. That is, a claim does not have to recite the word "calculating" in order to be considered a mathematical calculation. For example, a step of "determining" a variable or number using mathematical methods or "performing" a mathematical operation may also be considered mathematical calculations when the broadest reasonable interpretation of the claim in light of the specification encompasses a mathematical calculation”.
The limitations of transforming a RNA expression dataset based on characteristics of a different RNA expression dataset by performing dimensionality reduction, generating adaptation factors, and correcting a RNA expression dataset based on the adaptation factors are mathematical concepts because these steps encompass mathematical calculations. The instant disclosure provides that dimensionality reduction is performed by principal component analysis which is a series of mathematical calculations to generate principal components (i.e., eigenvectors/ adaptation factors of the covariance matrix of the RNA expression datasets covariance matrix) and an encoded values from the dataset (i.e., reduced RNA expression datasets) (see instant disclosure [0127] and [0134]). The instant disclosure provides that correcting a RNA expression dataset based on the adaptation factors is performed by performing a linear transformation) which is a series of mathematical calculations (see instant disclosure [0088], [0098], [0109]-[0111], [0128] and [0135]).
Dependent claims 2-6, 8, 9, 11, 13-17, and 19-30 further limit the mental process/mathematical concept recited in the independent claim but do not change their nature as a mental process/mathematical concept. Claims 2-6, 9, 11, 14-17, 19-30 further recite limitations about the abstract data that is being processed by the judicial exceptions. It is noted that further limiting abstract data which is processed by mathematical calculations does not change the nature of the mathematical concepts. Further, claims 7, 8, 12, 13, and 30 recite a law of nature for reciting the event occurrence is a risk of cancer recurrence or a response to treatment which is a recitation of a naturally occurring correlation between a subject’s RNA expression and cancer recurrence or response to a treatment.
(Step 2A Prong 2)
Claims found to recite a judicial exception under Step 2A, Prong 1 are then further analyzed to determine if the claims as a whole integrate the recited judicial exception into a practical application or not (Step 2A, Prong 2). Integration into a practical application is evaluated by identifying whether there are any additional elements recited in the claim and evaluating those additional elements to determine whether they integrate the exception into a practical application.
The additional element in claims 1 of receiving a first dataset and receiving a record and the additional element in claim 12 of receiving a second record does not integrate the judicial exceptions into a practical application because these steps are adding insignificant extra solution of data gathering (see MPEP 2106.05(g)). It is noted that the limitations which limit the content of the data being received in claims 1-6, 9, 11, 12, 14-26, and 30 falls under the abstract idea and the content does not change the active step of receiving data in a computer environment.
The additional element in claim 1 of a generic computer, generating a machine learning molecular model, providing data to a machine learning molecular model, and an artificial intelligence engine in the computer system implementing the generated machine learning model to perform judicial exceptions and the additional elements in claim 12 of providing data to the generated machine learning molecular model and an artificial intelligence engine in the computer system implementing the generated machine learning model to perform judicial exceptions does not integrate the judicial exception into a practical application because this mere instructions to apply the judicial exception to a generic computer without an improvement to computer technology (see MPEP 2106.04(d)(1), example 47, and MPEP 2106.05(f)). The limitations of generating the machine learning model (e.g., the model is generated to make the abstract predictions of predicting an event occurrence of a subject using data) and using the machine learning molecular model provide nothing more than mere instructions to implement the abstract idea on a computer because the claims recite only the ideas of a solution or outcome (i.e., the claims fail to recite details of how predicting an event occurrence of a subject is accomplished or how generating the machine learning model to predict an event occurrence of a subject is accomplished). It is noted that the content of the data (i.e., transformed RNA expression data) and the predicted classification of a subject’s event occurrence based on the RNA expression data fall under the abstract idea of a mental process because the human mind is capable of performing a classification on a subject’s event occurrence based on a patient’s RNA expression data.
Thus, the additional elements do not integrate the judicial exceptions into a practical application and claims 1-17 and 19-30 are directed to the abstract idea.
(Step 2B)
Claims found to be directed to a judicial exception are then further evaluated to determine if the claims recite an inventive concept that provides significantly more than the judicial exception itself (Step 2B). The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception because:
The additional element in claims 1 of receiving a first dataset and receiving a record and the additional element in claim 12 of receiving a second record is conventional see MPEP 2106.05(b) and 2106.05(d)(II). It is noted that the limitations which limit the content of the data being received in claims 1-6, 9, 11, 12, 14-26, and 30 falls under the abstract idea and the content does not change the active step of receiving data in a computer environment.
The additional element in claim 1 of a generic computer, generating a machine learning molecular model, providing data to a machine learning molecular model, and an artificial intelligence engine in the computer system implementing the generated machine learning model to perform judicial exceptions and the additional elements in claim 12 of providing data to the generated machine learning molecular model and an artificial intelligence engine in the computer system implementing the generated machine learning model to perform judicial exceptions amounts to mere instructions to apply the judicial exceptions to a generic computer which is conventional (see MPEP 2106.05(b) and 2106.05(d)(II)).
The combination of receiving data and using a generic computer to perform judicial exceptions is conventional see MPEP 2106.05(b) and 2106.05(d)(II).
Thus, the additional elements are not sufficient to amount to significantly more than the judicial exception because they are conventional see MPEP 2106.05(b) and 2106.05(d)(II).
Response to Arguments
Applicant's arguments filed 12 June 2026 have been fully considered but they are not persuasive.
Argument 1 (Step 2A, Prong 1):
Applicant argues that reducing a dimensionality of first and second datasets to generate first and second reduced datasets, generating first and second sets of adaptation factors related to the first and second reduced datasets, and correcting a first dataset based on the first and second sets of adaptation factors, like the training using a backpropagation algorithm or gradient descent algorithm in Example 47, is not something that can be practically performed in the mind. See
Office action at 9-10. Instead, consistent with the court's discussion in Synopsis, it represents a "manipulation of data that... could not conceivably be performed in the human mind or with pencil and paper." Synopsis v. Mentor Graphics, 839 F.3d at 1148 (Reply p. 11).
This argument has been fully considered but found to be not persuasive. The claims recite mental processes of performing a transformation on RNA expression data by reducing the dimensionality of RNA expression datasets, generating adaptation factors, correcting the RNA expression datasets based on the adaptation factors, and generating a classification of a subjects event occurrence based on the RNA expression data, wherein the new subjects event occurrence is a cancer recurrence or response to treatment. The claimed steps are recited to encompass any method for achieving reducing a dimensionality of first and second datasets to generate first and second reduced datasets, generating first and second sets of adaptation factors related to the first and second reduced datasets, and correcting a first dataset based on the first and second sets of adaptation factors. The human mind is capable of analyzing RNA expression data in a manner which results in reducing a dimensionality of RNA expression datasets (which may encompasses a judgment that filters out data based on a criteria such as expression data known to not be informative for the classification of a subjects cancer recurrence or response to treatment), generating adaptation factors (which encompasses calculating eigenvectors and the human mind is capable of calculating eigenvectors), and correcting RNA expression data (which encompasses performing a linear transformation on numerical values and the human mind is able to perform a linear transformation on numerical values). The claims do not recite limitations in the process of transforming the RNA expression data which precludes the human mind from performing. Thus, the claims recite mental processes.
Argument 2 (Step 2A, Prong 1):
Applicant argues a machine learning molecular model is not a mental construct and, as such, the steps of "generating" that model and generating a predicted classification of a new subject's event occurrence "by an artificial intelligence engine in the first computer system implementing [a] generated machine learning molecular model" are not ones that can "practically be performed in the human mind," as demonstrated by the Office's own eligibility examples. For example, referring again to claim 2 in Example 47, the Office considers the limitation "using the trained ANN" to represent an additional element, separate from any alleged mental process (Reply p. 12).
This argument has been fully considered and it is noted that in view of claim 2 and Example 47 the limitations of generating a machine learning model, providing data to the machine learning model, and using the generated machine learning model are interpreted as being mere instructions to apply the judicial exception to a generic computer (see MPEP 2106.05(f) and Example 47) due to the claims only reciting the ideas of a solution or outcome (i.e., the claims fail to recite details of how predicting an event occurrence of a subject is accomplished or how generating the machine learning model to predict an event occurrence of a subject is accomplished). It is noted that the limitations of “transforming the first set of RNA expression data…” and generating a predicted classification of the new subject’s event occurrence, wherein the new subject’s event occurrence is a cancer recurrence or a response to treatment fall under the abstract idea of a mental process and mathematical concept.
Argument 3 (Step 2A, Prong 1):
Applicant argues that in formulating the rejection which identifies steps of reciting mathematical concepts, the Office improperly reads the processes of principal component analysis, gradient descent optimization, minimizing the Euclidean distance between factors of different data sets, linear regression, and logistic regression into the claims. These elements are not recited in claim 1. In this regard, Example 48, claim 2 governs the proper analysis. In that case, the Office instructs: Even though the disclosure explains that stitching could be performed by an overlap-add method, which is a mathematical operation, the claim recites no details of how the stitching is performed. Additionally, while the claim recites variables, variables on their own are not mathematical relationships, formulas, or calculations. Therefore, the combining step is merely based on or involves a mathematical concept but does not recite a mathematical concept (Reply p. 13).
This argument has been fully considered but found to be not persuasive. It is noted that Example 48, claim 2 has a different fact pattern than the instant case because example 48 claim 2 is a process of processing speech signals, while the instant case is processing RNA expression which in light of the specification is represented as numerical values arranged in a matrix (see instant disclosure [0090], [0098], and [0130]). As described in further detail below, each step of transforming the RNA expression data encompasses mathematical calculations and the entire limitation of performing this transformation encompasses an embodiment which is purely mathematical. It is further noted the limitations described below are not read into the claims to improperly limit the claims but to provide what each step encompasses in light of the specification and to show these steps themselves encompass embodiments which are purely mathematical calculations.
The MPEP states at 2106.04(a)(2)(I)(C) that “There is no particular word or set of words that indicates a claim recites a mathematical calculation. That is, a claim does not have to recite the word "calculating" in order to be considered a mathematical calculation. For example, a step of "determining" a variable or number using mathematical methods or "performing" a mathematical operation may also be considered mathematical calculations when the broadest reasonable interpretation of the claim in light of the specification encompasses a mathematical calculation”.
The instant claims recite transforming the first set of RNA expression data and describes that this transformation is achieved through the steps of reducing a dimensionality of the first and second datasets to generate first and second reduced datasets (the BRI of this step of dimensionality reduction in light of the specification encompasses performing PCA which is a series of mathematical calculations that performs a linear transformation on numerical data to represent the numerical data in a reduced form), generating a first set of adaptation factors related to the first reduced data set (the BRI of this step of generating a first set of adaptation factors in light of the specification encompasses generating eigenvectors of the reduced data set and the process of generating eigenvectors of a numerical dataset is a mathematical calculation see instant disclosure [0014]), generating a second set of adaptation factors related to the second reduced data set (the BRI of this step of generating a first set of adaptation factors in light of the specification encompasses generating eigenvectors of the reduced data set and the process of generating eigenvectors of a numerical dataset is a mathematical calculation see instant disclosure [0014]), and correcting the first dataset based on the first and second sets of adaptation factors (the BRI of this step of correcting the first dataset based on the first and second sets of adaptation factors in light of the specification encompasses performing a linear transformation which is a mathematical calculation which maps numerical data from one vector space defined by eigenvectors/adaptation factors of a first dataset to the vector space defined by eigenvectors/adaptation factors of a second data set see instant disclosure [0098] and [0109]-[0111]). Thus, each step which defines the process of transforming the first set of RNA expression data based at least in part on characteristics of a second dataset having a second set of RNA expression data in the claim encompasses mathematical calculations in light of the specification and this limitation transforming the first set of RNA expression data as a whole encompasses an embodiment which is purely mathematical calculations.
Argument 4 (Step 2A, Prong 1):
Applicant argues that the claims do not recite and are not directed to any correlation between natural phenomena and are not focused on merely observing or detecting a natural phenomenon, let alone without reciting any other meaningful, non-routine steps. Instead, Applicant's claims are focused on addressing shortcomings in the use of machine learning models due to biases, domain shifts, target shifts, etc., in the machines and methods used to generate the inputs for those machine learning models. See Specification at [0003]-[0013]. The steps of Applicant's claims that carry out those shortcomings, including the transformation of sets of expression data by reducing a dimensionality of those data sets, generating adaptation factors related to those data sets, and correcting a first data set based on those adaptation factors are just the sort of meaningful, non-routine steps (Reply p. 14-15).
This argument has been fully considered but found to be not persuasive. The claims recite a natural occurring correlation because the claims recite a correlation between RNA expression data (biomarkers) and a classification of a subject’s cancer recurrence or response to treatment. Applicant states that the MPEP distinguishes between claims that “describe a natural ability or quality of a product, or describe a natural process” and those that “recite a law of nature or natural phenomenon” indicating that the difference is the former are eligible because they “are not focused on merely observing or detecting” any such correlation or phenomenon. It is noted the instant claims do not describe a natural ability or quality of a product or describe a natural process. The instant claims have been identified as detecting a naturally occurring correlation between RNA expression data and a patients classification of cancer recurrence or response to treatment. Although the claims recite additional judicial exceptions of analyzing the data (i.e., transforming the RNA expression data) to observe or detect the naturally occurring correlation, this does not change the nature of the naturally occurring correlation recited in the claims of correlating RNA expression data (biomarkers) with a classification of a subjects cancer recurrence or response to treatment.
Argument 1 (Step 2A, Prong 2 and Step 2B):
Applicant argues the claims are directed to the problem that a machine learning model trained on one of those datasets - one of the additional elements in the claims - produces inaccurate results when implemented using the other dataset. Applicant argues contrary to the allegation in the Office action, this is not merely a "problem in abstract data analysis itself when making predictions using multiple datasets with different biases," but, instead, is one rooted in the way in which the machine learning model that implements such data is trained. Office action at 23. In this regard, the recitation of a machine learning molecular model and an artificial intelligence engine are not merely the application of generic computer components but instead are integral pieces of the problem to be solved. Applicant further argues that machine learning model problem is a technical problem, and Applicant's claimed solution to that problem is an improvement to the functioning of a computer (Reply p. 15-16).
This argument has been fully considered but found to be not persuasive. As discussed in the rejection above, the additional elements of “generating a molecular model” to perform the judicial exception of predicting an event occurrence of a subject and “providing data to the machine learning model… using the generated machine learning model” to perform the judicial exception of predicting an event occurrence of a subject constitutes as mere instructions to apply the judicial exception of predicting a subjects event occurrence based on RNA expression data to a generic computer because the claims fail to recite the details of how predicting an event occurrence of a subject is accomplished or how generating the machine learning model to predict an event occurrence of a subject is accomplished (see MPEP 2106.05(f) and Example 47 claim 2). Further, the argued improvement which addresses inaccurate results when implementing a model to make predictions on datasets with different biases, domain shifts, covariate shifts, or other dataset specific phenomenon is provided solely by the abstract data transformation (i.e., the judicial exception alone) which is applied to the generically recite machine learning model to perform the judicial exception of making a prediction of a subjects event occurrence. Thus, the argued improvement that the machine learning model provides better accuracy when making the prediction of a subjects event occurrence comes solely from the use of corrected RNA expression data produced using mathematical transformations (i.e., better data is being applied to the generically recited machine learning model). The claims do not provide details of how the model itself is generated to make this abstract prediction or details of how the model itself accomplishes this abstract prediction in manner which provides that the model itself is responsible for the improvement in accuracy. Thus, the argued improvement is interpreted as being provided by the judicial exception alone of generating an improved RNA dataset which is generically applied to a machine learning model/computer.
The fact pattern of the instant case is different than the fact pattern presented in Ex Parte Desjardins (hereinafter Desjardins). The claims at issue in Desjardins provided details of how the machine learning model itself operates which was found to solve the technical problem of “catastrophic forgetting” in machine learning models. In contrast, the instant claims do not provide details of how the machine learning model operates or how the machine learning model is generated, rather the instant claims provide details of generating an improved RNA expression dataset (which falls under the abstract idea) which is then applied to the generically recited machine learning model to perform the judicial exception of generating the prediction of a subjects event occurrence.
Argument 2 (Step 2A, Prong 2 and Step 2B):
Applicant argues the MPEP recognizes "[c]omponents or methods, such as measurement devices or techniques, that generate new data" as supporting the conclusion that a claim is directed to improvements to existing technology. MPEP § 2106.05(a)(II)(vi) (citing Electric Power Group, LLC v. Alstom, S.A., 830 F.3d 1350, 1355 (Fed. Cir. 2016)). Here, the recited transformation of the first set of RNA expression, which involves reducing a dimensionality of first and second datasets, generating first and second adaptation factors, and subsequently correcting the first dataset based on those adaptation factors, which are all done to address the shortcomings in machine learning models and in the relevant sampling technology, are precisely the types of methods or techniques used to generate new data that that section contemplates (Reply p. 16-17).
This argument has been fully considered but found to not be persuasive. The instant claims do not provide components or methods, such as measurement devices or techniques, that generate new data, rather the instant claims provide limitations of performing a series of mathematical transformations on data produced using known measurement devices and known techniques to make a prediction of a classification of a subjects cancer recurrence or response to treatment. Unlike the provided example of “measurement devices or techniques” which would be the devices or processes used to collect the data, the mathematical transformations in the instant claims fall under how the data already collected is processed during an analysis (e.g. the claims do not recite measurement devices or techniques which generate new data but rather provide how the data collected is processed in an analysis pipeline).
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.
The rejection below is newly recited. It is noted that Korsunsky et al. (Nat Methods 16, 1289–1296 (2019); newly recited) provides a Methods section at the end of the paper with no page numbers. There is 5 pages of this Methods section (each page in this section is referred to as page 1 of Methods, page 2 of Methods, etc.).
The rejection below has been modified necessitated by amendment.
Claims 1-17 and 19-30 are rejected under 35 U.S.C. 103 as being unpatentable over Sakellaropoulos et al. (Cell Reports, Volume 29, Issue 11, 3367 - 3373.e4 (2019); newly recited) in view of Korsunsky et al. (Nat Methods 16, 1289–1296 (2019); newly recited).
Claim 1 is directed to receiving, at a first computer system associated with a first entity, a first dataset having a first set of RNA expression data generated from a first set of sequencing equipment,
Sakellaropoulos et al. shows receiving gene expression of GDSC cell lines form the Array Express repository which was generated from a first set of sequencing equipment (Sakellaropoulos et al. page e1 section “Gene expression cell line and patient datasets”).
transforming the first set of RNA expression data based at least in part on characteristics of a second dataset having a second set of RNA expression data generated from a second set of sequencing equipment,
Sakellaropoulos et al. shows performing batch correction between GDSC cell line data and clinical datasets to correct for batch effects (Sakellaropoulos et al. page e2). Sakellaropoulos et al. shows the GDSC cell line data and clinical datasets are gathered from different sequencing equipment such as microarray sequencing equipment and Next generation RNA-seq sequencing equipment (Sakellaropoulos et al. page e1-e2).
generating a molecular model trained from the transformed first set of RNA expression, receiving a record associated with the new subject the record having RNA expression data generated from the second set of sequencing equipment, providing the RNA expression data of the record to the generated model and generating a predicted classification of the new subject’s risk of event occurrence, wherein the new subject’s event occurrence is a cancer recurrence or a response to treatment.
Sakellaropoulos et al. shows training machine learning models Deep Neural Network only on the cell line gene expression data (after batch correction between the cell line and clinical datasets) to determine the machine learning models optimal models (Sakellaropoulos et al. page 3368 right col. and page e2). Sakellaropoulos et al. shows the optimal models were utilized to predict z-score normalized IC50 values for each patient in the clinical dataset from the patient’s gene expression data which is interpreted as an event occurrence of a response to treatment (Sakellaropoulos et al. page e2).
Sakellaropoulos et al. does not show the transforming comprising reducing a dimensionality of the first and second datasets to generate first and second reduced datasets, generating a first set of adaptation factors related to the first reduced dataset, generating a second set of adaptation factors related to the second reduced dataset, and correcting the first dataset based on the first and second sets of adaptation factors, wherein at least one of the first set of sequencing equipment or methods used to generate the first set of RNA expression data from the first set of sequencing equipment introduce a dataset specific nature into the first set of RNA expression data that is not represented in the second dataset,
Like Sakellaropoulos et al., Korsunsky et al. shows a batch correction method for RNA expression datasets. Korsunsky et al. shows a process of performing batch correction by using principal component analysis (PCA) to embed cells (represented by gene expression data) into a space with reduced dimensionality (Korsunsky et al. page 1290 Figure 1). Korsunsky et al. shows that the process uses eigenvalue-scaled eigenvectors produced by the principal component analysis (which is interpreted as adaptation factors) as the low-dimensional embedding for batch correction (Korsunsky et al. “Harmony” section page 1 left col. of Methods). Korsunsky et al. shows correcting datasets (i.e. batch correction process) which is based on the low-dimensional embedding (eigenvalue-scaled eigenvectors/the adaptation factors) (Korsunsky et al. page 1290 Figure 1, “Harmony” section page 1 left col. of Methods, and “Linear mixture model correction” section on pages 3-4 of Methods). Korsunsky et al. the batch correction process is utilized for batch correction between different sequencing equipment (“Analysis details” section page 4 right col. of Methods).
Dependent claim 2 is directed to wherein the first sequencing equipment comprises microarray sequencing equipment and the second sequencing equipment comprise next generation sequencing equipment. Dependent claim 3 is directed to wherein the first sequencing equipment comprises next generation sequencing equipment and the second sequencing equipment comprise microarray sequencing equipment.
Sakellaropoulos et al. in view of Korsunsky et al. shows utilizing data produced by microarray sequencing equipment and next generation sequencing equipment (Sakellaropoulos et al. page e1-e2). Further, Sakellaropoulos et al. in view of Korsunsky et al. shows the ability of performing batch correction between different sequencing equipment technology (Korsunsky et al. page 4 right col. of Methods).
Dependent claim 4 is directed to wherein the first dataset is a public dataset and the second dataset is a laboratory specific dataset. Dependent claim 5 is directed to wherein the first dataset is a laboratory-specific dataset and the second dataset is a public dataset. Dependent claim 6 is directed to wherein the first dataset is a laboratory-specific dataset and the second dataset is a second laboratory specific dataset.
Sakellaropoulos et al. in view of Korsunsky et al. shows the use of RNA expression datasets produced from different laboratories (Sakellaropoulos et al. page e1 section “Gene expression cell line and patient datasets”) and shows the use of RNA expression datasets from public databases (Sakellaropoulos et al. page e1 section “Gene expression cell line and patient datasets”). Sakellaropoulos et al. in view of Korsunsky et al. shows that the data from the different labs are also public data from databases (Sakellaropoulos et al. page e1 section “Gene expression cell line and patient datasets”).
Dependent claim 9 is directed to wherein the first dataset is a germline dataset from a laboratory in a first location and the second dataset is a germline dataset from a laboratory in a second location. Dependent claim 11 is directed to wherein the record further comprises DNA mutation data generated from the second set of sequencing equipment. Claim 26 is directed to wherein the record further comprises pathology imaging features from a pathology slide.
Sakellaropoulos et al. in view of Korsunsky et al. shows gathering TCGA datasets from a reference which includes DNA variants from datasets such as germline variants (Sakellaropoulos et al. page e2 section TCGA dataset). Sakellaropoulos et al. in view of Korsunsky et al. shows a OCCAMS dataset was gathered for prediction which includes corresponding histological response data (Sakellaropoulos et al. page e2).
Claim 10 is directed wherein correcting the first dataset based on the adaptation factors comprises generating a transform between the first set of adaptation factors and the second set of adaptation factors, encoding two or more eigenvalues of the second dataset with the generated transform, and providing the eigenvalues of the second dataset as the adapted second dataset.
Sakellaropoulos et al. in view of Korsunsky et al. shows generating a transform between the first set of adaptation factors and the second set of adaptation factors by generating cell-specific correction values which is a linear combination of dataset correction factors weighted by the cell’s soft cluster assignments made and further shows a process of reference mapping where query datasets can be mapped onto a reference dataset (Korsunsky et al. page 1290 Figure 1 and page 3 right col of Methods). Sakellaropoulos et al. in view of Korsunsky et al. shows cells are embedded into a low-dimensional space as a result of PCA analysis by performing PCA on high-dimensional gene expression matrix and use the eigenvalue-scaled eigenvectors as the low dimensional embedding input for batch correction (Korsunsky et al. page 1 left col. of Methods). Sakellaropoulos et al. in view of Korsunsky et al. shows multiplying the cell embeddings by the eigenvalues to avoid given eigenvectors equal variance (i.e., eigenvalue-scaled eigenvectors) (Korsunsky et al. page 4 right col. of Methods).
Dependent claim 12 is directed to receiving a record form a second new patient with RNA expression data generated from a third set of sequencing equipment, transforming the RNA expression data and providing the transformed RNA expression data to the model to predict classification of the second new subjects event occurrence.
Sakellaropoulos et al. in view of Korsunsky et al. shows the optimal models were utilized to predict z-score normalized IC50 values for each patient in the clinical dataset from the patient’s gene expression data which is interpreted as an event occurrence of a response to treatment (Sakellaropoulos et al. page e2). This shows that the steps are repeated for each patient in the clinical dataset to predict the response to treatment
Dependent claim 14 is directed to wherein the first set of sequencing equipment and the second set of sequencing equipment sequences up to 140,000 transcripts. Dependent claim 15 is directed to wherein the first set of sequencing equipment and the second set of sequencing equipment sequences between 10 genes and 20,000 genes.
The BRI of these claims do not recite an active step of sequencing up to 140,000 transcripts or sequencing between 10 and 20,000 genes and is a product by process limitation of how the sequencing equipment produces the first and second datasets. Sakellaropoulos et al. in view of Korsunsky et al. shows the datasets include transcripts and sequences from 16,445 genes (Sakellaropoulos et al. page 3369 Figure 2).
Dependent claim 16 is directed to wherein characteristics of first set of sequencing equipment sequences have differences from the second set of sequencing equipment sequences. Dependent claim 17 is directed to wherein the characteristics are measured by variance of each gene or by heterogeneity. Dependent claim 19 is directed to wherein the characteristics are measured across high-expression genes. Dependent claim 20 is directed to wherein the characteristics occur in at least one of the FGFR2, MAP3K1, TNRC9, BRCA1, and BRCA 2 genes.
The BRI of these claims do not recite an active step of measuring variance of each gene, measuring heterogeneity, or measuring across high-expression genes, they are descriptive of how the underlying characteristics differences can be identified but are not active steps of the method or limitations to what the characteristics are (just how they may be measured). Sakellaropoulos et al. in view of Korsunsky et al. shows that the GDSC cell line data and clinical datasets are gathered from different sequencing equipment such as microarray sequencing equipment and Next generation RNA-seq sequencing equipment (Sakellaropoulos et al. page e1-e2). Sakellaropoulos et al. in view of Korsunsky et al. further shows variation in the ERK MAPK signaling which includes the MAP3K1 gene (Sakellaropoulos et al. page 3369 Figure 2). It is interpreted that due to the difference in the technological equipment used to gather the data the characteristic between sequences will be different and will occur in all genes sequenced.
Dependent claim 21 is directed to wherein the RNA expression data is from gene expression across different cancer types. Dependent claim 25 is directed to wherein the RNA expression data is from a tumor cell.
Sakellaropoulos et al. in view of Korsunsky et al. shows utilizing RNA expression data from cancer-cell lines acquired from the Genomics of Drug Sensitivity in Cancer (GDSC) database which includes cancer cell-lines across different cancer types (Sakellaropoulos et al. page 3368 and page 3369 Figure 2A).
Dependent claim 22 is directed wherein the different cancer types comprise at least two of breast, colorectal, pancreatic, lung, and bladder cancers. Dependent claim 23 is directed to wherein the different cancer types comprise at least two of squamous and immunogenic. Dependent claim 24 is directed to wherein the different cancer types comprise adenocarcinoma and neuroendocrine cancers.
Sakellaropoulos et al. in view of Korsunsky et al. shows that the different cancer types include breast, bladder, lung, and pancreatic (Sakellaropoulos et al. page 3369 Figure 2A). Sakellaropoulos et al. in view of Korsunsky et al. further shows the different cancer types also include Lung NSCLC squamous cell carcinoma and Melanoma (which is interpreted as an immunogenic cancer) (Sakellaropoulos et al. page 3369 Figure 2A). Sakellaropoulos et al. in view of Korsunsky et al. further shows the different cancer types also include Lung NSCLC adenocarcinoma and Lung small cell carcinoma (which is a type of neuroendocrine cancer) (Sakellaropoulos et al. page 3369 Figure 2A).
Dependent claims 27 and 28 are directed to using subsets of the transformed RNA expression data for training. Dependent claim 29 wherein the training dataset excludes gene variants which are not informative to the machine molecular model.
Sakellaropoulos et al. in view of Korsunsky et al. shows using 5-fold cross validation for training the machine learning model which splits the RNA expression data into subsets for training (Sakellaropoulos et al. page e2). Sakellaropoulos et al. in view of Korsunsky et al. shows selecting highly variable genes for training the machine learning model which is interpreted as excluding gene variants that are not informative ((Sakellaropoulos et al. page e2).
Dependent claim 30 is directed to wherein the dataset specific nature reflects a bias.
Sakellaropoulos et al. in view of Korsunsky et al. shows batch performing batch correction which corrects for batch effects which are differences between characteristics in datasets which is a bias (Sakellaropoulos et al. page e2 and Korsunsky et al. page 1 left col. “Methods”).
Dependent claim 7 is directed to predicting a subject’s outcome to treatments based at least in part on the prediction of the new subject’s risk of cancer recurrence or predicting the new subject’s risk of cancer recurrence based at least in part on the classification of the new subject’s response to treatment. Dependent claim 8 is directed to wherein the prediction of the new subject’s risk of cancer recurrence is based at least in part on a cancer type or cancer subtype, or wherein the classification of the new subject’s response to treatment comprises predicting a high-risk of a cancer type. Dependent claim 13 is directed to wherein the classification of the new subject’s risk of cancer recurrence comprises an estimate on a duration of time until the new subject’s risk of cancer recurrence occurs or wherein the classification of the new subject’s response to treatment further comprises an estimate on a duration of time until the new subject’s response to treatment occurs.
Sakellaropoulos et al. in view of Korsunsky et al. shows the optimal models were utilized to predict z-score normalized IC50 values for each patient in the clinical dataset from the patient’s gene expression data which is interpreted as an event occurrence of a response to treatment (Sakellaropoulos et al. page e2). Sakellaropoulos et al. in view of Korsunsky et al. performing Kaplan-Meyer survival analysis to contrast the groups of the lowest and highest IC50 (Sakellaropoulos et al. page e3). Sakellaropoulos et al. in view of Korsunsky et al. shows Kaplan-Meyer survival plots for specific drugs in different cohorts with time under treatment in days (for Bortezomib, Paclitaxel, and Cisplatin-TCGA) and weeks (for PARP-Inhibitor, and Cisplatin-OCCAMS) on the x-axis and the ratio of the surviving patients on the y-axis (Sakellaropoulos et al. Supplementary Figure 2.). It is implicitly shown that response to treatment is related to cancer recurrence in patients who have a beneficial response initially but then respond less which leads to a decrease patient survival (Sakellaropoulos et al. Supplementary Figure 2e (patients treated with Paclitaxel) which shows a steady patient survival for a period of time in the cohort with Low IC50 values followed by a steep decrease in patient survival which indicates less response to treatment/less effective of minimizing cancer cell growth (in the context of Paclitaxel which is an agent to stop cancer cell growth). This shows the relationship between response to treatment and survival rates over a duration of treatment time.
An invention would have been obvious to one or ordinary skill in the art if some motivation in the prior art would have led that person to modify reference teachings to arrive at the claimed invention. It would have been obvious to one of ordinary skill in the art before the effective filling date of the invention to have modified the batch correction process of Sakellaropoulos et al. with the particular batch correction method as shown in Korsunsky et al. because this would allow for method to use of a batch correction process that is computationally efficient, requires less memory compared to other algorithms, and is capable of analyzing large datasets on personal computers (Korsunsky et al. page 1290 right col.). One would have a reasonable expectation of success for this modification because Sakellaropoulos et al. shows the use of a batch correction process to correct for batch effects while Korsunsky et al. shows a computationally efficient batch correction process to correct for batch effects.
Response to Arguments
Applicant's arguments filed 12 June 2026 have been fully considered but they are not persuasive.
Applicant argues that the Office action admits that Sakellaropoulous does not teach or suggest any of the limitations in this quoted portion of claim 1 and cites Korsunsky for their alleged disclosure. See Office action at 28-29. Korsunsky, however, does not teach or suggest the steps of reducing a dimensionality of the first and second datasets to form first and second reduced datasets. It also does not teach or suggest generating separate first and second sets of adaptation factors related to the first and second reduced datasets, respectively. On the contrary, Korsunsky expressly states that it "avoid[s] all batch-sensitive preprocessing steps" but instead "simply concatenate[s] the data and perform[s] PCA on the combined dataset." Korsunsky at 1296 (emphasis added) (Reply p. 17).
This argument has been fully considered but found to be not persuasive. Although Korsunsky shows performing concatenation (which is a process of linking data sequentially where one end of the first data set is linked to the beginning of the second dataset) then performing PCA, these steps are still a process of reducing a dimensionality of the first and second datasets. Further, the Korsunsky shows deriving a set of adaptation factors related to the first reduced dataset and a set of adaptation factors related to the second reduced dataset by showing a process that uses eigenvalue-scaled eigenvectors produced by the principal component analysis (which is interpreted as adaptation factors) as the low-dimensional embedding for batch correction (Korsunsky et al. “Harmony” section page 1 left col. of Methods) because this process produced eigenvectors that are related to the first reduced dataset and eigenvectors that are related to the second reduced dataset. The claims do not require the datasets to be independently processed when producing this information.
Conclusion
No claims are allowed.
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 JONATHAN EDWARD HAYES whose telephone number is (571)272-6165. The examiner can normally be reached M-F 9am-5pm.
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.
/J.E.H./Examiner, Art Unit 1685
/KAITLYN L MINCHELLA/Primary Examiner, Art Unit 1685
Prosecution Insights