DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Status of Claims
This action is in reply to the present action filed on 09/18/2025.
Claims 11, 13, 20, 23, and 32 have been amended.
Claims 12 and 24 have been canceled.
Claims 11, 13-23, and 25-33 are currently pending and have been examined.
Continued Examination
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 03/02/2026 has been entered.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 11, 13-23, and 25-33 are rejected under 35 USC § 101 as being directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
Step 1 Analysis:
Independent Claims 11 and 23 are directed toward methods and therefore fall into one of the four statutory categories. Dependent Claims 13-22 and 25-33 also recite methods and therefore also fall into one of the four statutory categories.
Step 2A Analysis- Prong One:
Claim 11, which is representative of the inventive concept, recites:
A computer-implemented method for training a machine-learning model to predict type 2 Diabetes Mellitus (T2DM) status in a subject, the method comprising:
a) reading a training dataset corresponding to a plurality of biological samples, each biological sample comprising a liquid biopsy sample, the training dataset comprising, for each biological sample:
i) a first measure of a first gene methylation biomarker detected from circulating cell free DNA (ccfDNA) of its liquid biopsy, the first gene methylation biomarker comprises GCK (Glucokinase),
and ii) a second measure of a second gene methylation biomarker detected from ccfDNA of its liquid biopsy, the second gene methylation biomarker comprising one or more of IAPP (Islet Amyloid Polypeptide-Amylin) and KCNJ11 (Potassium Inwardly Rectifying Channel Subfamily J Member 11), and
iii) one or more demographic or clinical parameters associated with the biological sample, and iv) a type 2 Diabetes Mellitus (T2DM) status associated with the biological sample;
and b) providing the training dataset to a machine-learning classifier, thereby training the machine- learning classifier to provide a prediction of a T2DM status of a subject based on subject data, the subject data comprising:
i) a third measure of the first gene methylation biomarker detected from ccfDNA of a liquid biopsy sample from the subject,
ii) a fourth measure for the second gene methylation biomarker detected from ccfDNA of the liquid biopsy sample from the subject, and
iii) one or more demographic parameters of the subject,
wherein the first gene methylation biomarker and second gene methylation biomarker are associated with the pancreas.
Independent Claim 23 additionally recites the following limitations:
A computer-implemented method of predicting diabetes mellitus type 2 (T2DM) in a subject, the method comprising:
i) reading a first measure of a first gene methylation biomarker detected from circulating cell free DNA (ccfDNA) of a liquid biopsy sample of a biological sample of the subject, the first gene methylation biomarker comprising GCK (Glucokinase),
ii) reading a second measure of a second gene methylation biomarker detected from ccfDNA of the liquid biopsy sample, the second gene methylation biomarker comprising at least one of IAPP and KCNJ11, and
iii) reading one or more demographic or clinical parameters of the subject;
iv) providing, to a machine-learning classifier, the first measure, the second measure, and the one or more demographic parameters, wherein the machine-learning classifier is trained to predict a type 2 Diabetes Mellitus (T2DM) status, the training being based on a training data set comprising a third measure of the first gene methylation biomarker, a fourth measure of the second gene methylation biomarker, the one or more demographic parameters, and a T2DM status;
and v) obtaining, from the machine-learning classifier, a prediction of T2DM of the subject, wherein the first gene methylation biomarker and second gene methylation biomarker are associated with the pancreas.
The limitations as shown in underline above, given the broadest reasonable interpretation, recite the abstract idea of certain methods of organizing human activity because they recite managing personal behavior or relationship or interactions between people (i.e. social activities, teaching, and following rules or instructions – in this case, following steps to take patient specific data and generate a prediction of Type II Diabetes Mellitus status), e.g. see MPEP 2106.04(a)(2)(II). Any limitations not identified as part of the abstract idea are deemed “additional elements”, and will be discussed in further detail below.
Dependent Claims 13, 17-22, 25, and 29-33 recite additional limitations directed toward the abstract idea. For example, Claims 13 and 25 recite the dataset comprising the third gene methylation biomarker and which genes are measured, Claim 17 recites the first, second, third, and four measures correspond to indices of methylation levels, Claim 18 recites calculating measures based on demethylation index, Claim 19 recites calculating measures based on a percentage of the gene methylation biomarkers, Claims 20 and 32 recite the types of demographic parameters that can be included, Claim 21 recites diagnosing or prognosing T2DM in the subject, Claims 22 and 33 recite measuring beta-cell destruction and the prediction of the status of T2DM based on said measure, Claims 29 recites the first and second measure correspond to indices of methylation levels, Claim 30 recites calculating the percentage of the first and second biomarker in the sample, Claim 31 recites the details regarding the liquid biopsy sample. These limitations only serve to further narrow the abstract idea, and a claim may not preempt abstract ideas, even if the judicial exception is narrow, e.g. see MPEP 2106.04. Additionally, any limitations in dependent Claims 13-22 and 25-33 not addressed above are deemed additional elements to the abstract idea and will be further addressed below. Hence, dependent Claims 13, 17-22, 25, and 29-33 are nonetheless directed toward fundamentally the same abstract idea as independent Claims 11 and 23.
Step 2A Analysis - Prong Two:
Claims 11 and 23 are not integrated into practical application because the additional elements (i.e. the non-underlined portions presented in prong one- in this case, the machine learning classifier of Claims 11 and 23) are recited at a high level of generality (i.e. as a generic processor performing generic computer functions) such that they amount to no more than mere instructions to apply an exception using generic computer parts. For example, applicant’s specification explains that different Machine learning tools (machine learning classification algorithms) are used to build the predictive models. Any machine learning classification algorithm can be used and in different embodiments the following: Decision Tree, k-Nearest Neighbors (k-NN), Gradient Boosting Machine (GBM),…(p. 8, lines 16-20). Accordingly, these additional elements, when considered separately and as an ordered combination, do not integrate the abstract idea into practical application because they do not impose any meaningful limits on the abstract idea. Therefore Claims 11 and 23 are directed toward an abstract idea without practical application.
Dependent Claims 13-16, 22, and 26-28 recite additional elements. Claims 13, 22, 25, and 33 narrow the previously recited additional element of the machine-learning classifier. Claim 13 narrows the previously recited additional element of the machine-learning classifier and specifies the machine learning classifier provides the prediction of the diabetes status for the subject based on the sixth measure. Claims 22 and 33 narrow the previously recited additional element of the machine-learning classifier and specifies the machine learning classifier provides the measure of beta-cell destruction and a prediction of T2DM status for the subject when the measure indicates T2DM. Claim 25 narrows the previously recited machine-learning classifier and specifies the machine learning classifier receives a fifth measure of a third gene methylation biomarker. Claims 14 and 26 recite a new additional element of a Random Forest model and specify the machine learning classifier comprises a Classification Random Forest (RF). Claims 15 and 27 recite a new additional element of a Ridge Logistic Regression Model and specify the machine learning classifier comprises a Ridge Logistic Regression model. Claims 16 and 28 recite new additional elements of a Decision Tree, a k-Nearest Neighbors (k-NN) model, a Gradient Boosting Machine (GBM), a linear kernel Support Vector Machine (SVM-linear), a Radial Basis Function (RBF) kernel Support Vector Machine, an Artificial Neural Network (ANN), a Multifactor Dimensionality Reduction (MDR), a naive Bayes model, a Classification And Regression Tree (CART), a Support Vector Machine (SVM), a Random Forest (RF), and Logistic Regression (LR) and specify the machine learning classifier comprises one or more of the listed models. However, these additional elements are used in their expected fashion, so they do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on the abstract idea. These limitations amount to no more than mere instructions to apply an exception, and hence, do not integrate the aforementioned abstract idea into practical application.
Step 2B Analysis:
The claims, whether considered individually or as an ordered combination, do not include any additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the integration of the abstract idea into a practical application, the additional element of the machine-learning classifier of Claims 11 and 23 amount to no more than mere instructions to apply the exception using generic computer components. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept (“significantly more”). MPEP 2106.05(I)(A) indicates that merely stating “apply it” or an equivalent to the abstract idea cannot provide an inventive concept (“significantly more”). Accordingly, even in combination, these additional elements do not provide significantly more. As such, Claims 11 and 23 are not patent eligible.
Dependent Claims 17-21, and 29-32 do not recite any additional elements and solely narrow the abstract idea. Claim 17 narrows the abstract idea by specifying that the four measures correspond to indices of methylation levels. Claim 18 narrows the abstract idea by specifying calculating the four measures based on a demethylation index. Claims 20 and 32 narrow the abstract idea by specifying the patient demographic parameters. Claims 21 narrows the abstract idea by specifying the diagnosing or prognosing of T2DM in the subject. Claim 29 narrows the abstract idea by specifying the first and second measure correspond to indices of methylation levels. Claim 31 narrows the abstract idea by specifying what types of fluids encompass the liquid biopsy sample.
Claims 13, 22, 25, and 33 recite previously cited additional elements, which are not eligible for the reasons stated above, and further narrow the abstract idea. Claim 13 narrows the previously recited additional element of the machine-learning classifier by specifying the machine learning classifier provides the prediction of diabetes status based on the sixth measure. Claims 22 and 33 narrow the previously recited additional element of the machine-learning classifier by specifying the machine-learning classifier provides a measure of the beta-cell destruction and predicts the T2DM status based on that measurement. Claim 25 narrows the previously recited additional element of the machine-learning classifier by specifying the fifth measure of the third gene methylation biomarker is provided to the machine-learning classifier.
Claims 14-16 and 36-28 recite new additional elements that narrow an already recited additional element. Claims 14 and 26 recite a new additional element of a Random Forest which the machine-learning classifier comprises. Claims 15 and 27 recite a new additional element of a Ridge Logistic Regression Model which is the machine-learning classifier. Claims 16 and 28 recite new additional elements of a Decision Tree, a k-Nearest Neighbors (k-NN) model, a Gradient Boosting Machine (GBM), a linear kernel Support Vector Machine (SVM-linear), a Radial Basis Function (RBF) kernel Support Vector Machine, an Artificial Neural Network (ANN), a Multifactor Dimensionality Reduction (MDR), a naive Bayes model, a Classification And Regression Tree (CART), a Support Vector Machine (SVM), a Random Forest (RF), and Logistic Regression (LR) and specifies the machine-learning classifier comprises at least one of the mentioned models. Hence, Claims 13-22 and 25-33 do not include any additional elements that amount to “significantly more” than the judicial exception.
Thus, taken alone, the additional elements do not amount to significantly more than the abstract idea identified above. Furthermore, looking at the limitations as an ordered combination does not add anything that is already present when looking at the elements taken individually, and there is no indication that the combination of elements improves the functioning of computer or improves any other technology, and their collective functions merely provide conventional computer implementation.
Therefore, whether taken individually or as an ordered combination, Claims 11, 13-23, and 25-33 are nonetheless rejected under 35 USC § 101 as being directed to a non-statutory subject matter.
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.
Claims 11-17, 20-21, 23-29, and 31-32 are rejected under 35 USC § 103 as being unpatentable over Lai et al. (Lai, H., Huang, H., Keshavjee, K. et al. Predictive models for diabetes mellitus using machine learning techniques. BMC Endocr Disord 19, 101 (2019).) in view of Ahmed et al. (Ahmed, S.A.H., Ansari, S.A., Mensah-Brown, E.P.K. et al. The role of DNA methylation in the pathogenesis of type 2 diabetes mellitus. Clin Epigenet 12, 104 (2020).) and Panagopoulou et al. (Panagopoulou, M., Karaglani, M., Balgkouranidou, I. et al. Circulating cell-free DNA in breast cancer: size profiling, levels, and methylation patterns lead to prognostic and predictive classifiers. Oncogene 38, 3387–3401 (2019).).
Regarding Claim 11, Lai discloses the following:
A computer-implemented method for training a machine-learning model to predict type 2 Diabetes Mellitus (T2DM) status in a subject, the method comprising: (Lai discloses the ability of our model to predict patients with Diabetes … is high with satisfactory sensitivity. These models can be built into an online computer program to help physicians in predicting patients with future occurrence of diabetes (p. 1, ¶ 0004). We applied our methods on this dataset to predict whether or not a patient has diabetes (p. 7, ¶ 0001).)
a) reading a training dataset corresponding to a plurality of biological samples, … the training dataset comprising, for each biological sample: (Lai discloses we first created a training dataset by randomly choosing 80% of all patients in the dataset and created a test dataset with the remaining 20% of patients. The training dataset has 10,647 patients and the test dataset has 2662 patients. We used the training dataset to train the model and used the test dataset to evaluate how well the model performs based on an unseen dataset. Using the training dataset and the 10-fold cross-validation method, we tuned the model hyperparameters to obtain the set of optimal hyperparameters that yields the highest area under the receiver operating characteristic curve (AROC) (p. 3, ¶ 0003).)
i) a first measure of a first … biomarker…, ii) a second measure of a second …biomarker…, (Lai discloses there are 215,544 records pertaining to patient visits in the dataset. The outcome variable is Diabetes Mellitus which is encoded a binary variable, with category 0 indicating patients with no DM and category 1 indicating patients with DM. The predictors of interest are: … TG (Triglycerides), FBS (Fasting Blood Sugar), sBP (Systolic Blood Pressure), HDL (High Density Lipoprotein), and LDL (Low Density Lipoprotein) (p. 2, ¶ 0005). These predictors are biomarkers, and these are interpreted as a first and second biomarker used in the trained machine learning model.)
iii) one or more demographic or clinical parameters associated with the biological sample, (Lai discloses the predictors of interest are: Sex, Age (Age at examination date), BMI (Body Mass Index), TG (Triglycerides), FBS (Fasting Blood Sugar), sBP (Systolic Blood Pressure) (p. 2, ¶ 0005).)
and iv) a type 2 Diabetes Mellitus (T2DM) status associated with the biological sample; (Lai discloses we present predictive models using Gradient Boosting Machine and Logistic Regression techniques to predict the probability of patients having DM based on their demographic information and laboratory results from their visits to medical facilities (p. 2, ¶ 0003).)
and b) providing the training dataset to a machine-learning classifier, thereby training the machine- learning classifier to provide a prediction of a T2DM status of a subject based on subject data, the subject data comprising: (Lai discloses using the optimal threshold we evaluated the performance of the final model on the test dataset (p. 3, ¶ 0004). The test dataset is interpreted as the use of the subject data.)
i) a third measure of the first…biomarker …from the subject, ii) and/or a fourth measure for the second… biomarker from the … from the subject, and (Lai discloses there are 215,544 records pertaining to patient visits in the dataset. The outcome variable is Diabetes Mellitus which is encoded a binary variable, with category 0 indicating patients with no DM and category 1 indicating patients with DM. The predictors of interest are: … TG (Triglycerides), FBS (Fasting Blood Sugar), sBP (Systolic Blood Pressure), HDL (High Density Lipoprotein), and LDL (Low Density Lipoprotein) (p. 2, ¶ 0005). These predictors are biomarkers, and these are interpreted as a first and second biomarker used in the trained machine learning model.)
iii) and/or one or more demographic parameters of the subject, (Lai discloses the predictors of interest are: Sex, Age (Age at examination date), BMI (Body Mass Index), TG (Triglycerides), FBS (Fasting Blood Sugar), sBP (Systolic Blood Pressure) (p. 2, ¶ 0005).)
Lai does not disclose the following limitations met by Ahmed:
…each biological sample comprising a liquid biopsy sample, (Ahmed teaches this study used blood samples from patients with T2DM and healthy (control) individuals (p. 10, ¶ 0005).)
…gene methylation biomarker… (Ahmed teaches a study was conducted to investigate the DNA methylation of certain candidate genes and whether differential levels of DNA methylation in the blood could be used as biomarkers for T2DM and/or MetS (p. 11, ¶ 0004).)
…the first gene methylation biomarker comprises GCK (Glucokinase),… (Ahmed teaches studies have found that one of the main roles of insulin is to stimulate GCK expression to activate relevant glycolytic genes, thus increasing glucose utilisation. Indeed, insulin resistance due to downregulated glucokinase activities in patients with T2DM has been reported, indicating that GCK is a T2DM susceptibility gene. Whether the degree of DNA methylation in 11 CpG sites of the hepatic Gck promoter influences its expression and activity and whether age influences its expression and age-related diabetes were investigated in Wistar rats (p. 8, ¶ 0005-6).)
… gene methylation biomarker comprising one or more of IAPP (Islet Amyloid Polypeptide-Amylin) and KCNJ11 (Potassium Inwardly Rectifying Channel Subfamily J Member 11), (Ahmed teaches DNA hypermethylation of the micro RNA (miRNA) cluster in the imprinted 14q32 locus was found to substantially downregulate its expression in T2DM pancreatic β cells. The expression of this miRNA cluster was found to play an important role in regulating the expression of two T2DM-relevant genes, namely tumour protein p53-inducible nuclear protein-1 (TP53INP1) and islet amyloid polypeptide (IAPP) (p. 19, ¶ 0003).)
…gene methylation biomarker associated with the pancreas (Ahmed teaches insulin is a peptide hormone secreted by pancreatic β cells upon nutrient uptake. It regulates blood glucose levels by enhancing glucose uptake and glycolysis in skeletal and adipose tissues (p. 4, ¶ 0005). The peroxisome proliferator-activated receptor gamma coactivator-1 alpha (PPARGC1A) is a transcriptional coactivator of various transcription factors…PPARGC1A controls the activity of a wide range of transcription factors functioning in various cellular and metabolic processes… PPARGC1A is expressed predominantly in tissues with high energy demand including the…pancreas…(p. 6, ¶ 0007). GRB10 is expressed in a variety of tissues, with the highest expression in pancreas (p. 10, ¶ 0007). The Examiner interprets these gene methylation biomarkers as being associated with the pancreas.)
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have modified the method for training a machine learning model based on biomarkers and patient demographics to predict patient diabetes status as disclosed by Lai to incorporate the biomarkers being obtained via a liquid biopsy sample and specifically using gene methylation biomarkers and these biomarkers being associated with the pancreas as taught by Ahmed. This modification would create a method which is capable of more accurately diagnosing T2DM as it incorporates the role of DNA methylation in the development of the disease (see Ahmed, p. 4, ¶ 0004).
Lai and Ahmed do not teach the following limitations met by Panagopoulou:
…gene methylation biomarker detected from circulating cell free DNA (ccfDNA)… (Panagopoulou teaches recent novel approaches implicating “liquid biopsies” such as blood circulating cell-free DNA (ccfDNA), have been considered to provide biosources of potential clinically relevant information, meeting the need for a convenient, minimally-invasive advancement in the route of precision medicine [2, 3]. Numerous studies from our and other groups attempted to validate specific cancer-related gene methylation detected in ccfDNA as a biomarker for early cancer diagnosis, accurate prognosis, and dynamic drug response monitoring [8–10] (p. 2, ¶ 0001-2).)
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have modified the method for training a machine learning model based on biomarkers and patient demographics to predict patient diabetes status as disclosed by Lai to incorporate the biomarkers being obtained via circulating cell free DNA as taught by Panagopoulou. This modification would create a method which is capable of extracting information of biological characteristics of ccfDNA including methylation patterns (see Panagopoulou, p. 2, ¶ 0003).
Regarding Claim 13, Lai, Ahmed, and Panagopoulou teach the limitations as seen in the rejection of Claim 11 above. Lai further discloses
the subject data further comprise a sixth measure of the third … biomarker, and the machine-learning classifier is further configured to provide the prediction of the diabetes status for the subject based on the sixth measure. (Lai discloses the outcome variable is Diabetes Mellitus, which is encoded a binary variable, with category 0 indicating patients with no DM and category 1 indicating patients with DM. The predictors of interest are: … TG (Triglycerides), FBS (Fasting Blood Sugar), sBP (Systolic Blood Pressure), HDL (High Density Lipoprotein), and LDL (Low Density Lipoprotein) (p. 2, ¶ 0005). These predictors are biomarkers, and any can be selected as the third biomarker used in the trained machine learning model.)
Lai does not disclose the following limitations met by Ahmed:
wherein the training dataset further comprises a fifth measure of a third gene methylation biomarker, the third gene methylation biomarker measuring methylation of one or more of the following genes: INS (insulin), IAPP, KCNJ11, and ABCC8 (ATP Binding Cassette Subfamily C Member 8), or any other gene related to the pancreas (Ahmed teaches the role of DNA methylation was further examined in another study using human islets from T2DM and nonT2DM donors. The findings indicate that the glucose stimulated insulin secretion (GSIS), insulin mRNA, and insulin content were reduced in the pancreatic islets of T2DM in comparison with the non-T2DM donors (p. 5, ¶ 0002). The candidate genes were fat mass and obesity-associated (FTO), KCNJ11, …These genes were chosen based on their functional relevance to certain metabolic processes, including glucose metabolism (p.11, ¶ 0004). The expression of this miRNA cluster was found to play an important role in regulating the expression of two T2DM-relevant genes, namely tumour protein p53-inducible nuclear protein-1 (TP53INP1) and islet amyloid polypeptide (IAPP) (p. 19, ¶ 0003).)
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have modified the method for training a machine learning model based on biomarkers and patient demographics to predict patient diabetes status as disclosed by Lai to incorporate the use of a third biomarker being either INS, KCNJ11, or IAPP as taught by Ahmed. This modification would create a method which is capable of more accurately diagnosing T2DM as it incorporates the role of DNA methylation in the development of the disease (see Ahmed, p. 4, ¶ 0004).
Regarding Claim 14, Lai, Ahmed, and Panagopoulou teach the limitations as seen in the rejection of Claim 11 above. Lai further discloses:
wherein the machine-learning classifier comprises a Classification Random Forest (RF). (Lai discloses to compare the performance of the obtained Logistic Regression and GBM models with other machine-learning techniques, we used the same training dataset, test dataset, and procedure on the Rpart and Random Forest techniques. The results in Table 3 show that the GBM model performs the best based on highest AROC value, followed by the Logistic Regression model and the Random Forest model (p. 4, ¶ 0003-4).)
Regarding Claim 26, this claim recites limitations that are substantially similar to those recited in Claim 14 above; thus, the same rejection applies.
Regarding Claim 15, Lai, Ahmed, and Panagopoulou teach the limitations as seen in the rejection of Claim 11 above. Lai and Ahmed do not teach the following limitation met by Panagopoulou:
wherein the machine-learning classifier comprises a Ridge Logistic Regression model. (Panagopoulou teaches for classification modeling, JADBio tries SVM [50] with full polynomial and Gaussian kernels, random forests [51], ridge logistic regression [52], and decision trees [53] (p. 13, ¶ 0004).)
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have modified the method for training a machine learning model based on biomarkers and patient demographics to predict patient diabetes status as disclosed by Lai to incorporate the use of a Ridge Logistic Regression model as taught by Panagopoulou. This modification would create a method which is capable of providing a diabetes status prediction that is as accurate as possible (see Panagopoulou, p. 5, ¶ 0002).
Regarding Claim 27, this claim recites limitations that are substantially similar to those recited in Claim 15 above; thus, the same rejection applies.
Regarding Claim 16, Lai, Ahmed, and Panagopoulou teach the limitations as seen in the rejection of Claim 11 above. Lai further discloses:
wherein the machine learning classifier comprises one or more of a Decision Tree, a k-Nearest Neighbors (k-NN) model, a Gradient Boosting Machine (GBM), a linear kernel Support Vector Machine (SVM-linear), a Radial Basis Function (RBF) kernel Support Vector Machine, an Artificial Neural Network (ANN), a Multifactor Dimensionality Reduction (MDR), a naive Bayes model, a Classification And Regression Tree (CART), a Support Vector Machine (SVM), a Random Forest (RF), and Logistic Regression (LR). (Lai discloses we built predictive models using Logistic Regression and Gradient Boosting Machine (GBM) techniques (p. 1, ¶ 0002). Machine learning methods, such as logistic regression, artificial neural network, and decision tree were used by Meng et al. [12] to predict DM and pre-diabetes. The accuracy was reported to be 77.87% using a decision tree model; 76.13% using a logistic regression model; and 73.23% using the Artificial Neural Network (ANN) procedure. Other machine learning methods, such as Random Forest, Support Vector Machines (SVM), k-nearest Neighbors (KNN), and the naïve Bayes have also been used as in [6–8, 10, 11, 21] (p. 2, ¶ 0002).)
Regarding Claim 28, this claim recites limitations that are substantially similar to those recited in Claim 16 above; thus, the same rejection applies.
Regarding Claim 17, Lai, Ahmed, and Panagopoulou teach the limitations as seen in the rejection of Claim 11 above. Lai further discloses:
wherein the first, second, third, and fourth measures (Lai discloses there are 215,544 records pertaining to patient visits in the dataset. The outcome variable is Diabetes Mellitus which is encoded a binary variable, with category 0 indicating patients with no DM and category 1 indicating patients with DM. The predictors of interest are: … TG (Triglycerides), FBS (Fasting Blood Sugar), sBP (Systolic Blood Pressure), HDL (High Density Lipoprotein), and LDL (Low Density Lipoprotein) (p. 2, ¶ 0005). These predictors are biomarkers, and these are interpretted as the four biomarkers which measurements are used in the trained machine learning model.)
Lai does not disclose the following limitations met by Ahmed:
correspond to respective indices of methylation levels. (Ahmed teaches analysis of DNA methylation status of IGFBP-1 and its association with serum IGFBP-1 levels in T2DM found that the DNA methylation levels of six CpG sites were higher in T2DM patients compared with the control individuals. Newly diagnosed patients with T2DM with a familial history of the disease showed higher IGFBP-1 DNA methylation levels compared with those without a familial history of the disease (p. 8, ¶ 0007).)
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have modified the method for training a machine learning model based on biomarkers and patient demographics to predict patient diabetes status as disclosed by Lai to incorporate the measurement of the methylation levels of the biomarkers as taught by Ahmed. This modification would create a method which is capable of more accurately diagnosing T2DM as the measurement of methylation levels shows a correlation to diabetes patients (see Ahmed, p. 19, ¶ 0002).
Regarding Claim 20, Lai, Ahmed, and Panagopoulou teach the limitations as seen in the rejection of Claim 11 above. Lai further discloses:
wherein the one or more demographic or clinical parameters of the training data set and the one or more demographic or clinical parameters of the subject data each further comprise a body-mass index (BMI), smoking status, age, sex, glucose level, C-peptide level, HbAlc, diabetes duration, diabetes complication status, diabetes therapy status, insulin use, circulating cell free DNA concentration or any combination thereof. (Lai discloses the predictors of interest are: Sex, Age (Age at examination date), BMI (Body Mass Index), …FBS (Fasting Blood Sugar), sBP (Systolic Blood Pressure) (p. 2, ¶ 0005).)
Regarding Claim 32, this claim recites limitations that are substantially similar to those recited in Claim 20 above; thus, the same rejection applies.
Regarding Claim 21, Lai, Ahmed, and Panagopoulou teach the limitations as seen in the rejection of Claim 11 above. Lai further discloses:
wherein the method comprises diagnosing or prognosing T2DM in the subject (Lai discloses we present predictive models using Gradient Boosting Machine and Logistic Regression techniques to predict the probability of patients having DM based on their demographic information and laboratory results from their visits to medical facilities (p. 2, ¶ 0003).)
Regarding Claim 23, Lai discloses:
A computer-implemented method of predicting diabetes mellitus type 2 (T2DM) in a subject, the method comprising: (Lai discloses the ability of our model to predict patients with Diabetes … is high with satisfactory sensitivity. These models can be built into an online computer program to help physicians in predicting patients with future occurrence of diabetes (p. 1, ¶ 0004). We applied our methods on this dataset to predict whether or not a patient has diabetes (p. 7, ¶ 0001).)
i) reading a first measure of a first … biomarker from a liquid biopsy sample of a biological sample of the subject, ii) and/or reading a second measure of a second … biomarker …, (Lai discloses there are 215,544 records pertaining to patient visits in the dataset. The outcome variable is Diabetes Mellitus which is encoded a binary variable, with category 0 indicating patients with no DM and category 1 indicating patients with DM. The predictors of interest are: … TG (Triglycerides), FBS (Fasting Blood Sugar), sBP (Systolic Blood Pressure), HDL (High Density Lipoprotein), and LDL (Low Density Lipoprotein) (p. 2, ¶ 0005). These predictors are biomarkers, and these are interpretted as a first and second biomarker used in the trained machine learning model.)
iii) and/or reading one or more demographic parameters of the subject; (Lai discloses the predictors of interest are: Sex, Age (Age at examination date), BMI (Body Mass Index), TG (Triglycerides), FBS (Fasting Blood Sugar), sBP (Systolic Blood Pressure) (p. 2, ¶ 0005).)
iv) providing, to a machine-learning classifier, the first measure, the second measure, and the one or more demographic parameters, wherein the machine-learning classifier is trained to predict a type 2 Diabetes Mellitus (T2DM) status, the training being based on a training data set comprising a third measure of the first gene methylation biomarker, a fourth measure of the second gene methylation biomarker, the one or more demographic parameters, and a T2DM status; and (Lai discloses we first created a training dataset by randomly choosing 80% of all patients in the dataset and created a test dataset with the remaining 20% of patients. The training dataset has 10,647 patients and the test dataset has 2662 patients. We used the training dataset to train the model and used the test dataset to evaluate how well the model performs based on an unseen dataset (p. 3, ¶ 0003). Using the optimal threshold, we evaluated the performance of the final model on the test dataset (p. 3, ¶ 0004). The test dataset is interpretted as the use of the subject data.)
v) obtaining, from the machine-learning classifier, a prediction of T2DM of the subject, (Lai discloses the ability of our model to predict patients with Diabetes … is high with satisfactory sensitivity. These models can be built into an online computer program to help physicians in predicting patients with future occurrence of diabetes (p. 1, ¶ 0004). We applied our methods on this dataset to predict whether or not a patient has diabetes (p. 7, ¶ 0001).)
Lai is silent regarding the biomarker being a gene methylation biomarker as taught by Ahmed:
each biological sample comprising a liquid biopsy sample, (Ahmed teaches this study used blood samples from patients with T2DM and healthy (control) individuals (p. 10, ¶ 0005).)
…gene methylation biomarker… (Ahmed teaches a study was conducted to investigate the DNA methylation of certain candidate genes and whether differential levels of DNA methylation in the blood could be used as biomarkers for T2DM and/or MetS (p. 11, ¶ 0004).)
…the first gene methylation biomarker comprises GCK (Glucokinase),… (Ahmed teaches studies have found that one of the main roles of insulin is to stimulate GCK expression to activate relevant glycolytic genes, thus increasing glucose utilisation. Indeed, insulin resistance due to downregulated glucokinase activities in patients with T2DM has been reported, indicating that GCK is a T2DM susceptibility gene. Whether the degree of DNA methylation in 11 CpG sites of the hepatic Gck promoter influences its expression and activity and whether age influences its expression and age-related diabetes were investigated in Wistar rats (p. 8, ¶ 0005-6).)
… gene methylation biomarker comprising one or more of IAPP and KCNJ11,… (Ahmed teaches DNA hypermethylation of the micro RNA (miRNA) cluster in the imprinted 14q32 locus was found to substantially downregulate its expression in T2DM pancreatic β cells. The expression of this miRNA cluster was found to play an important role in regulating the expression of two T2DM-relevant genes, namely tumour protein p53-inducible nuclear protein-1 (TP53INP1) and islet amyloid polypeptide (IAPP) (p. 19, ¶ 0003).)
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have modified the method for training a machine learning model based on biomarkers and patient demographics to predict patient diabetes status as disclosed by Lai to incorporate the biomarkers being obtained via a liquid biopsy sample and specifically using gene methylation biomarkers as taught by Ahmed. This modification would create a method which is capable of more accurately diagnosing T2DM as it incorporates the role of DNA methylation in the development of the disease (see Ahmed, p. 4, ¶ 0004).
Lai and Ahmed do not disclose the following limitations met by Panagopoulou:
wherein the first gene methylation biomarker and second gene methylation biomarker are detected in circulating cell free DNA (ccfDNA) from the liquid biopsy sample. (Panagopoulou teaches recent novel approaches implicating “liquid biopsies” such as blood circulating cell-free DNA (ccfDNA), have been considered to provide biosources of potential clinically relevant information, meeting the need for a convenient, minimally-invasive advancement in the route of precision medicine [2, 3]. Numerous studies from our and other groups attempted to validate specific cancer-related gene methylation detected in ccfDNA as a biomarker for early cancer diagnosis, accurate prognosis, and dynamic drug response monitoring [8–10] (p. 2, ¶ 0001-2).)
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have modified the method for training a machine learning model based on biomarkers and patient demographics to predict patient diabetes status as disclosed by Lai to incorporate the biomarkers being obtained via circulating cell free DNA as taught by Panagopoulou. This modification would create a method which is capable of extracting information of biological characteristics of ccfDNA including methylation patterns (see Panagopoulou, p. 2, ¶ 0003).
Regarding Claim 25 Lai, Ahmed, and Panagopoulou teach the limitations as seen in the rejection of Claim 24 above. Lai further discloses:
wherein the method comprises diagnosing or prognosing T2DM in the subject. (Lai discloses we present predictive models using Gradient Boosting Machine and Logistic Regression techniques to predict the probability of patients having DM based on their demographic information and laboratory results from their visits to medical facilities (p. 2, ¶ 0003).)
the subject data further comprise a sixth measure of the third … biomarker, and the machine-learning classifier is further configured to provide the prediction of the diabetes status for the subject based on the sixth measure. (Lai discloses the outcome variable is Diabetes Mellitus, which is encoded a binary variable, with category 0 indicating patients with no DM and category 1 indicating patients with DM. The predictors of interest are: … TG (Triglycerides), FBS (Fasting Blood Sugar), sBP (Systolic Blood Pressure), HDL (High Density Lipoprotein), and LDL (Low Density Lipoprotein) (p. 2, ¶ 0005). These predictors are biomarkers, and any can be selected as the third biomarker used in the trained machine learning model.)
reading a fifth measure of the third … biomarker from the … the subject; and providing the fifth measure to the machine-learning classifier. (Lai discloses there are 215,544 records pertaining to patient visits in the dataset. The outcome variable is Diabetes Mellitus which is encoded a binary variable, with category 0 indicating patients with no DM and category 1 indicating patients with DM. The predictors of interest are: … TG (Triglycerides), FBS (Fasting Blood Sugar), sBP (Systolic Blood Pressure), HDL (High Density Lipoprotein), and LDL (Low Density Lipoprotein) (p. 2, ¶ 0005). These predictors are biomarkers, and these are interpretted as a third biomarker used in the trained machine learning model.)
Lai does not disclose the following limitations met by Ahmed:
wherein the training dataset further comprises a sixth measure of a third gene methylation biomarker, the third gene methylation biomarker measuring methylation of one or more of the following genes: INS (insulin), IAPP, KCNJ11, and ABCC8 (ATP Binding Cassette Subfamily C Member 8), (Ahmed teaches the role of DNA methylation was further examined in another study using human islets from T2DM and nonT2DM donors. The findings indicate that the glucose stimulated insulin secretion (GSIS), insulin mRNA, and insulin content were reduced in the pancreatic islets of T2DM in comparison with the non-T2DM donors (p. 5, ¶ 0002). The candidate genes were fat mass and obesity-associated (FTO), KCNJ11, …These genes were chosen based on their functional relevance to certain metabolic processes, including glucose metabolism (p.11, ¶ 0004). The expression of this miRNA cluster was found to play an important role in regulating the expression of two T2DM-relevant genes, namely tumour protein p53-inducible nuclear protein-1 (TP53INP1) and islet amyloid polypeptide (IAPP) (p. 19, ¶ 0003).)
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have modified the method for training a machine learning model based on biomarkers and patient demographics to predict patient diabetes status as disclosed by Lai to incorporate the use of a third biomarker being either INS, KCNJ11, or IAPP as taught by Ahmed. This modification would create a method which is capable of more accurately diagnosing T2DM as it incorporates the role of DNA methylation in the development of the disease (see Ahmed, p. 4, ¶ 0004).
Regarding Claim 29, Lai, Ahmed, and Panagopoulou teach the limitations as seen in the rejection of Claim 23 above. Lai further discloses:
wherein the first and second measures (Lai discloses there are 215,544 records pertaining to patient visits in the dataset. The outcome variable is Diabetes Mellitus which is encoded a binary variable, with category 0 indicating patients with no DM and category 1 indicating patients with DM. The predictors of interest are: … TG (Triglycerides), FBS (Fasting Blood Sugar), sBP (Systolic Blood Pressure), HDL (High Density Lipoprotein), and LDL (Low Density Lipoprotein) (p. 2, ¶ 0005). These predictors are biomarkers, and these are interpretted as the two biomarkers which measurements are used in the trained machine learning model.)
Lai does not disclose the following limitations met by Ahmed:
correspond to respective indices of methylation levels. (Ahmed teaches analysis of DNA methylation status of IGFBP-1 and its association with serum IGFBP-1 levels in T2DM found that the DNA methylation levels of six CpG sites were higher in T2DM patients compared with the control individuals. Newly diagnosed patients with T2DM with a familial history of the disease showed higher IGFBP-1 DNA methylation levels compared with those without a familial history of the disease (p. 8, ¶ 0007).)
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have modified the method for training a machine learning model based on biomarkers and patient demographics to predict patient diabetes status as disclosed by Lai to incorporate the measurement of the methylation levels of the biomarkers as taught by Ahmed. This modification would create a method which is capable of more accurately diagnosing T2DM as the measurement of methylation levels shows a correlation to diabetes patients (see Ahmed, p. 19, ¶ 0002).
Regarding Claim 31, Lai, Ahmed, and Panagopoulou teach the limitations as seen in the rejection of Claim 23 above. Lai does not disclose the following limitation met by Ahmed:
wherein the liquid biopsy sample comprises one or more bodily fluids selected from blood, serum, plasma, or any combination thereof. (Ahmed teaches this study used blood samples from patients with T2DM and healthy (control) individuals, along with the human liver carcinoma cell line HepG2 and liver tissue from male KK and KK-Cg-Ay /J mice of three age groups (p. 10, ¶ 0004).)
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have modified the method for training a machine learning model based on biomarkers and patient demographics to predict patient diabetes status as disclosed by Lai to incorporate the biomarkers being obtained via a blood sample as taught by Ahmed. This modification would create a method which is capable of more accurately diagnosing T2DM as it incorporates the role of DNA methylation in the development of the disease (see Ahmed, p. 4, ¶ 0004).
Claims 18, 22, and 33 are rejected under 35 USC § 103 as being unpatentable over Lai, Ahmed, and Panagopoulou in view of Akirav et al. (US 20160369340 A1).
Regarding Claim 18, Lai, Ahmed, and Panagopoulou teach the limitations as seen in the rejection of Claim 11 above. Lai, Ahmed, and Panagopoulou do not teach the following limitation met by Akirav:
calculating the first, second, third, and fourth measures based on a demethylation index. (Akirav teaches a ratio representing the relative abundance of demethylated DNA is described as the demethylation index (DMI) (Akirav et. al. (12)) shown in FIG. 14B. Alternately, DNA probes which are specific for methylated and demethylated MOG gene may be used to quantitatively determine the DMI [0266].)
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have modified the method for training a machine learning model based on biomarkers and patient demographics to predict patient diabetes status as disclosed by Lai to incorporate calculating a demethylation index as taught by Akirav. This modification would create a method which is capable of more accurately diagnosing T2DM as it determines the methylation status of a target biomarker to assist in determining disease status and progression (see Akirav, ¶ 0069, 0045).
Regarding Claim 22, Lai, Ahmed, and Panagopoulou teach the limitations as seen in the rejection of Claim 11 above. Lai further discloses:
and wherein the machine- learning classifier provides the prediction of T2DM status (Lai discloses the main contribution of our research study was proposing two predictive models using machine-learning techniques, Gradient Boosting Machine and Logistic Regression, in order to identify patients with high risk of developing DM (p. 8, ¶ 0002).)
Lai, Ahmed, and Panagopoulou do not teach the following limitation met by Akirav:
wherein the machine-learning classifier is further configured to provide a measure of β-cell destruction, …for the subject when the measure of β-cell destruction indicates T2DM. (Akirav teaches a method is provided for the detection of extrapancreatic circulating β cell, or β cell-derived DNA that is indicative of acute and chronic β cell or oligodendrocyte destruction, and thus provides an early biomarker for β cell or oligodendrocyte death in human tissues, serum and other bodily fluids, such as plasma, lymph, saliva, urine, cerebrospinal fluid, tears, and perhaps sweat. This strategy may prove useful for monitoring β cell or oligodendrocyte destruction in individuals at risk for the development of diabetes or multiple sclerosis [0046].)
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have modified the method for training a machine learning model based on biomarkers and patient demographics to predict patient diabetes status as disclosed by Lai to incorporate determining a measure of beta-cell destruction as taught by Akirav. This modification would create a method which is capable of more accurately diagnosing T2DM as it determines the amount of beta cell loss to assist in determining disease status and progression (see Akirav, ¶ 0031).
Regarding Claim 33, this claim recites limitations that are substantially similar to those recited in Claim 22 above; thus, the same rejection applies.
Claims 19 and 30 are rejected under 35 USC § 103 as being unpatentable over Lai, Ahmed, and Panagopoulou in view of Ding et al. (US 20150275300 A1).
Regarding Claim 19, Lai, Ahmed, and Panagopoulou teach the limitations as shown in the rejection of Claim 11 above. Lai, Ahmed, and Panagopoulou do not teach the following limitation met by Ding:
calculating the first, second, third, and fourth measure based on a percentage of the first gene methylation biomarker and second gene methylation biomarker in the liquid biopsy sample. (Ding teaches a method of determining the methylation levels of a biomarker/biomarker region comprising the steps of: (a) treating a sample… (b) calculating the percentage of unmodified cytosine residues over the total number of modified and unmodified cytosine residues in order to determine the methylation levels of a biomarker/biomarker region (Claim 47).)
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have modified the method for training a machine learning model based on biomarkers and patient demographics to predict patient diabetes status as disclosed by Lai to incorporate calculating the percentage of the gene biomarker in the sample as taught by Ding. This modification would create a method which is capable of accurately determining methylation levels of a biomarker (see Ding, ¶ 0023).
Regarding Claim 30, Lai, Ahmed, and Panagopoulou teach the limitations as shown in the rejection of Claim 23 above. Lai, Ahmed, and Panagopoulou do not teach the following limitation met by Ding:
calculating the first and second measure based on a percentage of the first gene methylation biomarker and second gene methylation biomarker in the liquid biopsy sample. (Ding teaches a method of determining the methylation levels of a biomarker/biomarker region comprising the steps of: (a) treating a sample… (b) calculating the percentage of unmodified cytosine residues over the total number of modified and unmodified cytosine residues in order to determine the methylation levels of a biomarker/biomarker region (Claim 47).)
It would have been obvious to a person having ordinary skill in the art prior to the effective filing date of the claimed invention to have modified the method for training a machine learning model based on biomarkers and patient demographics to predict patient diabetes status as disclosed by Lai to incorporate calculating the percentage of the gene biomarker in the sample as taught by Ding. This modification would create a method which is capable of accurately determining methylation levels of a biomarker (see Ding, ¶ 0023).
Response to Arguments
Regarding rejections under 35 USC 101 to Claims 11, 13-23, and 25-33, Applicant’s arguments have been considered but are not persuasive. The rejection has been updated in light of the amendments above.
Applicant argues with respect to Step 2A Prong One, Claim 1 is not an abstract idea because it provides it is directed to a technical pipeline of training a machine-learning model that use specific scientific measurements and computer- implemented processing steps in a diagnostic context. The claim is therefore not directed to certain methods of organizing human activity, as asserted by the Examiner, because the method is not rules or frameworks for social, commercial, or interpersonal behavior. Applicant further submits that Claim 11, as currently amended, further is not directed to certain methods of organizing human activity, as asserted by the Examiner, because it recites that the "training dataset corresponds to...biological samples...comprising ...a first measure of [GCK]...[and] a second measure of [IAPP and/or KCNJ11]." As such, the claims are directed to training a domain-specific diagnostic machine-learning model. This is analogous to Claim 2 of Example 39 where the claim did not recite a judicial exception because it did not recite a mathematical concept, mental process or method of organizing human activity (see Applicant’s Remarks, p. 9-10).
Regarding (a), Examiner respectfully disagrees. The additional elements are not considered when evaluating whether there is a judicial exception present. Instead, the additional elements are omitted, and the limitations that remain are evaluated for the presence of a judicial exception, and in this case, the limitations are steps that could be reasonably carried out by a person behind a generic computer. The additional elements (i.e., the machine learning) are evaluated to determine if the additional elements integrate the abstract idea into a judicial exception in step 2A Prong Two. In this case, the abstract idea is identified as being the steps of reading a training dataset corresponding to a plurality of biological samples and providing a prediction of a diabetes status based on subject data. The prediction of a diabetes status is deemed to be similar to “a mental process that a neurologist should follow when testing a patient for nervous system malfunctions, In re Meyer, 688 F.2d 789, 791-93, 215 USPQ 193, 194-96 (CCPA 1982)”, which is an example of managing personal behavior in a claim under the ‘Managing Personal Behavior or Relationships or Interactions Between People’ subheading of the ‘Certain Methods of Organizing Human Activity” grouping of abstract ideas, per MPEP 2106.04(a)(2)(II)(C).
Applicant argues, with respect to Step 2A Prong Two, the claims now recite "training dataset corresponds to...biological samples...comprising...a first measure of [GCK]...[and] a second measure of [IAPP and/or KCNJ11]." At least by virtue of reciting these elements, Claim 1 is directed to a practical application because the claims are directed to improving a specific technical field of a minimally-invasive, tissue- of-origin diagnostic inference for T2DM. The improvement is claimed with concrete, domain- specific measurements used in a training dataset.
The claim requires, "for each biological sample," measurements "detected from [ccfDNA] of its liquid biopsy," where "the first gene methylation biomarker comprises GCK" and "the second ... comprises ...IAPP [and/or] KCNJl1," and that those biomarkers are "associated with the pancreas." The above recited elements are not a generic "environment," field-of-use gloss, or a recitation to "apply it with a computer." Rather, the recitations are a specific technical data structure and method of training a machine learning model tied to tissue-of-origin methylation signals for pancreas (p. 11).
Regarding (b), Examiner respectfully disagrees. Examiner notes that the distributed nature of the claimed invention is not, in itself, dispositive in determining whether the claimed invention recites an abstract idea because the concept of managing data associated with Type 2 Diabetes is not a technological solution to a technological problem – that is, the concept of diabetes diagnostics have existed since long before the advent of computer technology, and thus, cannot properly be considered a technological improvement and/or an improvement to the computer itself. A claim that integrates a judicial exception into a practical application will apply, rely on, or use the judicial exception in a manner that imposes a meaningful limit on the judicial exception, such that the claim is more than a drafting effort designed to monopolize the judicial exception (see MPEP § 2106.04(d) - Integration of a Judicial Exception Into A Practical Application). The court has provided limitations that are indicative that an additional element (or combination of elements) may have integrated the exception into a practical application and limitations that did not integrate a judicial exception into a practical application (see MPEP § 2106.04(d)(I) – Relevant Considerations for Evaluating Whether Additional Elements integrate a Judicial Exception into a Practical Application) wherein the claims may amount to (1) improvements to the functioning of a computer, (2) improvements to a technological field, (3) applying the judicial exception to a particular machine (as evaluated above in ¶ ), (4) transforming or reducing a particular article to a different state or thing, (5) unconventional activity or steps that confine the claim to a particular useful application, or (6) other meaningful limitations beyond generally linking the use of the judicial exception to a particular technological environment. Here the instant claims seem more analogous to "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea, as discussed in MPEP § 2106.05(f).
As outlined above, the additional elements of the machine learning system which uses a training dataset did not meaningfully limit the abstract idea because they merely linked the use of an abstract idea to a particular technological environment (i.e., “implementation via computers”) and were recited at a high level of generality (MPEP 2106.05(e)). When making a determination of meaningful limitations, the Examiner should consider if additional elements that provide an inventive concept to the claim as a whole. Here, the combination of the additional elements does not provide an inventive computation environment – the machine learning classifier is utilized in a predictable manner by applying inputting data into the model, outputting a prediction, and reporting the results. The amendments to the independent claim specifying what the first and second biomarkers entail are details of the training dataset; neither the claims nor specification provide any detail on how these particular biomarker types would serve to improve a training dataset more than any other type of data or biomarker types. Similarly, it is not clear how the details of the first and second biomarkers would improve a specific technical field of a minimally-invasive, tissue- of-origin diagnostic inference for T2DM, considering that the claims do not require collecting or obtaining the data in such a way that would be “invasive”.
Applicant argues the claims apply any alleged abstract concept in a concrete pipeline designed to produce pancreas-specific T2DM predictions from ccfDNA methylation signals using a specially trained machine-learning model. The USPTO's examples reflect this approach: Example 47 (network intrusion detection) and Example 48 (speech separation) find eligibility at Step 2A, Prong Two based on technical-field improvements. Accordingly, the claim is patent eligible under Prong Two because it recites a practical application of the judicial exception. For this reason alone, the claim rejections based on 35 U.S.C. §101 should be withdrawn (p. 11).
Regarding (c), Examiner respectfully disagrees. Claim 3 of Example 47 was found to be eligible because it provided a technical improvement to a computer or technical field because in claim 3, the recited limitation provides a specific technique for solving an intrinsically technical problem of network intrusions. Because this was an invention that specifically addressed and solved a problem specific to a technical field, the claim was found to be eligible.
Claim 3 of Example 48 similarly is eligible under step 2A prong two because it addresses and solves the technical problem of distinguishing speech sources in a computer setting when the claim is examined as a whole.
Conversely to these examples, in the instant claims, there is no technical problem present and no improvement to the functioning of a computer or any other related technical field.
Applicant argues, with respect to Step 2B, the claims include limitations that amount to "significantly more" because "an inventive concept can be found in the non-conventional and non-generic arrangement of known, conventional pieces." Bascom Global Internet Services v. AT&T Mobility, 827 F. 3d 1341, 1350 (Fed. Cir. 2016). In particular, Applicant's Claim 11 has been amended to recite "reading a training dataset corresponding to a plurality of biological samples, each biological sample comprising a liquid biopsy sample, the training dataset comprising, for each biological sample: i) a first measure of a first gene methylation biomarker detected from circulating [ccfDNA] of its liquid biopsy, the first gene methylation biomarker comprises [GCK], ii) a second measure of a second gene methylation biomarker detected from ccfDNA of its liquid biopsy, the second gene methylation biomarker comprising [IAPP and/or KCNJl1], iii) ...demographic or clinical parameters associated with the biological sample, and iv) a [T2DM] status associated with the biological sample; and b) providing the training dataset to a machine-learning classifier, thereby training the machine-learning classifier to provide a prediction of a T2DM status of a subject based on subject data." These elements constitute more than the "abstract idea" identified in the Office Action, and provide "specific limitation[s] other than what is well-understood, routine and conventional in the field." (See 2014 Interim Guidance on Patent Subject Matter Eligibility at I.B.1.). In particular, Claim 1 as currently amended recites "for each biological sample," measurements "detected from [ccfDNA] of its liquid biopsy," where "the first gene methylation biomarker comprises GCK" and "the second ... comprises .IAPP [and/or] KCNJl1," and that those biomarkers are "associated with the pancreas." Such claims clearly do not seek to tie up any judicial exception. The Examiner does not provide evidence that the above elements are well-understood, routine, or conventional (p. 12).
Regarding (d), Examiner respectfully disagrees. Firstly, Examiner notes that if it was asserted in the office action that the additional elements of the claims were well-understood, routine, conventional, then the Examiner would have to provide Berkheimer evidence to support this statement. However, the Office Action groups the additional elements of the claims as generic computer components using the term “apply it” (or an equivalent) which does not require any evidence as support. MPEP 2106.07(a)(III) states the courts consider the determination of whether a claim is eligible (which involves identifying whether an exception such as an abstract idea is being claimed) to be a question of law. Rapid Litig. Mgmt. v. CellzDirect, 827 F.3d 1042, 1047, 119 USPQ2d 1370, 1372 (Fed. Cir. 2016); OIP Techs. v. Amazon.com, 788 F.3d 1359, 1362, 115 USPQ2d 1090, 1092 (Fed. Cir. 2015); DDR Holdings v. Hotels.com, 773 F.3d 1245, 1255, 113 USPQ2d 1097, 1104 (Fed. Cir. 2014); In re Roslin Institute (Edinburgh), 750 F.3d 1333, 1335, 110 USPQ2d 1668, 1670 (Fed. Cir. 2014); In re Bilski, 545 F.3d 943, 951, 88 USPQ2d 1385, 1388 (Fed. Cir. 2008) (en banc), aff’d by Bilski v. Kappos, 561 U.S. 593, 95 USPQ2d 1001 (2010). Thus, the court does not require "evidence" that a claimed concept is a judicial exception, and generally decides the legal conclusion of eligibility without resolving any factual issues. FairWarning IP, LLC v. Iatric Sys., 839 F.3d 1089, 1097, 120 USPQ2d 1293, 1298 (Fed. Cir. 2016) (citing Genetic Techs. Ltd. v. Merial LLC, 818 F.3d 1369, 1373, 118 USPQ2d 1541, 1544 (Fed. Cir. 2016)); OIP Techs., 788 F.3d at 1362, 115 USPQ2d at 1092; Content Extraction & Transmission LLC v. Wells Fargo Bank, N.A., 776 F.3d 1343, 1349, 113 USPQ2d 1354, 1359 (Fed. Cir. 2014). In some cases, however, the courts have characterized the issue of whether additional elements are well-understood, routine, conventional activity as an underlying factual issue upon which the legal conclusion of eligibility may be based. See, e.g., Interval Licensing LLC v. AOL, Inc., 896 F.3d. 1335, 1342, 127 USPQ2d 1553, 1557 (Fed. Cir. 2018) (patent eligibility is a question of law that may contain underlying issues of fact), Berkheimer v. HP, Inc., 881 F.3d 1360, 1368, 125 USPQ2d 1649, 1654 (Fed. Cir. 2018) (issue of whether additional elements are well-understood, routine, conventional activity is factual).
Furthermore, this section of the MPEP additionally states at Step 2A Prong Two or Step 2B, there is no requirement for evidence to support a finding that the exception is not integrated into a practical application or that the additional elements do not amount to significantly more than the exception unless the examiner asserts that additional limitations are well-understood, routine, conventional activities in Step 2B. The Examiner has not asserted the limitations as well-understood, routine, conventional activities and because of this, does not need to provide Berkheimer evidence.
Furthermore, Examiner notes that in this case, the limitations identified by the Applicant ("for each biological sample," measurements "detected from [ccfDNA] of its liquid biopsy," where "the first gene methylation biomarker comprises GCK" and "the second ... comprises .IAPP [and/or] KCNJl1," and that those biomarkers are "associated with the pancreas") are abstract and therefore cannot provide significantly more. An improvement to an abstract idea of the type of biomarkers being analyzed does not amount to an improvement to technology or a technical field (see MPEP § 2106.05(a)(III) stating “it is important to keep in mind that an improvement in the abstract idea itself (e.g. a recited fundamental economic concept) is not an improvement in technology. For example, in Trading Technologies Int’l v. IBG, 921 F.3d 1084, 1093-94, 2019 USPQ2d 138290 (Fed. Cir. 2019), the court determined that the claimed user interface simply provided a trader with more information to facilitate market trades, which improved the business process of market trading but did not improve computers or technology.”). There is no indication in the instant disclosure that the involvement of a computer assists in improving the technology for the outlined problem statement. Here, the improvement is to a diagnosis of diabetes. The instant application and claim language fail to detail how a computer aids the method, the extent to which the computer aids the method, or the significance of a computer to the performance of the method. Merely adding generic computer components to perform the method is not sufficient.
Applicant argues the ordered combination of Claim 11 is significantly more than any alleged abstract idea. Considered as a whole, the claims recite more than "apply it on a computer." The claims require a technically constrained input source (ccfDNA from liquid biopsy), specified pancreas-associated biomarkers at training and inference ("the first ... comprises GCK"; "the second ... comprises one or more of IAPP ... and KCNJ11"), incorporation of clinical/demographic parameters, and training/applying a classifier to "obtain[] ... a prediction of T2DM." This non-generic arrangement meaningfully limits any alleged exception and, as an ordered combination, amounts to "significantly more." The USPTO's examples similarly recognize that when steps do not represent merely gathering data for comparison, but instead set up a sequence of events that address unique problems in the field, the claims are directed to significantly more. Such is the case here, where specific biomarkers are gathered in an arranged in a training dataset in a per-subject manner and are used to train a machine-learning model. Thus, as in Bascom, the claims should be deemed patent eligible under Step 2B. (See, generally, Nov. 2016 Memo at 3.) (p. 12-13).
Regarding (e), Examiner respectfully disagrees. The claims do not provide “specifically more” because there is no improvement to a computer or technology provided. Further, the invention does not provide a non-conventional arrangement of known pieces. The 2024 USPTO Guidance Update “an improvement can be provided by one or more additional elements or by the additional element(s) in combination with the recited judicial exception. An exemplary case illustrating such an improvement is McRO, Inc. v. Bandai Namco Games America Inc., 837 F.3d 1299 (Fed. Cir. 2016), which is discussed extensively in the MPEP at, e.g., 2106.04(d)(1) and 2106.05(a). In McRO, the claims were to a rule-based system to animate the lip synchronization and facial expressions of three-dimensional characters. The Federal Circuit relied on the specification's explanation of how the claimed rules enabled the automation of specific animation tasks that previously could not be automated. The court indicated that it was the incorporation of the particular claimed rules in computer animation that ‘improved [the] existing technological process.’” Accordingly, the specification of the instant claims does not provide details regarding how the claims improve the existing technological process. The steps of training a machine learning model with historical data, inputting the target data and outputting a recommendation for the target is how the fundamental execution of machine learning.
Regarding rejections under 35 USC 103 to Claims 11-33, Applicant’s arguments have been considered but are not persuasive. The rejection has been updated in light of the amendments above.
Applicant argues the Office acknowledges Lai fails to disclose (1) a biological sample being a liquid biopsy, (2) the biomarker being a gene methylation biomarker, and (3) the gene methylation biomarker being associated with the pancreas. The use of a liquid biopsy ccfDNA sample enables determining changes in gene methylation markers of cells specific to the pancreas. Measurement of cell specific changes through markers such as GCK, IAPP, and KCNJ11 provide higher discrimination performance for tissue specific changes. When combined with demographic data as claimed, the method enables early and precise detection of type 2 diabetes mellitus in patients. As discussed supra, the method demonstrates an average precision of 0.951 (95%) in discriminating between diabetes patients and healthy subjects. In contrast, Lai discloses a receiver operating characteristic curve (ROC) ranging from 84.0% -84.7% and sensitivity ranging from 71.6% - 73.4% misclassification rate of about 18.9% (see Lai at page 3, column 2, paragraphs 3-4). Lai states that its authors found the most important predictors in their model to be fasting blood glucose, high-density lipoprotein, body mass index, and triglycerides (see Lai at page 4, columns 1 and 2). The physiological changes that Lai uses in its model can be conflated by other pathological diseases in a patient and may arise long after significant beta-cell destruction, however. Such conflations prevent the specific and early detection of beta-cell destruction, which in turn can prevent earlier intervention for patients. Lai fails to teach or suggest use of the described methods with ccfDNA, the prediction of T2DM using a panel of gene methylation markers, or the combination of such features with patient pathophysiological and demographic data.
Regarding (a), in response to applicant's arguments against the references individually, one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986). The 103 rejection cites Ahmed and Panagopoulou to teach these limitations not disclosed by Lai.
Applicant argues the Office asserts Ahmed addresses the deficiencies of Lai discussed above. The Office asserts Ahmed discloses use of liquid biopsies for measurement of gene methylation biomarkers related to T2DM. Notably, Ahmed discusses the detection of diabetes by separate measurement of GCK, INS, and KCNJ11 methylation data from peripheral blood leukocytes but not in ccfDNA. As recited in Applicant's Specification as filed, "[m]ethylation is a tissue specific- event and its detection in liquid biopsy material (ccfDNA) differs dramatically that its detection in genomic DNA from other sources such as blood cells. Gene methylation status detected in ccfDNA can reflect methylation status in the tissue of origin, in this case the beta-pancreatic islets." ([0010]). ccfDNA can be released into circulation through normal cell processes (e.g., through exosomes) and through death processes making elucidation of the relationship of certain methylation profiles difficult. Ahmed acknowledges the complexities of measuring gene methylation expression profiles. In particular, Ahmed describes the difficulty in dissecting whether certain profiles are the cause, or the result of disease development, and that "the methylation landscape, although associated with T2DM, can either be one of the contributing factors in development of T2DM or the result of the reaction of B cells to T2DM hyperglycemia" (see Ahmed, Future Scope, first paragraph). Notably, when discussing the gene methylation profiles disclosed in the review, Ahmed acknowledges the majority of the DNA profiles were determined using peripheral blood samples (see, Ahmed, page 20, column 1, paragraph 5). Peripheral blood samples include whole blood cells and ccfDNA from cell types throughout the body, which prohibits early, sensitive, and accurate detection of beta-cell destruction, an indication of T2DM, using a model trained from pancreatic specific methylation biomarkers found in ccfDNA.
Not only does Ahmed fail to disclose classification of a diabetes diagnosis through a panel of biomarkers, but Ahmed also fails to resolve the conflation of genomic DNA from other sources when measuring methylation changes, and fails to combine these markers with a clinicopathological or demographic/lifestyle parameters as in the pending claims to build and use a predictive model. Therefore, even if a person of ordinary skill in the art starting from Lai looked to Ahmed to incorporate pancreatic biomarkers, as the Office asserts, there is no teaching or suggestion to utilize ccfDNA gene methylation biomarkers, let alone those as claimed, in combination with patient data for detection of beta-cell destruction. Further, given the complexity of gene methylation expression profiles as acknowledged by Ahmed, even if a person was motivated to combine Lai and Ahmed as suggested by the Office, they would not expect to achieve the sensitivity and accuracy of detection as set forth above and described in the application as filed achieved by the use of machine learning (p. 16).
Regarding (b), Examiner respectfully disagrees. With regards to Applicant’s arguments that there is no teaching or suggestion to utilize ccfDNA gene methylation biomarkers in Ahmed, Examiner notes that one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986). In this case, Panagopoulou is used to teach the examination of circulating cell free DNA. Furthermore, Examiner notes that the sensitivity and accuracy described above of the instant invention is not apparent in the claims, and the data that is gathered is recited in a generic manner.
Applicant argues The Office acknowledges Lai and Ahmed fail to teach use of ccfDNA, but asserts Panagopoulou teaches ccfDNA for providing clinically relevant information for cancer related gene methylation as a biomarker for cancer diagnosis. Applicant respectfully disagrees. As previously noted in the response to non-final Office Action filed on September 18, 2025, Panagopoulou discusses cancer biomarkers, rather than cell/organ-specific gene methylation markers recited by Applicant's Claim 11. Panagopoulou discusses the detection of methylation on ccfDNA related to mutations and other cancer-related molecular events to build cancer-associated signatures. Cancers have a unique disease pathology where tumor cells hijack various cellular processes and can result in aberrant growth, undifferentiation and metastases throughout the body. When comparing ccfDNA concentrations alone, Panagopoulou discovered levels of ccfDNA are significantly higher in patients with tumors and metastases relative to healthy controls demonstrating a disease-specific impact of cancer on ccfDNA. Breast cancer, as discussed in Panagopoulou in relation to determining disease progression, is an entirely different tissue and unrelated disease compared to the instant claims predicting T2DM by measuring biomarkers in ccfDNA samples. There is no indication that a person of ordinary skill would look to Panagopoulou's description of cancer biomarkers of a specific cancer type to modify methods to measure beta-cell destruction. Even if a person of ordinary skill was motivated to look to Panagopoulou in view of Lai and Ahmed, there is no indication that the approach of utilizing ccfDNA in cancer could successfully be modified for early and specific detection of these biomarkers related to T2DM (p. 17).
Regarding (c), Examiner respectfully disagrees. In response to applicant’s argument that there is no teaching, suggestion, or motivation to combine the references, the examiner recognizes that obviousness may be established by combining or modifying the teachings of the prior art to produce the claimed invention where there is some teaching, suggestion, or motivation to do so found either in the references themselves or in the knowledge generally available to one of ordinary skill in the art. See In re Fine, 837 F.2d 1071, 5 USPQ2d 1596 (Fed. Cir. 1988), In re Jones, 958 F.2d 347, 21 USPQ2d 1941 (Fed. Cir. 1992), and KSR International Co. v. Teleflex, Inc., 550 U.S. 398, 82 USPQ2d 1385 (2007). In this case, the references Lai and Panagopoulou are analogous references because the references both address using biomarkers and identifier as information for formulating an analysis of a disease and identifying individuals at risk.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to OLIVIA R GEDRA whose telephone number is (571)270-0944. The examiner can normally be reached Monday - Friday 8:00am-5:00pm.
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, Peter H Choi can be reached at (469)295-9171. 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.
/OLIVIA R. GEDRA/Examiner, Art Unit 3681
/PETER H CHOI/Supervisory Patent Examiner, Art Unit 3681
Prosecution Insights