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
Applicant's amendment filed 03/31/2026 is acknowledged. Claims 1, 49-50, 52, 55-56, 60-61, and 63 have been amended. Claims 2-48 previously were cancelled. Claims 1 and 49-67 are pending in the instant application and the subject of this final office action.
All of the amendments and arguments have been reviewed and considered. Any rejections or objections not reiterated herein have been withdrawn in light of amendments to the claims or as discussed in this office action.
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Previous Rejection
Status of Prior Rejections/Objections:
The 112(a) written description rejection is withdrawn in view of the arguments and upon further search and consideration, including Wu, 2004, as cited in co-pending application number 18/344,616 and the prior art made of record in the conclusion below.
The 112(d) rejections of claims 49-57 and 60-67 are withdrawn in view of the amendments to claims 49 and 60 and upon further consideration.
The 101 rejections of claims 1 and 49-67 are maintained and modified/clarified in view of the amendments.
The prior art rejection(s) under 35 USC 102 directed to claim(s) 1 and 49-56 as being anticipated by Olsen are withdrawn in view of the amendments.
The prior art rejection(s) under 35 USC 103 directed to claim(s) 1 and 49-67 as being unpatentable over Olsen are maintained and modified in view of the amendments.
The double patenting rejections over co-pending application 18/344,616 in view of Olsen are maintained and modified in view of the amendments.
New Ground(s) of Rejections
The new ground(s) of rejections were necessitated by applicant’s amendment of the claims.
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claim Rejections - 35 USC § 101
Claims 1 and 49-67 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception without significantly more. The claim(s) recite(s) abstract ideas and natural phenomena. This judicial exception is not integrated into a practical application. The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception.
The following three inquiries are used to determine whether a claim is drawn to patent-eligible subject matter:
Step 1. Is the claim directed to a process, machine, manufacture, or composition of matter?
Yes, the claims are directed to process.
Step 2A, prong 1. Does the claim recite a law of nature, a natural phenomenon, or an abstract idea (recognized judicial exceptions)?
The claims are directed to classifying a liver disease of a subject based on a methylation pattern comprising assigning a liver disease state to the subject based on the methylation pattern, as recited in claim 1. Assigning encompasses the abstract idea of a mental process (i.e., evaluation of data or information to reach a conclusion or make a judgement).
Additionally, the assertion of assigning a liver disease state classification based on a methylation pattern also recites a natural phenomenon of the correlation between a liver disease and methylation statuses of a set of CpG sites of target sequence.
Step 2A, prong 2. Is the judicial exception(s) integrated into a practical application?
Regarding claims 1, the claim recites obtaining a set of CpG methylation status at a set of CpG sites of target sequences from a … sample of the subject by performing any or a combination of generically recited detection assays without obtaining a liver biopsy. MPEP 2106.04(d) discusses that mere data gathering of to obtain data input for a judicial exception (see MPEP 2106.05(g)) is not sufficient to integrate a judicial exception into a practical application. Selection of a particular type of data (i.e., subjects who are not undergoing a liver biopsy) also does not amount to an inventive concept as discussed in 2106.05(g).
Regarding claims 57-59, the claims recite that the methylation pattern further distinguishes between grades of fibrosis. The step of identification in claim 1, step (b) encompasses an abstract idea of a mental process so performing to identify additional categories fails to integrate the claims into a practical application.
Regarding claims 49 and 60, the claims recite first that the identifying is performed using a model and that places limitations directed to a choice of a subset of for the training data used. It is noted that models encompass calculations such as simple machine learning algorithms that may be performed by hand, i.e., an abstract idea.
In addition, choice of data, even should the data be novel, does not represent integration into practical application. See MPEP 2106.05(g) and e.g., SAP Am., Inc. v. InvestPic, LLC, 898 F.3d 1161, 1168 (Fed. Cir. 2018); Elec. Power, 830 F.3d at 1353 (“[W]e have treated collecting information, including when limited to particular content (which does not change its character as information), as within the realm of abstract ideas.” (citing Internet Pats. Corp. v. Active Network, Inc., 790 F.3d 1343, 1349 (Fed. Cir. 2015); OIP Techs., Inc. v. Amazon.com, Inc., 788 F.3d 1359, 1363 (Fed. Cir. 2015); Content Extraction, 776 F.3d at 1347; Digitech Image Techs., LLC v. Elecs. for Imaging, Inc., Case: 23-2437 Document: 51 Page: 14 Filed: 04/18/2025 RECENTIVE ANALYTICS, INC. v. FOX CORP. 15 758 F.3d 1344, 1351 (Fed. Cir. 2014); CyberSource Corp. v. Retail Decisions, Inc., 654 F.3d 1366, 1370 (Fed. Cir. 2011))).”
Regarding claims 50 and 61, the claims recite administering a therapy selected to treat a disease corresponding to the liver disease state. MPEP 2106.04(d) requires that a treatment be particular for such a limitation to integrate the claim into a practical application. The limitation of “a therapy selected to treat a disease” is not particular and encompasses treatments that have no more than a nominal or insignificant relationship to the disease corresponding to the liver disease state.
Regarding claims 51 and 62, the claims recite that the methylation pattern comprises particular modifications at sites of the target sequences. As with claims 57-59, such limitations further limit the calculations that encompass those performed by hand and thus fail to integrate the claims into a practical limitation.
Regarding claims 52-54 and 63-65, the claims limit the assignment step, such that it remains directed to an abstract idea, encompassing a mental process of forming a judgment (claims 52 and 63) or performing a calculation that could be performed by hand (claims 53-54 and 64-65), and thus also fail to integrate the claims into a practical limitation.
Regarding claims 55-56 and 66-67, the claims recite generic classes of well-known machine learning algorithms that are capable of being performed—at least on a small scale—by hand. See Pandey (Pandey D, et al. Machine learning algorithms: a review [Internet]. Mach Learn. 2019 [cited 2025 Sep 26]. Available from: https://d1wqtxts1xzle7.cloudfront.net/59817925/IRJET-V6I217620190621-73452-h5pwu7-libre.pdf; as cited in the IDS dated 04/01/2026); Artley (Artley B. Training a neural network by hand [Internet]. 2022 [cited 2025 Sept 26]. Available from: https://towardsdatascience.com/training-a-neural-network-by-hand-1bcac4d82a6e/; as cited in the IDS dated 04/01/2026); and Breiman (Breiman L. Random Forests [Internet]. University of California, Berkeley; 2001 Jan [cited 2025 Sep 26]. Available from: https://www.stat.berkeley.edu/~breiman/randomforest2001.pdf; as cited in the IDS dated 04/01/2026). For this reason, they too fail to integrate the claims into a practical application as these limitations encompass abstract ideas (e.g., mental processes).
Step 2B. Does the claim amount to significantly more?
The claims are directed to judicial exception(s). Further limitations themselves encompass abstract ideas, are directed to mere data gathering/selections of data, simple calculations, or are directed to treatment or machine learning techniques are recited at a high level (such that they encompass those that may be performed within the human mind with or without the aid of paper/generic computer), which are therefore equivalent to mere directions to apply the judicial exception. Therefore, no claim amounts to significantly more.
Claim Rejections - 35 USC § 103
Claim(s) 1 and 49-67 is/are rejected under 35 U.S.C. 103 as being unpatentable over Olsen (WO 2020/150258 A1; published 07/23/2020; cited on the IDS dated 03/13/2024).
Regarding claim 1, Olsen teaches a method of classifying a liver disease of a subject (entire document, e.g., Fig. 1) comprising:
performing a sequencing reaction (claims 26-27, 30, claim 79-80; para [0019-20]; see also at least [0103-104] for array, PCR, and variants of PCR) to obtain a set of CpG methylation status at a set of CpG sites of target sequences from a cfDNA sample (para [0017-18]; [0099]; [0194]; [0204]: “A cell-free biological sample is obtained…from a subject…processed and assayed for nucleic acid molecules (e.g., nucleic acid molecules derived from the cell-free biological sample)”; [0201]: “Methylation values are calculated for 10,508 analyzed CpGs”; Fig. 1)
identifying a methylation patterns based on the set of CpG methylation statuses, wherein the methylation pattern distinguishes between NAFL [instant “NALFD”], NASH, and cirrhosis (para [0194]: “Group A (…NAFL/Fibrosis F0-1), Group B (NASH/F0-1), and Group C (NASH/F3-4)”; [0193]: “a panel of cfDNA methylation markers that predict the presence of fibrosis and NASH is identified”; para [0178]: “based at least in part on a data set obtained using the methods…identifying the presence or absence of cirrhosis”; [0093]: “identifying liver disease by processing biological samples…subjects may include patients with NAFLD who are at increased risk of having SH and/or fibrosis” [i.e., those with NAFL]; [0202]; Table 3; [0154])
assigning a liver disease state to the subject based on the methylation pattern, wherein the liver disease state comprises one of: NAFL [instant “NALFD”], NASH, and cirrhosis (Fig. 1; para [0040]; “the presence or the severity of the liver disease is identified according to a Non-Alcoholic Fatty Liver Disease Activity Score [NAS] composite comprising steatosis, lobular inflammation, and hepatocellular ballooning”; [0154]: “an NAS value of >5 including both steatosis and hepatocyte ballooning is considered indicative of NASH”; [0178]: “identifying the presence or absence of cirrhosis”; [0072]; [0197]; claim 10).
Olsen teaches the current “imperfect gold standard” of liver biopsy” carries a non-negligible risk of complications, relative high costs, and sample biases (para [0002]) and that there is a need for non-invasive diagnostic tests that can effectively detect severity of liver diseases such as steatohepatitis and discriminate between different severities of liver disease (para [0003]).
Olsen teaches high concordance in tissue and plasma (para [0195]) and that the use of liver tissue is orthogonal validation of putative DMRs and demonstrates the utility of cfDNA analysis in detecting disease-specific signals (para [0196]).
Olsen teaches identifying the severity of the disease may include an identification of NAFLD, steatohepatitis, steatosis, fibrosis, or a combination thereof and may be based on differentially methylated regions in a dataset (para [0163]).
Olsen teaches configuring a machine learning algorithm to identify a disease and severity of the disease, including classifying apparently healthy subjects (para [0146]) and correctly identifying subjects known to have the severity of the liver disease (para [0145]) at a given sensitivity and specificity (para [0145-146]).
Olsen teaches that NAFLD exists as a spectrum ranging from NAFL to NASH and that patients may progress to cirrhosis (para [0002]).
Olsen teaches using machine learning algorithms that have more than binary outputs, including continuous output values that may indicate a presence and/or severity of a liver disease of a subject (para [0125-126]).
Olsen teaches identifying healthy subjects (para [0142], [0146], [0156], [0160]) and that a methylation profile [i.e., pattern] indicative of a healthy subject may be identified (para [0166]).
Olsen does not explicitly teach training an ML model to identify healthy, “NAFLD” [NAFL], NASH, and cirrhosis based off a training dataset with subjects of each classification.
However, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to have simply substituted the classifications of the embodiment of Olsen using fibrosis and steatohepatitis for a multi-class/continuous classifier using the disease spectrum identified in Olsen from healthy to at least cirrhosis (including training such a classifier on a dataset comprising samples from each) as such would have been a matter of substituting one set of classes for another to obtain predictable results, and would further be motivated by the use of at least healthy individuals to perform an assessment of sensitivity/specificity and/or the identification of the presence of the disease, as taught by Olsen, and to provide a diagnostic for NASH and associated conditions relevant the monitoring thereof, especially that avoids the risks and costs of biopsies, as taught and suggested by Olsen. Alternatively,/additionally, such would have been obvious to try as the number of disease classes in the spectrum is a finite number of predictable solutions, wherein the artisan would have been motivated by the need in the art identified by Olsen to discriminate between different severities of liver disease, where the liver disease may be NAFLD (i.e., including NAFL and NASH and may progress to F4/cirrhosis) and wherein Olsen teaches healthy samples for evaluating the machine learning model. Such would have been predictable at least because Olsen anticipates multi-class classifiers and teaches means of grading patients to meet disease criteria using liver biopsy to validate samples and orthogonally validate putative DMRs.
It further would have been obvious to the POSTIA before the EFD of the claimed invention to have avoided liver biopsies in the diagnostic method of Olsen at least because Olsen teaches that liver biopsies carry a risk of complications and bias and have high cost. Such would have been predictable as Olsen teaches liver biopsy methylation sites as orthogonal verification of the plasma methylation sites, such that the artisan would not anticipate the need for them in the “production” model once the methylation sites were verified/model trained.
Regarding claim 49, Olsen teaches processing a data set comprising a methylation profile of one or more genomic regions with a machine learning algorithm to identify a presence or severity of a liver disease in the subject (para [0204]) and the regions may be a subset of a plurality of CpG methylation statuses (para [0201]), and that the subset may be selected based on a feature subset search [i.e., feature selection] (para [0122-123]; [0151]; [0201]).
Regarding claim 57, Olsen does not explicitly teach identifying a methylation pattern distinguishes between F0-F3 and F4 fibrosis. It is noted that the disease state of cirrhosis comprises F4 (Table 2).
Olsen rectifies this by teaching that identifying the presence or severity of the liver disease comprises determining a presence or absence of fibrosis or significant fibrosis in the subject (para [0002]), where significant fibrosis may be defined according to the Non-Alcoholic Fatty Liver Disease Score criteria (para [0002]).
Olsen teaches identifying a presence or absence of cirrhosis in the subject (para [0025]) and that F4 on the Non-Alcoholic Fatty Liver Disease Score criteria corresponds to “cirrhosis” (Table 2).
Olsen teaches that the machine learning algorithm may comprise a binary output (para [0125]) and identifying [methylation] markers associated with a fibrosis severity of disease (para [0120]).
Olsen also teaches setting a threshold at F2 for significant fibrosis (para [0008]). Olsen teaches adjusting the parameters including cutoff values used to classify a sample to improve performance or other characteristics (para [0150]).
Olsen teaches that the present of liver fibrosis has become increasingly recognized as reliable predictive marker of NAFLD outcomes (para [0002]) and that there is a need for non-invasive diagnostic tests that can effectively detect and discriminate different severities of liver disease (para [0003]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to substitute the threshold of F4 for the threshold of F2 in the method of Olsen to identify markers that discriminate between cirrhosis and no cirrhosis on the NAS scoring system (i.e., F0-3 and F4), motivated by the desire to create a non-invasive diagnostic test for NASH cirrhosis in view of the ability to reliably predict NAFLD outcomes. There would have been a strong expectation of success as Olsen teaches performing and optimizing relevant machine learning operations on relevant data.
Alternatively, choosing the F4 threshold in the binary model of significant fibrosis would have been obvious to try as there was a demonstrated need in the market to discriminate different severities of liver disease, a finite number of identified thresholds for fibrosis where fibrosis could predictably predict outcomes, and the artisan had a reasonable expectation of success given the teachings of Olsen in performing and optimizing relevant machine learning operations on relevant data. The artisan would have been so motivated by the desire to further improve/tune the performance of the model.
It is further noted that a choice of threshold for “significant fibrosis” may also be considered a matter of routine optimization. See MPEP 2144.05(II).
Regarding claims 58-59, Olsen does not explicitly teach identifying a methylation pattern that further distinguishes between F0, F1, F2, F3, and F4. It is noted that the disease state of cirrhosis comprises F4 (Table 2; instant claim 59).
Olsen rectifies this by teaching that the output values may comprise continuous values that may indicate severity of a liver disease of a subject and may comprise Non-Alcoholic Fatty Liver Disease Activity Score of the subject (para [0126]).
Olsen teaches that the Non-Alcoholic Fatty Liver Disease Activity Score values comprise F0, F1, F2, F3, and F4 (Table 2; para [0194]). Olsen teaches adjusting the parameters including cutoff values used to classify a sample to improve performance or other characteristics (para [0150]).
Olsen teaches that the presence of liver fibrosis has become increasingly recognized as reliable predictive marker of NAFLD outcomes (para [0002]); that challenges may be encountered in using imaging and current blood-based tests to monitor modest changes in fibrosis progression (para [0003]; and that there is a need for non-invasive diagnostic tests that can effectively detect and discriminate different severities of liver disease (para [0003]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have tried using the “continuous” fibrosis scores from the NAS rather than a binary system with a threshold, in view of the recognized need for diagnostic tests that can differentiate between different severities of liver disease and the reliability of fibrosis as a predictable marker of NAFLD outcomes, wherein there are is a finite number of fibrosis categories in the NAS system and ways to choose a number of classes (i.e., 5, corresponding to F0, F1, F2, F3, and F4) within that. The artisan would have been so motivated by the desire to further improve/tune the performance of the model, as suggested by Olsen.
It is noted that the choice of the number of classes within the fibrosis classification is also, therefore, held to be a matter of routine optimization. See MPEP 2144.05(II).
Regarding claim 60, Olsen teaches processing a data set comprising a methylation profile of one or more genomic regions [i.e., methylation pattern] with a machine learning algorithm to identify a presence or severity of a liver disease in the subject (para [0204]) and the regions may be a subset of a plurality of CpG methylation statuses (para [0201]), and that the subset may be selected based on a feature subset search [i.e., feature selection] (para [0122-123]; [0151]; [0201]).
Olsen teaches a subset of a plurality of input values may be identified indicating the genomic regions whose methylation profiles are most influential or most important for making high quality classifications, such as identification of severity of a liver disease (para [0151]).
Olsen teaches adjusting the parameters of a trained machine learning algorithm including cutoff values used to classify a sample to improve/tune performance, accuracy, sensitivity, specificity, or other characteristics of identifying a presence and/or severity of the disease (para [0150]).
Thus, as discussed above in claim 58, it follows that this method would be applied to the F0, F1, F2, F3, and F4 classes such by adjusting parameters including cutoff values to classify a sample and choosing output classes such as the NAS values corresponding to the fibrosis values, Olsen would have identified methylation pattern distinguishes among them using a training dataset comprising a plurality of CpG methylation statuses based on feature selection, according to the same motivations as claim 58 as well as obtaining a high quality classification.
It is further noted that routine optimization would be applied to enable such subset to be capable of distinguishing between all of them. See MPEP 2144.05(II).
Regarding claims 50 and 61, Olsen teaches that the methods may also be used as a part of assessing a therapeutic regimen comprising proving a therapeutic regimen to a subject based on the identification of the presence and/or severity of the liver disease (para [0109-110]).
Regarding claims 51 and 62, Olsen teaches that the methylation pattern may be derived from whole genome bisulfite sequencing (para [0194] and [0201]) and that a bisulfite conversion process converts unmethylated cytosines and optionally 5hmC to uracils while preserving at least 5mC (para [0101]). Olsen teaches that the absolute or relative amount of methylation within nucleic molecules may be determined by sequencing and processed to identify hypo and/or hypermethylated regions of the one or more genomic regions (para [0102]).
Thus, Olsen teaches that the methylation pattern comprises at least 5mC at individual sites of the target sequences.
Regarding claim 52 and 63, Olsen teaches that continuous output values that may indicate a presence and/or severity or a liver disease such as the NAS score of a subject may comprise a probability value (para [0126]). As the NAS score is linked to a threshold that indicates NASH or NAFL, as discussed above, Olsen teaches that the liver disease state may be a probability (instant claim 52).
It also follows that the fibrosis NAS score, with chosen thresholds as discussed above, may also be probability (instant claim 63).
Regarding claims 53 and 64, Olsen teaches that a machine learning algorithm may be configured to identify a disease and/or a severity of the disease with a sensitivity of at least about 5-99%, including at least about 75% (para [0145] and [0056]).
Olsen further teaches that the algorithm may be adjusted or tune to improve performance, sensitivity, specificity, or AUC of identifying a presence and/or severity of the disease (para [0150]), wherein the algorithm may be adjusted or tune by configuring parameters or during or after the training process (para [0150]).
Olsen likewise teaches ROC curves comprising TPR [sensitivity] vs. FPR [specify] and an area under the curve (AUC) [Fig. 3-4].
Thus, while Olsen does not explicitly show the assignment of the liver disease state with sensitivity in the embodiments where the liver disease state is i) NASH or NALFD or ii) one of F0-F4, the choice of optimizing the model until a sensitivity for assignment of 75% is reached would have been obvious to one of ordinary skill in the art before the effective filing date, motivated by the desire to improve the performance of the model, as taught by Olsen. There would have been a strong expectation of success as Olsen teaches means of optimizing machine learning models to improve sensitivity.
Further, in view of the relationship between sensitivity and specificity demonstrated by the ROC curve and the methods of optimizing each detailed by Olsen, picking a sensitivity threshold for the model for assigning a disease state is likewise held to be matter of routine optimization. See MPEP 2144.05(II).
Regarding claims 54 and 65, Olsen teaches that a machine learning algorithm may be configured to identify a disease and/or a severity of the disease with a specificity of at least about 5-99%, including at least about 75% (para [0146] and [0056]).
Olsen further teaches that the algorithm may be adjusted or tune to improve performance, sensitivity, specificity, or AUC of identifying a presence and/or severity of the disease (para [0150]), wherein the algorithm may be adjusted or tune by configuring parameters or during or after the training process (para [0150]).
Olsen likewise teaches ROC curves comprising TPR [sensitivity] vs. FPR [specify] and an area under the curve (AUC) [Fig. 3-4].
Thus, while Olsen does not explicitly show the assignment of the liver disease state with a sensitivity and specificity in the embodiments where the liver disease state is i) NASH or NALFD or ii) one of F0-F4, the choice of optimizing the model until a sensitivity and specificity for assignment of 75% is reached would have been obvious to one of ordinary skill in the art before the effective filing date, motivated by the desire to improve the performance of the model, as taught by Olsen. There would have been a strong expectation of success as Olsen teaches means of optimizing machine learning models to improve sensitivity and specificity.
Further, in view of the relationship between sensitivity and specificity demonstrated by the ROC curve and the methods of optimizing each detailed by Olsen, picking a sensitivity and specificity threshold for the model for assigning a disease state is likewise held to be matter of routine optimization. See MPEP 2144.05(II).
Regarding claims 55-56 and 66-67, Olsen teaches that the trained algorithm may comprise a random forest (para [0058] and [0201]; instant claim 55 and 66) or a neural network (para [0058]; instant claims 66-67).
Claims 1 are 49-67 are provisionally rejected on the ground of nonstatutory double patenting as being unpatentable over claims 1-14, 20-27, 31-35, and 38 of copending Application No. 18/344,616 in view of Olsen (WO 2020/150258 A1; published 07/23/2020; cited on the IDS dated 03/13/2024). This is a provisional nonstatutory double patenting rejection.
Both sets of claims are directed to a method of classifying a liver disease of a subject comprising obtaining a set of CpG methylation status in a cfDNA sample of a subject; identifying a methylation pattern based on the set of CpG methylation statuses, wherein the methylation pattern distinguishes between states that comprise non-liver disease/healthy state, NAFLD, NASH, and cirrhosis.
Claim 26 of ‘616 recites sequencing and claim 27 recites a DNA array, PCR, qPCR, and digital PCR.
Claim 1 of ‘616 recites identifying a methylation pattern by selecting a set of CpG sites based on feature selection such that the methylation is predictive of non-liver disease /healthy state, NAFLD, NASH, or cirrhosis. Thus, it teaches a pattern that distinguishes of non-liver disease /healthy state, NAFLD, NASH, and cirrhosis.
Claim 1 of ‘616 further teaches administering a therapy to treat the disease states. The claims of ‘616 further teach at least determining the presence of 5mc/5hmC; a probability belonging to the states; and further classifying a stage of fibrosis and a probability of F0, F1, F2, F3, or F4 fibrosis.
The ‘616 application fails to teach a sensitivity or specificity; that the model comprises a random forest or neural network; that the pattern distinguishes F0-F3 and F4 or each of F0-F4; and that the CpG methylation statuses are obtained without obtaining a liver biopsy.
Any additional limitations of the ‘616 claims are encompassed by the open claim language “comprising” found in the instant claims.
Olsen teaches the current “imperfect gold standard” of liver biopsy” carries a non-negligible risk of complications, relative high costs, and sample biases (para [0002]) and that there is a need for non-invasive diagnostic tests that can effectively detect severity of liver diseases such as steatohepatitis and discriminate between different severities of liver disease (para [0003]).
Olsen teaches high concordance in tissue and plasma (para [0195]) and that the use of liver tissue is orthogonal validation of putative DMRs and demonstrates the utility of cfDNA analysis in detecting disease-specific signals (para [0196]).
Olsen teaches identifying the severity of the disease may include an identification of NAFLD, steatohepatitis, steatosis, fibrosis, or a combination thereof and may be based on differentially methylated regions in a dataset (para [0163]).
Olsen teaches assigning a liver disease state with at least about 75% sensitivity and specificity, as discussed and cited in the rejection of claims 53-54 and 64-65 above, with the noted routine optimization. Olsen teaches that the model may comprise a random forest or neural network, as cited in claims 55-56 and 66-67 above.
Olsen further teaches assigning a threshold in a binary classifier for fibrosis and that the cutoff may be optimized, as cited and discussed in claim 57. Olsen teaches classifying based on “continuous” values such as NAS scores and the importance of monitoring small changes in fibrosis, as cited and discussed in claims 58-59.
Olsen teaches that machine learning algorithms may be adjusted or tuned to improve the performance or other characteristics (para [0150]) and that a subset of inputs is identified as most important to be included for making high quality classifications (para [0151]).
Therefore, it would have been obvious before the effective filing date to combine ‘616 with the method of Olsen, motivated by the desire to improve the performance of the method to make high quality classifications. It further would have been obvious, for the same reasons as cited in claims 57-59 to modify the method of ‘616 in view of Olsen to identify a methylation pattern that distinguishes F0-F3 from F4 and each of F0-F4, for the same motivations as above. There would have been a strong expectation of success as both are directed to machine learning methods using methylation datasets on diseases on the NAFLD spectrum.
It further would have been obvious to the POSTIA before the EFD of the claimed invention to have avoided liver biopsies in the combined method at least because Olsen teaches that liver biopsies carry a risk of complications and bias and have high cost. Such would have been predictable as Olsen teaches liver biopsy methylation sites as orthogonal verification of the plasma methylation sites, such that the artisan would not anticipate the need for them in the “production” model once the methylation sites were verified/model trained.
Response to Arguments
Applicant's arguments filed 03/31/2026 have been fully considered but they are not persuasive.
Regarding the art and double patenting rejections over Olsen, on pg. 9-11, Applicant argues that Olsen does not teach the amended claims because Olsen does not use healthy subjects to identify methylation patterns but instead uses patients from NAFLD with little to fibrosis, NASH with little to fibrosis, and NASH with moderate to severe fibrosis. Applicant alleges Olsen would not have been able to distinguish a healthy subject from a subject not having NASH such as a patient with NAFLD. Applicant argues that the NAS score of Olsen is “obtained at least in part upon a liver biopsy analysis” which is precluded by the amended claim.
In response to the arguments directed to healthy subjects, while Olsen does not utilize healthy subject in the classifier of Example 1, Olsen does teach correctly having identified healthy subjects and/or subject without a liver disease in para [0142], [0146], [0156], [0160] and discusses methylation profiles [i.e., patterns] of healthy subjects in para [0166]. As discussed in the amended rejection above, it would have been obvious and predictable to substitute the classification outputs of that classifier for the spectrum of healthy, NAFL, NASH, and cirrhosis, as taught by Olsen, and/or obvious to try a trained model with those classifiers, wherein it also would have been predictable to utilize only the cfDNA of Olsen at least because of the need in the art taught by Olsen for a diagnostic for NASH and associated conditions to enable effective monitoring by non-invasive diagnostics, as suggested by Olsen. Applicant’s arguments directed to the knowledge of the artisan of translating pairwise methylation patterns into multi-class classifications are also noted. See also Wu (Wu, 2004), as cited below in the Conclusion.
Arguments that Olsen would not have been able to distinguish a healthy subject are not persuasive as Olsen explicitly contemplates assessing the accuracy of machine learning models using healthy subjects and teaches various metric for assessing such models, as cited in the paragraph above. Further, Olsen teaches multiple embodiments including obtaining a predicted NAS score in other embodiments based on the methylation sites (e.g., para [0163]). It is also noted that the amended claims do not preclude use of liver biopsy in the training.
Further, where Applicant may be alleging the inoperability of the prior art, it is noted that arguments by applicant cannot take the place of evidence in the record, as discussed in MPEP 716.01(c)(II). See also MPEP 716.07 on inoperability.
Regarding the 101 rejections, on pg. 6-9, Applicant argues that the amended claims integrate alleged judicial exception via the “without a liver biopsy”, which acts as a prophylaxis to prevent a procedure to obtain a liver biopsy to distinguish between the liver classification states. Applicant further argues that the amended claims recite “non-routine and unconventional elements” and argues that “tangibility is not necessary for eligibility under the Alice/Mayo test”. Applicant argues, alone or in combination, it was not routine or conventional at the time of filing to classify a liver disease state without obtaining a liver biopsy from the subject. Applicant additionally cites the 08/04/2025 memo on subject matter eligibility that a claim does not recite a mental process when it cannot be practically performed in the human mind, such as certain machine learning steps, and argues that amended claim 1 cannot practically be performed in the human mind.
In response to the argument regarding the integration of the judicial exception via the “without a liver biopsy” acting as a prophylaxis, the limitation applies regardless of the judicial exception. As it encompasses all applications of the judicial exception(s), it does not integrate the claim under the exception identified in MPEP 2106.04(d)(2) as it is not particular to the JE. Further, such serves to select the particular type of data used in the JE and thus serves merely as extra-solution activity.
Where arguments are directed to non-routine and conventional elements, the choice of obtaining the data necessary to perform the judicial exception without a biopsy encompasses a selection of a data type. As discussed in MPEP 2106.05(g), selection of a particular data source of type to be manipulated does not amount to an inventive concept, particularly when the data gathering (under Step 2B) is also well known and conventional, e.g., generically recited amplifying and sequencing nucleic acids (see MPEP 2106.05(d)(II)), which has been taught in at least Olsen (e.g., orthogonal cfDNA), as previously cited. The choice of classification bins, which is central to the judicial exception(s), cannot be significantly more, as discussed in MPEP 2106.05(I): An inventive concept "cannot be furnished by the unpatentable law of nature (or natural phenomenon or abstract idea) itself. Instead, an "inventive concept" is furnished by an element or combination of elements that are recited in the claim in addition to (beyond) the judicial exception.
Thus, there remains no more than the judicial exception(s) and insignificant extra-solution activity. As such, there is not significantly more than the judicial exception(s).
Regarding the practicality of performing machine learning steps “in the human mind”, as discussed in MPEP 2106.04(a)(2)(III), the courts do not distinguish between mental processes that are performed entirely in the human mind and mental processes that require a human to use a physical aid or that recite process performed on a (generic) computer. The claim does not require any particular number of statuses, patterns, etc. that would render the method impractical to perform in the human mind. As taught by AI by Hand (AI by Hand. One-node neural network by hand [Internet]. YouTube; 2026 [cited 2026 Oct 14]. Available from: https://www.youtube.com/shorts/Lnk0f16fkv0. [0:00-2:44], especially [2:39]), a trained machine learning model such as a neural network may be calculated at each node by hand using a simple algorithm comprising addition and multiplication. See also: Pandey, 2019; Artley, 2022; and Breiman, 2001, as cited in the 101 above.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Wu (Wu, Ting-Fan et al., Probability Estimates for Multi-class Classification by Pairwise Coupling. Journal of Machine Learning Research. 5: 975-1005 (2004)) teaches pairwise coupling is a popular multi-class classification method (Abstract) and that many authors propose using two-class classifiers for multi-class classification (Introduction, para 1).
Hardy (Hardy, 2017; as cited on the IDS dated 03/13/2024) teaches non-invasively stratifying liver fibrosis severity in patients with NAFLD based on patients with different severities and healthy patients (Abstract). Hardy teaches that similar results were found in patients with cirrhosis vs. [healthy] controls (pg. 1324, col 1, para 1; Abstract). Hardy suggests assessing the biomarker for early NAFLD stages [i.e., NAFL, instant NAFLD] (pg. 1327, col 1, para 2).
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 Emma R Hoppe whose telephone number is (703)756-5550. The examiner can normally be reached Mon - Fri 11:00 am - 7:00 pm.
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, Anne Gussow can be reached at (571) 272-6047. 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.
/EMMA R HOPPE/Examiner, Art Unit 1683
/NANCY J LEITH/Primary Examiner, Art Unit 1636
