DETAILED ACTION
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Herein “the previous Office action” refers to the Non-Final Rejection filed 12/23/2025.
Amendments Received
Amendments to the claims were received on 3/23/2026, and have been entered.
Priority
As detailed on the Filing Receipt filed 6/16/2021, the instant application claims priority to as early as 1/19/2018. At this point in prosecution, the claims are accorded the earliest claimed priority date.
Information Disclosure Statement
The Information Disclosure Statements filed on 2/3/2026, 3/19/2026, and 5/28/2026 are in compliance with the provisions of 37 CFR 1.97 and have been considered in full. Signed copies of the IDS are included with this Office Action.
Claim Status
Claims 1-9, 13, 17 and 24-46 are canceled.
Claims 10-12, 14-16 and 18-23 are pending, and under examination.
Withdrawn Objections/Rejections
The rejection of claims 10-12, 14-16 and 18-20 under 35 USC § 103, as being unpatentable over Martin, in view of Al-Murrani, Friesen and Struck, is hereby withdrawn in view of Applicant’s amendment of the claims and persuasive argument that the applied art does not teach amended limitations pertaining to reduced trajectories of training data and mixture-of-experts learning (Remarks filed 3/23/2026 at pg. 6, para. 4 – pg. 8, para. 1).
The rejection of claim 21 under 35 USC § 103, as being unpatentable over Martin, in view of Al-Murrani, Friesen, Struck and Oehmcke, is hereby withdrawn in view of Applicant’s amendment of the claims and persuasive argument that Oehmcke does not remedy the noted deficiency of the other art (pg. 8, para. 4).
The rejection of claims 22-23 under 35 USC § 103, as being unpatentable over Martin, in view of Al-Murrani, Friesen, Struck and Lipton, is hereby withdrawn in view of Applicant’s amendment of the claims and persuasive argument that Lipton does not remedy the noted deficiency of the other art (pg. 8, para. 4 – pg. 9, para. 1).
The provisional rejection of claims 10-12 and 16 on the ground of nonstatutory double patenting as being unpatentable over claims of co-pending Application No. 17/250,389, in view of Al-Murrani, Friesen and Martin, is hereby withdrawn in view of Applicant’s amendment of the claims and persuasive argument that Martin, Al-Murrani and Friesen fail to teach or suggest the amended claim 10 as detailed (pg. 9, paras. 3-4).
Claim Interpretation
The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language, in light of the specification, as it would be understood by one of ordinary skill in the art (MPEP 2111-2111.01).
Claim 10 recites a number of ‘wherein’ clauses, including the following:
“wherein the feline has a probability score based on age and biomarker levels of one or more biomarkers, the biomarker levels comprising urine specific gravity level, a creatinine level, a urine protein level, a blood urea nitrogen (BUN) or urea level, a white blood cell count (WBC), or urine pH” (lines 10-13);
“wherein the probability score is obtained by executing a machine learning algorithm to transform the age and the biomarker levels to determine the probability score indicating that the feline has or is likely to develop CKD” (lines 14-14-16);
“wherein the machine learning algorithm was trained using a mixture-of-experts (MOE) learning technique based [on] an ensemble of a plurality of prediction models built from a plurality of respective sampled subsets of a training dataset, the training dataset comprising,. for each sample feline diagnosed with CKD, medical information associated with a plurality of reduced trajectories, each reduced trajectory being a temporally ordered subset of clinical visit records truncated at a point up to two years prior to a CKD diagnosis of the sample feline, and wherein each of the sampled subsets comprises, for each sample feline, a single reduced trajectory randomly selected from the plurality of reduced trajectories” (lines 17-24); and
“wherein the probability score is greater than a threshold value indicating the feline has or is at risk of developing CKD, wherein the threshold value is determined by applying a linear discriminant analysis to a plurality of medical records of a plurality of felines, wherein the plurality of medical records comprise ages and measurements of the one or more biomarkers” (lines 28-31).
Claim scope is not limited by claim language that suggests but does not require steps to be performed, or by claim language that does not limit a claim to a particular structure. It is unclear if the above clauses, as written, require steps to be performed or limit the particular structure of the method. See ‘Claim Rejections – 35 USC § 112’ section for further details.
For purposes of applying prior art, the above clauses are interpreted as structurally limiting (e.g., requiring a user of the method to obtain a probability score in the recited manner). Please note that, were the above clauses not interpreted as limiting, numerous pieces of prior art applied herein would be considered unnecessary to render the claims as obvious. For example, independent claim 10 would be considered obvious under 35 USC § 103, over the combination of Al-Murrani and Friesen alone.
Claim Objections
Claim 10 is objected to because of the following minor informality:
With respect to claim 10, the recited phrase “based an ensemble” (line 18) should be amended to, e.g., “based on an ensemble”.
Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 USC § 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
Claims 10-12, 14-16 and 18-23 rejected under 35 USC § 112(b) as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor, or a joint inventor, regards as the invention. The new grounds of rejection presented herein were necessitated by Applicant’s amendment of the claims (filed 3/23/2026) to introduce further ‘wherein’ language, prompting reconsideration of claim scope as a whole.
With respect to claim 10 and dependents thereof, there is uncertainty regarding scope of the claim. The claim is directed to a method of treating chronic kidney disease (CKD) in a feline in need thereof, comprising administering a particular nutrient-supplemented treatment diet to the feline in need thereof. The claim further recites numerous ‘wherein’ clauses having uncertain limiting effect on claim scope, including the following:
“wherein the feline has a probability score based on age and biomarker levels of one or more biomarkers, the biomarker levels comprising urine specific gravity level, a creatinine level, a urine protein level, a blood urea nitrogen (BUN) or urea level, a white blood cell count (WBC), or urine pH” (lines 10-13);
“wherein the probability score is obtained by executing a machine learning algorithm to transform the age and the biomarker levels to determine the probability score indicating that the feline has or is likely to develop CKD” (lines 14-14-16);
“wherein the machine learning algorithm was trained using a mixture-of-experts (MOE) learning technique based [on] an ensemble of a plurality of prediction models built from a plurality of respective sampled subsets of a training dataset, the training dataset comprising,. for each sample feline diagnosed with CKD, medical information associated with a plurality of reduced trajectories, each reduced trajectory being a temporally ordered subset of clinical visit records truncated at a point up to two years prior to a CKD diagnosis of the sample feline, and wherein each of the sampled subsets comprises, for each sample feline, a single reduced trajectory randomly selected from the plurality of reduced trajectories” (lines 17-24); and
“wherein the probability score is greater than a threshold value indicating the feline has or is at risk of developing CKD, wherein the threshold value is determined by applying a linear discriminant analysis to a plurality of medical records of a plurality of felines, wherein the plurality of medical records comprise ages and measurements of the one or more biomarkers” (lines 28-31).
For example, it is unclear if the language “the probability score is obtained by executing a machine learning algorithm… [which] was trained using a mixture-of-experts (MOE) learning technique” (lines 14-18) actively requires a user of the claimed method to train the machine learning algorithm and obtain the probability score by executing said algorithm, in the recited manner, or whether this language is merely informative regarding the provenance of an associated probability score.
Moreover, a feline “in need []of” receiving a diet designed to treat CKD encompasses any feline that has, or is at risk of developing, CKD. It is unclear how the above requirements, e.g., that the feline “has a probability score” (line 10) that “is greater than a threshold value indicating that the feline has or is at risk of developing CKD” (lines 28-29), further limit the subject feline. The intangible quality of “ha[ving] a probability score”, of the recited provenance and significance, does not appear to materially alter the structure of the feline.
Thus, the scope of the claims is indefinite.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 USC §§ 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 USC § 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 USC § 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 USC § 102(b)(2)(C) for any potential 35 USC § 102(a)(2) prior art against the later invention.
Claims 10-12, 14-16 and 18-20 are rejected under 35 USC § 103 as being unpatentable over Martin et al (US 2015/0178639; published 7/25/2015; previously cited), in view of Al-Murrani (US 2013/0344196; published 12/26/2013; previously cited) , Friesen et al (US 2009/0197804; published 8/6/2009; previously cited) Struck et al (US 2012/0003752; published 1/5/2012; previously cited), Maathuis et al (Preproceedings of the 29th Benelux Conference on Artificial Intelligence (BNAIC 2017), pp. 326-340, University of Groningen; published 2017) and Wang et al (arXiv 1611.04578v1 [cs.LG]; published 11/14/2016). The new grounds of rejection presented herein were necessitated by Applicant’s amendment of the claims (filed 3/23/2026) to incorporate new limitations.
Claim 10 recites “[a] method for treating chronic kidney disease (CKD) in a feline in need thereof… comprising”:
a) “administering a treatment diet to the feline in need thereof, the treatment diet comprising a potassium supplement diet having a potassium level between 0.1 mg/1000 kcal and 50 mg/1000 kcal and at least one of”: diets including levels within specific recited ranges of phosphorous, protein, sodium, potassium or polyunsaturated fatty acids, an anti-oxidant supplement diet, or a vitamin B supplement diet;
b) “wherein the feline has a probability score based on age and biomarker levels of one or more biomarker levels… comprising urine specific gravity level, a creatinine level, a urine protein level, a blood urea nitrogen (BUN) or urea level, a white blood cell count (WBC), or urine pH;
c) “wherein the probability score is obtained by executing a machine learning algorithm to transform the age and the biomarker levels to determine the probability score indicating that the feline has or is likely to develop CKD”;
d) “wherein the machine learning algorithm was trained using a mixture-of-experts (MOE) learning technique based [on] an ensemble of a plurality of prediction models built from a plurality of respective sampled subsets of a training dataset, the training dataset comprising, for each sample feline diagnosed with CKD, medical information associated with a plurality of reduced trajectories, each reduced trajectory being a temporally ordered subset of clinical visit records truncated at a point up to two years prior to a CKD diagnosis of the sample feline, and wherein each of the sampled subsets comprises, for each sample feline, a single reduced trajectory randomly selected from the plurality of reduced trajectories”; and
e) “wherein the probability score is greater than a threshold value indicating the feline has or is at risk of developing CKD, wherein the threshold value is determined by applying a linear discriminant analysis to a plurality of medical records of a plurality of felines, wherein the plurality of medical records comprise ages and measurements of the one or more biomarkers”.
With respect to claim 10, Martin discloses “methods for correcting a data set and classifying the data set” (Abstract) comprising:
b) “output[ting]… a set of predicted classifications” (pg. 6, para. 0042) for “data received… [that] may represent any experimental or otherwise obtained data from which a classification may be drawn, such as… a variety of phenotypic characteristics such as levels of any biologically significant analyte… [e.g.] the amount of a… polypeptide” wherein “each data set comprises a plurality of patient samples… Sample types include… urine… [and] can be derived from a living organism” (pg. 4, paras. 0026 and 0028-30), i.e., generating patient probability scores based on biomarker levels comprising a urine protein level;
c) wherein “the processor generates a first classifier for the training data set by applying a machine learning technique to the training data set and the training class set, and generates a first test class set by classifying the elements in the test data set according to the first classifier” (pg. 1, para. 0008) and “classes… can include… presence/absence of a disease in the source of the sample… risk for a disease; [and] likelihood of recurrence of disease” (pg. 4, para. 0028), i.e., classifications may be obtained by executing a machine learning algorithm upon the biomarker levels to determine a probability score indicating that a subject has or is likely to develop a disease;
d) wherein “the data sets may include expression level data for a disease condition and for a control condition… [and] include data samples that each correspond to one of a plurality of biological state classes… includ[ing]… presence/absence of a disease in the source of the sample” (pg. 4, para. 0028), i.e., a training dataset may comprise information relating to biomarker levels from sets of subjects that have been, and have not been, diagnosed with a disease and “The classification engine may use any one or more known machine-learning algorithms… including… Random Forest techniques” (pg. 5, para. 0032), i.e., the applied machine learning algorithm may build an ensemble comprising a plurality of prediction models; and
e) wherein “a threshold may be set such that [a] first test class set and [a] second test class set are said to differ if at least a predetermined number of elements in the first test class set differs from the corresponding elements in the second test class set” (pg. 2, para. 0011) and “The classification engine may use… linear discriminant analysis techniques” (pg. 5, para. 0032), i.e., classifications may be discriminated based on comparison to a threshold value that may be determined by applying a linear discriminant analysis to the received data.
Martin states that “the data sets may be relatively heterogeneous when considering characteristics outside of the characteristic defining the biological state class. Factors contributing to heterogeneity include, but are not limited to, biological variability due to… age” (pg. 4, para. 002). In this way, Martin at least suggests analysis of subject age.
Martin does not specifically disclose combined analysis of age and biomarker levels of felines; or determining classifications indicating subject risk of developing CKD. Neither does Martin disclose administering a treatment diet as claimed; combined analysis of age and biomarker levels; training using a mixture-of-experts learning technique as claimed; nor training on sampled subsets of reduced trajectories as claimed.
Al-Murrani discloses “a method for… prognosis of kidney disorder in a feline” (pg. 16, para. 0204), and teaches “utilizing at least one relevant biomarker isolated and measured from a biological test sample taken from such feline” (pg. 2, para. 0021), wherein “the level of expression of the one or more biomarkers… is normalized” (pg. 6, para. 0063). Al-Murrani exemplifies “acute or chronic abnormal loss of kidney function, such as renal failure” (pg. 10, para. 0157), i.e., CKD, and “measurements… including… urinary protein levels” (pg. 3, para. 0022).
Al-Murrani further teaches application comprising “predict[ing], detect[ing], and diagnos[ing] in such feline a change from a normal state to an abnormal state leading to a kidney disorder” (pg. 13, para. 0189), and teaches that their method is “predictive of a decline in renal function, as might otherwise be diagnosed at a later time” (pg. 3, para. 0023).
Additionally, Al-Murrani teaches “providing a kidney protective diet… to a feline having a kidney disorder as diagnosed or diagnosable by” i.e., responsive to, “the method[s]” (pg. 6, paras. 0079-0080), wherein “the kidney protective diet comprises one or more of the following modifications relative to a standard feline diet: Reduced phosphorus, Reduced levels of protein, Reduced sodium, Increased levels of omega-3 fatty acids”, i.e., polyunsaturated fatty acids, “Increased levels of B-complex vitamins, [and] Increased antioxidants” (pg. 6, paras. 0088-0094).
In particular, Al-Murrani exemplifies a treatment diet providing “from about 3.6 to about 7.9 g/100 kcal ME protein” (pg. 6, para. 0096), i.e., a protein level between about 36 to about 79g/1000 kcal. This disclosed range overlaps the claimed specific range of a protein level less than 70g/1000 kcal. Where prior art discloses a range which overlaps or approaches a claimed range, the prior art disclosure renders the claimed range as obvious (MPEP 2144.05 § I).
Additionally, Al-Murrani exemplifies a diet providing “from about 0.04 to about 0.17 g/100 kcal ME phosphorus” (pg. 6, para. 0096), i.e., a phosphorus level between about 0.4 to about 1.7 g/1000 kcal. This disclosed range falls within (i.e., overlaps) the claimed specific range of a phosphorus level between 0.1 g/1000 kcal and 20 g/1000 kcal.
Additionally, Al-Murrani exemplifies a diet providing “from about 0.008 to about 0.07 g/100 kcal ME sodium” (pg. 6, para. 0096), i.e., a sodium level between about 8 mg to about 70 mg/1000 kcal. This disclosed range falls within (i.e., overlaps) the claimed specific range of a sodium level between 0.1 mg/1000 kcal and 50 mg/1000 kcal.
Al-Murrani does not teach administering a treatment diet providing a potassium level between 0.1 mg/1000 kcal and 50 mg/1000 kcal. Neither does Al-Murrani teach combined analysis of age and biomarker levels; training using a mixture-of-experts learning technique as claimed; nor training on sampled subsets of reduced trajectories as claimed.
Friesen discusses administration of food compositions for preventing or treating kidney disease (Abstract), wherein the patient is preferably a canine or feline (para. 0013), and exemplifies food compositions comprising one or more food ingredients containing protein in amounts of from about 5% to about 30%, sodium in amounts of from about 0.01% to about 2%, potassium in amounts of from about 0.01% to about 2%, and phosphorus in amounts of from about 0.2 to about 1% dry matter weight (para. 0025). The low end of the disclosed range of potassium, 0.01% potassium, is equivalent to 100 mg potassium per 1 kg dry matter.
Friesen also mentions a utilized standard of 3.3 kcal/g dry food (para. 0054, Table caption), which equals 1000 kcal per about 0.303 kg dry matter. According to this standard, 0.01% potassium by dry weight would be equivalent to a potassium level of about 30.3 mg/1000 kcal. Friesen thus teaches administering a treatment diet having a potassium level of about 30.3 mg/1000 kcal, which falls within the claimed range of 0.1 mg/1000 kcal to 50 mg/1000 kcal.
Friesen teaches combination of the discussed food compositions with one or more renal diagnostic devices and methods for evaluating the presence and severity of renal dysfunction, including assessment of urine specific gravity, creatinine levels, albuminuria, proteinuria, blood urea nitrogen (BUN) concentration and/or urine pH (pg. 5, para. 0043). Friesen also presents study results indicating that administering food compositions containing relatively low levels of protein, sodium and potassium as disclosed is beneficial for preventing and/or treating feline kidney disease (pg. 7, para. 0059; pg. 8, para. 0062; pg. 9, para. 0068).
Friesen does not describe analysis of feline age, and thus does not teach combined analysis of age and biomarker levels. Neither does Friesen teach training using a mixture-of-experts learning technique as claimed; nor training on sampled subsets of reduced trajectories as claimed.
Struck discusses “assays and in vitro methods for prediction of the progression of primary chronic kidney disease (CKD)” (Abstract) and teaches that teaches “prediction of the progression… may be improved by… determining and using the level of… laboratory parameter[s]… selected from… creatinine… [or] proteinuria… [and] at least one clinical parameter… selected from… age” (pg. 3, para. 0039). Struck specifies that “the term ‘patient’ as used herein refers to a living human or non-human organism… Thus, the methods and assays described herein are also applicable to both human and veterinary disease” (pg. 6, para. 0089).
Struck exemplifies determination of cut-off values through “Kaplan-Meier curve analyses of… patients with CKD… stratified into two groups according to the median of [biomarker levels] at baseline” (pg. 8, para. 0111). Struck depicts their results in the form of “Kaplan-Meier plots (FIG. 11-14), where[in] the occurrence of events i.e. in the present case: kidney disease progression, over time is depicted for… the investigated patient population… separated in two subgroups each” (pg. 4, para. 0069).
In this way, Struck advantageously teaches combined analysis of age and biomarker levels of non-human subjects for prediction of kidney disease progression to CKD. Struck does not teach training using a mixture-of-experts learning technique as claimed; nor training on sampled subsets of reduced trajectories as claimed.
Maathuis evaluates “neural network approaches to perform a regression prediction task on chaotic time series”, and presents a modeling framework where “Different NNs”, neural networks, are “MLPs”, multilayer perceptrons, “Residual MLPs and Long Short-Term Memory (LSTM), are… embedded within a larger architecture, the Hierarchical Mixture of Experts (HME)” (pg. 326, Abstract; pg. 327, para. 3).
Maathuis discusses the difficulty and cross-field utility of chaotic time-series prediction, stating that “Many disciplines are concerned with time series prediction… Most problems in nature deal with nonlinear or chaotic data, which poses a challenge” (pg. 326, para. 1). Maathuis presents findings that their HME architecture outperforms considered baseline models on sparse time series data with high predictive accuracy to ground truth (pg. 337, Fig. 3c; pg. 338, Fig. 4).
Maathuis does not teach training on sampled subsets of reduced trajectories as claimed.
Wang presents “Earliness-Aware Deep Convolutional Networks (EA-ConvNets), an end-to-end deep learning framework, for early classification of time series data” (pg. 1, Abstract). Wang discusses implementation of the earliness-aware component of their architecture via a stochastic truncation process, wherein an index (
s
) is sampled from a geometric timestamp distribution (P) and applied to each time series (
T
i
) to generate a new, stochastically truncated time series (
T
i
↓
s
=
t
s
1
,
…
,
t
i
s
), which then forms the input for subsequent neural network architecture (pg. 4, para. 1). The described process of generating one stochastically-truncated time series (i.e., truncated at a point s sampled from a distribution P), from each raw input time series, is viewed as functionally equivalent to generating a plurality of time series (i.e., respectively truncated at each point s along the distribution P) from each raw time series and randomly selecting one from each plurality.
In this way, Wang teaches production of a training dataset comprising sampled subsets of reduced trajectories. If applied to the construction of a predictive model based on time series of feline clinical measurements, this process would produce training data as claimed.
Wang teaches that their early-aware framework yields significantly better predictions than various state-of-the-art methods for early time series classification, and is also competitive with state-of-the-art time series classification algorithms given fully observed time series data (pg. 1, Abstract; pg. 10, para. 2; pg. 11, Fig. 2). Wang particularly highlights the high accuracy of their framework in making extremely early predictions, based on even 60-80% missing time series (pg. 10, para. 1).
With respect to claim 11, Martin discloses that “the processor… interface[s] with computer peripheral devices (e.g., a video display, a keyboard, computer mouse, etc)” (pg. 7, para. 0055) and “output[s] the predicted classifications to a display device” (pg. 6, para. 0042), e.g., via a graphical user interface.
With respect to claim 12, Martin discloses that “instructions may initially be borne on… a remote computer… The remote computer can… send the instructions… using a modem. A communications device local to a computing device… can receive the data… and place the data on a system bus for the processor” (pg. 7, para. 0057), and “output[ting] the predicted classifications to… another device in communication… across a network” (pg. 6, para. 0042), i.e., receiving data from, and transmitting information to, a remote second system.
With respect to claim 14, Martin discloses “receiv[ing] a training data set and training class set, the training class set identifying a class associated with each of the elements in the training data set… [and] generat[ing] a first classifier… by applying a machine learning technique to the training data set and the training class set” (pg. 1, para. 0008). In other words, supervised learning.
With respect to claim 15, Martin discloses that “any machine-learning technique may be applied to generate the classifier” (pg. 5, para. 0041), wherein “In some implementations, no known classes for the data samples… are available, and thus… [a] class set is not provided to the classification engine” (pg. 4, para. 0025). In other words, unsupervised learning.
With respect to claim 16, Martin exemplifies “data sets… generated by collection of samples from the same population of patients at different times” (pg. 4, para. 0029).
With respect to claim 18, Martin discloses “divid[ing] a data set into a discovery/training set and a test/validation set” (pg. 1, para. 0004), and a “data pre-processing engine, which receives bulk data and splits the bulk data… into… data set[s]… [which] may be of the same… size[]” (pg. 4, para. 0031), i.e., folds. Martin further discloses performance of a “cross validation process in training” (pg. 8, para. 0070).
With respect to claim 19, Martin discloses “pool[ing] data from multiple sources to form a combined data set” (pg. 1, para. 0004), wherein “Data sets… can be collected from multiple sources… using different exclusion or inclusion criteria” (pg. 4, para. 0029).
With respect to claim 20, Martin discloses that “The classification engine may use any one or more known machine-learning algorithms… including… support vector machine techniques… Random Forest techniques, k-nearest neighbors techniques… neural network-based techniques” (pg. 5, para. 0032) and “linear classifiers such as… [a] naive Bayes classifier” (pg. 2, para. 0012).
An invention would have been obvious to one of ordinary skill in the art if some teaching in the prior art would have led that person to combine prior art reference teachings to arrive at the claimed invention. Before the effective filing date of the claimed invention, said practitioner would have utilized the biomarker analysis method of Martin to predict risk of contracting CKD in felines, as taught by Al-Murrani, because Al-Murrani teaches that analyzing biomarkers is an enabled method of predicting this risk (pg. 2, para. 0021; pg. 13, para. 0189).
Said practitioner would have achieved this combination of teachings by training a machine learning algorithm on biomarker data from felines diagnosed with CKD and not diagnosed with CKD, because Martin teaches that machine learning classifiers can be trained on associated biomarker data from grouped presence/absence samples to predict disease risk (pg. 4, paras. 0026 and 0028-29; pg. 5, para. 0031) and additionally one of ordinary skill in the art would have understood that a machine learning classifier must be trained on data representing the classes which it is intended to distinguish.
Additionally, said practitioner would have combined the disclosed dietary recommendations with the method of Martin because Al-Murrani teaches that these dietary recommendations have utility for treating felines at risk of kidney disease (pg. 6, paras. 0079-0080). Said practitioner would have had a reasonable expectation of success because Martin and Al-Murrani both discuss analyzing biomarker levels (e.g., urine protein levels) to prognose disease risk.
An invention would have been obvious to one of ordinary skill in the art if some teaching in the prior art would have led that person to combine prior art reference teachings to arrive at the claimed invention. Before the effective filing date of the claimed invention, said practitioner would have enhanced the biomarker analysis method of Martin, in view of Al-Murrani, by modifying the dietary recommendations of Al-Murrani to include the relatively low potassium level disclosed by Friesen, because Friesen teaches that administering food compositions containing relatively low levels of protein, sodium and potassium as disclosed is beneficial for preventing and/or treating feline kidney disease (paras. 0059, 0062 and 0068). Said practitioner would have had a reasonable expectation of success because Martin, Al-Murrani and Friesen all discuss analyzing biomarker levels (e.g., urine protein levels) to prognose disease risk.
An invention would have been obvious to one of ordinary skill in the art if some teaching in the prior art would have led that person to combine prior art reference teachings to arrive at the claimed invention. Before the effective filing date of the claimed invention, said practitioner would have implemented combined analysis of age and biomarker levels, as taught by Struck, to enhance the biomarker analysis method taught by Martin, in view of Al-Murrani and Friesen, because Struck teaches that prediction of CKD in non-human subjects can be improved by additionally analyzing a clinical parameter such as age (pg. 3, para. 0039). Said practitioner would have had a reasonable expectation of success because Martin, Al-Murrani, Friesen and Struck all discuss analyzing biomarker levels (e.g., urine protein levels) to prognose disease risk.
An invention would have been obvious to one of ordinary skill in the art if some teaching in the prior art would have led that person to combine prior art reference teachings to arrive at the claimed invention. Before the effective filing date of the claimed invention, said practitioner would have implemented use of a mixture of experts learning technique, as taught by Maathuis, within the biomarker analysis method taught by Martin, in view of Al-Murrani, Friesen, and Struck, because Maathuis teaches that mixture of experts is a technique with broad cross-domain applicability (pg. 326, para. 1; pg. 332, para. 3) which outperforms considered baseline models on sparse time series data with high predictive accuracy to ground truth (pg. 337, Fig. 3c; pg. 338, Fig. 4).
Said practitioner would have had a reasonable expectation of success in combining these teachings because Martin and Maathuis concern similar fields of endeavor, both discussing predictive modeling of time series data via ensemble techniques.
An invention would have been obvious to one of ordinary skill in the art if some teaching in the prior art would have led that person to combine prior art reference teachings to arrive at the claimed invention. Before the effective filing date of the claimed invention, said practitioner would have implemented training on sampled subsets of reduced trajectories, as taught by Wang, within the biomarker analysis method taught by Martin, in view of Al-Murrani, Friesen, Struck and Maathuis, because Wang teaches that training on sampled subsets of reduced trajectories yields significantly better predictions than various state-of-the-art methods for early time series classification, including highly accurate extremely early predictions (based on time series with majorities of data missing), and is competitive with state-of-the-art time series classification algorithms given fully observed time series data (pg. 1, Abstract; pg. 10, paras. 1-2; pg. 11, Fig. 2).
Said practitioner would have had a reasonable expectation of success in combining these teachings because Martin and Wang concern similar fields of endeavor, both discussing predictive modeling of time series data. Additionally, Maathuis and Wang both concern robust modeling of chaotic time series data.
In this way the disclosure of Martin, in view of Al-Murrani, Friesen, Struck, Maathuis and Wang, makes obvious the limitations of claims 10-12, 14-16 and 18-20. Thus, the invention is prima facie obvious.
Claim 21 is rejected under 35 USC § 103 as being unpatentable over Martin, in view of Al-Murrani, Friesen, Struck, Maathuis and Wang, as applied to claim 10 above, and further in view of Oehmcke et al (Proceedings of the 2016 International Joint Conference on Neural Networks, pp. 2774-2781, IEEE Xplore; published 11/3/2016; previously cited). The new grounds of rejection presented herein were necessitated by Applicant’s amendment of the claims (filed 3/23/2026) to incorporate new limitations.
With respect to claim 21, Martin discloses that “The classification engine may use any one or more known machine-learning algorithms… including… k-nearest neighbors techniques” (pg. 5, para. 0032) and “data sets… generated by collection of samples from the same population of patients at different times” (pg. 4, para. 0029), e.g., time series data. Martin does not disclose dynamic time warping.
Al-Murrani teaches “select[ing] among a number of algorithms for analyzing… data” (pg. 19, para. 0234). Al-Murrani does not teach dynamic time warping.
Friesen describes “diagnostic methods for evaluating the presence and severity of renal dysfunction” (pg. 5, para. 0043). Friesen does not teach dynamic time warping.
Struck teaches performance of various forms of “Statistical analysis” (pg. 9, para. 0126). Struck does not teach dynamic time warping.
Maathuis discusses evaluation of constituent expert models against a “simple but robust baseline (1-Nearest Neighbor)” (pg. 330, para. 1). Maathuis does not teach dynamic time warping.
Wang teaches discusses evaluation of constituent expert models against an ensemble of “kNN classifiers with fixed truncation length (Fixed 1-NN)” and notes superior performance of their EA-ConvNets architecture, positing that the superior performance may be attributable to the dynamic adaptation of truncation length (pg. 13, para. 2-3). Wang does not teach dynamic time warping.
Oehmcke presents “an accurate and efficient algorithm to impute missing values in multivariate time series data… [that] combin[es] the weighted k-Nearest Neighbors algorithm (kNN) with Dynamic Time Warping (DTW)… [and] creates an ensemble of imputation models with diversity methods such as bagging or by varying the penalty weights” (pg. 2774, r. column). Oehmcke characterizes their algorithm as an “DTWkNN ensemble algorithm” (pg. 2774, r. column).
Oehmcke teaches that “The algorithm employs Dynamic Time Warping as [the] distance metric instead of point-wise distance measurements” (pg. 2774, Abstract), uniquely allowing it to robustly evaluate multivariate data of different window widths (i.e., lengths) with consecutive missing values (pg. 2775, l. column – pg. 2777, l. column). Oehmcke evaluates performance of their algorithm on datasets with at least 5% consecutively missing, and presents findings of outperformance as compared to standard and state-of-the-art methods in 51% of all cases considered (pg. 2779, r. column – pg. 2780, r. column and Fig. 3).
An invention would have been obvious to one of ordinary skill in the art if some teaching in the prior art would have led that person to combine prior art reference teachings to arrive at the claimed invention. Before the effective filing date of the claimed invention, said practitioner would have implemented use of dynamic time warping, as taught by Oehmcke, within the biomarker analysis method taught by Martin, in view of Al-Murrani, Friesen, Struck, Maathuis and Wang, because Martin discloses employment of kNN techniques while Oehmcke teaches that implementation of dynamic time warping allows kNNs to accurately evaluate datasets with consecutive missing values and is more effective than other methods of achieving the same (pg. 2774, Abstract; pg. 2779, r. column – pg. 2780, r. column and Fig. 3).
Said practitioner would have had a reasonable expectation of success in combining these teachings because Martin and Oehmcke concern similar fields of endeavor, both discussing analysis of time series data using kNNs. Additionally, Maathuis and Oehmcke both particularly concern robust modeling of chaotic time series data using ensemble techniques.
In this way the disclosure of Martin, in view of Al-Murrani, Friesen, Struck, Maathuis, Wang and Oehmcke, makes obvious the limitations of claim 21. Thus, the invention is prima facie obvious.
Claims 22-23 are rejected under 35 USC § 103 as being unpatentable over Martin, in view of Al-Murrani, Friesen, Struck, Maathuis and Wang, as applied to claim 10 above, and further in view of Lipton et al (arXiv, article no. 1511.03677v7 [cs:LG], pp. 1-18; published 3/21/2017, previously cited). The new grounds of rejection presented herein were necessitated by Applicant’s amendment of the claims (filed 3/23/2026) to incorporate new limitations.
With respect to claim 22, Martin discloses that “The classification engine may use any one or more known machine-learning algorithms… including… neural-network based techniques” (pg. 2, para. 0012) and “data sets… generated by collection of samples from the same population of patients at different times” (pg. 4, para. 0029), e.g., time series data. Martin does not disclose recurrent neural networks with long short-term memory.
Al-Murrani teaches “select[ing] among a number of algorithms for analyzing… data” (pg. 19, para. 0234). Al-Murrani does not teach recurrent neural networks with long-short term memory.
Friesen describes “diagnostic methods for evaluating the presence and severity of renal dysfunction” (pg. 5, para. 0043). Friesen does not teach recurrent neural networks with long-short term memory.
Struck teaches performance of various forms of “Statistical analysis” (pg. 9, para. 0126). Struck does not teach recurrent neural networks with long-short term memory.
Maathuis teaches a hierarchical mixture of experts architecture wherein the constituent specialized neural networks (‘experts’) include a “recurrent Long Short-Term Memory NN (LSTM)” (pg. 329, para. 6 – pg. 330, para. 1), i.e., a recurrent neural network with long-short term memory. Maathuis thus demonstrates the feasibility of implementing an LSTM as an expert, within a mixture of experts ensemble, for predictive modeling of chaotic time series data. However, Maathuis does not discuss particular motivations for doing so.
Wang discusses adjacent work that utilizes the structure characteristic of long short term memory (LSTM) for early recognition tasks (pg. 7, para. 2), and suggests extension of their earliness-aware framework “using other neural models such as LSTMs, which are natural tools for deal[ing] with time series data” (pg. 14, para. 3). Wang does not elaborate particular functional aspects that make LSTMs ‘natural tools’ for this purpose.
Lipton discusses “Clinical medical data… [and] Recurrent Neural Networks (RNNs), particularly those using Long Short-Term Memory (LSTM)… powerful and increasingly effective models”, and “establish[es] the effectiveness of a simple LSTM network for modelling clinical data” (pg. 1, Abstract). Lipton elaborates the functional advantages of LSTMs in handling clinical time series data, teaching that “[Clinical] data is difficult to mine effectively, owing to varying length, irregular sampling and missing data… Trained once on raw time series, our models outperform several strong baselines” (pg. 1, Abstract). In other words, clinical time series data is often chaotic and LSTMs outperform other models in processing such data.
Additionally, Lipton teaches that “While we could not locate any published papers using LSTMs for multilabel classification in the medical domain, several papers use feedforward nets for this task. One of the earliest papers to investigate multi-task neural networks modeled risk in pneumonia patients” (pg. 3, section 2.3). In this way, Lipton suggests applicability of LSTMs to modeling of clinical risk.
With respect to claim 23, Wang states that their earliness-aware component draws inspiration from the prior idea of nested dropout, wherein a decay point is sampled from some distribution to stochastically drop nested sets of hidden units in a neural network (pg. 4, para. 1), and describes it as “a robust regularizor (akin to a ‘dropout’ mechanism)” (pg. 9, para. 4).
Wang additionally describes implementation of SpatialDropout (a regularization algorithm) on the final layer representation (
F
i
(T)) for each time series, before the final classifier, to avoid overfitting (pg. 5, para. 5). Wang does not further discuss particular dropout percentages implemented by this algorithm, and does not specifically teach employment of a regularization algorithm comprising 5% or more dropout.
Lipton teaches that “regularization, including judicious use of dropout, is crucial to our performance” (pg. 3, section 2.6) and exemplifies “dropout probability 0.5” (pg. 7, section 5; pg. 15, section B), i.e., 50%.
An invention would have been obvious to one of ordinary skill in the art if some teaching in the prior art would have led that person to combine prior art reference teachings to arrive at the claimed invention. Before the effective filing date of the claimed invention, said practitioner would have implemented use of a recurrent neural network featuring long short-term memory, within the biomarker analysis method taught by Martin, in view of Al-Murrani, Friesen, Struck, Maathuis and Wang, because Maathuis and Wang discuss employment of LSTMs for predictive modeling of chaotic time series data while Lipton teaches that LSTMs allow for evaluation of clinical time series datasets with varying lengths, sampling protocols, and/or missing values and is more effective than other methods of achieving the same (pg. 1, Abstract). In this way, prior art teaches that implementation of LSTM architecture would improve the applicability and performance of said biomarker analysis method.
Additionally, said practitioner would have implemented dropout of 5% or more, within the biomarker analysis method taught by Martin, in view of Al-Murrani, Friesen, Struck, Maathuis and Wang, because Wang teaches that implementation of dropout helps avoid overfitting while Lipton teaches that their described judicious use of dropout (i.e., dropout of 50%) is crucial to the favorable performance of their discussed LSTM architecture (pg. 3, section 2.6; pg. 7, section 5; pg. 15, section B).
Said practitioner would have had a reasonable expectation of success in combining these teachings because Martin, Maathuis, Wang and Lipton concern similar fields of endeavor, all discussing analysis of time series data using a neural network algorithm.
In this way the disclosure of Martin, in view of Al-Murrani, Friesen, Struck, Maathuis, Wang and Lipton makes obvious the limitations of claims 22-23. Thus, the invention is prima facie obvious.
Response to Arguments - Claim Rejections Under 35 USC § 103
In the Remarks filed 3/23/2026, Applicant traverses the rejections under 35 USC § 103 and presents supporting arguments.
Applicant alleges that Martin fails to disclose, teach or suggest that: (1) the training dataset comprises, for each sample feline, medical information associated with a plurality of reduced trajectories as claim a first set of sample felines that have been diagnosed with CKD either during or after a veterinary visit; (2) the machine learning algorithm was trained using a mixture-of-experts (MOE) technique based on an ensemble of a plurality of prediction models built on respective sampled subsets of a training dataset as claimed. Applicant further asserts that Al-Murrani, Friesen and Struck do not remedy the alleged deficiencies of Martin (pg. 6, para. 4 – pg. 7, para. 3).
These arguments are found persuasive insofar as the cited references are considered deficient with respect to the noted features of the amended claims. Consequently, the previous rejections under 35 USC § 103 have been withdrawn. Applicant’s amendment of the claims to incorporate new limitations, directed to these noted features, necessitated search and application of additional prior art. New rejections under § 103 have been issued in this Office action, applying additional references (Maathius and Wang) that are considered to remedy the deficiencies of the previously-applied art with respect to these features.
Applicant further asserts that each of Al-Murrani, Friesen, Struck, Oehmcke and Lipton concern divergent fields of endeavor and/or do not involve, reference or pertain to machine learning methodologies or applications, constructing a training dataset comprising reduced trajectories of clinical visit records as claimed, and/or feline CKD diagnosis (pg. 7, para. 3 – pg. 9, para. 1). As best understood, Applicant argues that the cited references cannot be properly considered to make obvious the instant claim limitations involving the features for which each reference is particularly relied upon by the Office, as they are not analogous art to Martin and/or the instant application.
The ‘field of endeavor’ is not limited to the specific point of novelty of an application, the narrowest possible conception of a field, or the particular focus within a given field (Netflix, Inc. v. DivX, LLC, 80 F.4th 1352, 1358-59 (Fed. Cir. 2023)). The field of endeavor of the claimed invention overlaps with that of several cited references and is not limited to, e.g., computer science.
Al-Murrani and Struck, for example, both concern prognosis of chronic kidney disorders based on analysis of measured biomarker levels. Although each reference further bears particular conceptual departures from the claimed invention, surely said prognosis can be reasonably viewed as a field of endeavor shared between Al-Murrani, Struck, and the claimed invention.
Moreover, prior art need not be from the same field of endeavor as a claimed invention to be considered analogous art (i.e., proper for combined application in an obviousness rejection) so long as it is reasonably pertinent to the problem faced by the inventor. As the Court held in KSR International Co. v. Teleflex Inc., 550 U.S. 398 (2007), “When a work is available in one field of endeavor, design incentives and other market forces can prompt variations of it, either in the same field or a different one. If a person of ordinary skill can implement a predictable variation, § 103 likely bars its patentability” (Id. at 417). All applied art is considered at least reasonably pertinent to the problem faced by the inventor, therefore this particular point of argument is not found persuasive. See MPEP 2141.01(a) and 2141 § I.
Double Patenting
The nonstatutory double patenting rejection is based on a judicially created doctrine grounded in public policy (a policy reflected in the statute) so as to prevent the unjustified or improper timewise extension of the “right to exclude” granted by a patent and to prevent possible harassment by multiple assignees. A nonstatutory double patenting rejection is appropriate where the conflicting claims are not identical, but at least one examined application claim is not patentably distinct from the reference claim(s) because the examined application claim is either anticipated by, or would have been obvious over, the reference claim(s). See, e.g., In re Berg, 140 F.3d 1428, 46 USPQ2d 1226 (Fed. Cir. 1998); In re Goodman, 11 F.3d 1046, 29 USPQ2d 2010 (Fed. Cir. 1993); In re Longi, 759 F.2d 887, 225 USPQ 645 (Fed. Cir. 1985); In re Van Ornum, 686 F.2d 937, 214 USPQ 761 (CCPA 1982); In re Vogel, 422 F.2d 438, 164 USPQ 619 (CCPA 1970); In re Thorington, 418 F.2d 528, 163 USPQ 644 (CCPA 1969).
A timely filed terminal disclaimer in compliance with 37 CFR 1.321(c) or 1.321(d) may be used to overcome an actual or provisional rejection based on nonstatutory double patenting provided the reference application or patent either is shown to be commonly owned with the examined application, or claims an invention made as a result of activities undertaken within the scope of a joint research agreement. See MPEP § 717.02 for applications subject to examination under the first inventor to file provisions of the AIA as explained in MPEP § 2159. See MPEP § 2146 et seq. for applications not subject to examination under the first inventor to file provisions of the AIA . A terminal disclaimer must be signed in compliance with 37 CFR 1.321(b).
The filing of a terminal disclaimer by itself is not a complete reply to a nonstatutory double patenting (NSDP) rejection. A complete reply requires that the terminal disclaimer be accompanied by a reply requesting reconsideration of the prior Office action. Even where the NSDP rejection is provisional the reply must be complete. See MPEP § 804, subsection I.B.1. For a reply to a non-final Office action, see 37 CFR 1.111(a). For a reply to final Office action, see 37 CFR 1.113(c). A request for reconsideration while not provided for in 37 CFR 1.113(c) may be filed after final for consideration. See MPEP §§ 706.07(e) and 714.13.
The USPTO Internet website contains terminal disclaimer forms which may be used. Please visit www.uspto.gov/patent/patents-forms. The actual filing date of the application in which the form is filed determines what form (e.g., PTO/SB/25, PTO/SB/26, PTO/AIA /25, or PTO/AIA /26) should be used. A web-based eTerminal Disclaimer may be filled out completely online using web-screens. An eTerminal Disclaimer that meets all requirements is auto-processed and approved immediately upon submission. For more information about eTerminal Disclaimers, refer to www.uspto.gov/patents/apply/applying-online/eterminal-disclaimer. Instant claims 10-12 and 16 are provisionally rejected on the ground of nonstatutory double patenting as being unpatentable over claims 81-83 and 85-86 of Application No. 17/250,389 (hereafter, “‘389”), in view of Al-Murrani, Friesen, Maathuis and Martin. ‘389 shares joint inventors (Bradley, Richard; Tagkopoulos, Ilias; Biourge, Vincent; Feugier, Alexandre; Delmotte, Sebastien) and a common assignee (Mars, Incorporated) with the instant application. The new grounds of rejection presented herein were necessitated by Applicant’s amendment of the claims (filed 3/23/2026) to incorporate new limitations.
Instant claim 10 recites “[a] method for treating chronic kidney disease (CKD) in a feline in need thereof… comprising”:
a) “administering a treatment diet to the feline in need thereof, the treatment diet comprising a potassium supplement diet having a potassium level between 0.1 mg/1000 kcal and 50 mg/1000 kcal and at least one of”: diets including levels within specific recited ranges of phosphorous, protein, sodium, potassium or polyunsaturated fatty acids, an anti-oxidant supplement diet, or a vitamin B supplement diet;
b) “wherein the feline has a probability score based on age and biomarker levels of one or more biomarker levels… comprising urine specific gravity level, a creatinine level, a urine protein level, a blood urea nitrogen (BUN) or urea level, a white blood cell count (WBC), or urine pH;
c) “wherein the probability score is obtained by executing a machine learning algorithm to transform the age and the biomarker levels to determine the probability score indicating that the feline has or is likely to develop CKD”;
d) “wherein the machine learning algorithm was trained using a mixture-of-experts (MOE) learning technique based [on] an ensemble of a plurality of prediction models built from a plurality of respective sampled subsets of a training dataset, the training dataset comprising, for each sample feline diagnosed with CKD, medical information associated with a plurality of reduced trajectories, each reduced trajectory being a temporally ordered subset of clinical visit records truncated at a point up to two years prior to a CKD diagnosis of the sample feline, and wherein each of the sampled subsets comprises, for each sample feline, a single reduced trajectory randomly selected from the plurality of reduced trajectories”; and
e) “wherein the probability score is greater than a threshold value indicating the feline has or is at risk of developing CKD, wherein the threshold value is determined by applying a linear discriminant analysis to a plurality of medical records of a plurality of felines, wherein the plurality of medical records comprise ages and measurements of the one or more biomarkers”.
With respect to instant claim 10, ‘389 claims systems for identifying susceptibility of a feline to developing chronic kidney disease (CKD), having functional limitations including:
b) analyz[ing] “input levels of one or more respective biomarkers from the feline and optionally an input level of an age of the feline… compris[ing] one or more of a urine specific gravity level, a creatinine level, a urine protein level, a blood urea nitrogen (BUN) or urea level, a white blood cell count (WBC), or urine pH… to derive a probability score of the feline developing CKD” (claim 81);
c) wherein analyzing the input level(s) and deriving the score is performed “by a classification algorithm”, and the score is “a probability score of the feline developing CKD” (claim 81);
d) “accessing a plurality of sampled subsets of a training dataset, the training dataset comprising, for each sample feline diagnosed with CKD, medical information associated with a plurality of reduced trajectories, each reduced trajectory being a temporally ordered subset of clinical visit records truncated at a point up to two years prior to a CKD diagnosis of the sample feline, and wherein each of the sample subsets comprises, for each sample feline, a single reduced trajectory randomly selected from the plurality of reduced trajectories; extracting, for each of the sampled subsets, medical information associated with [the] one or more biomarkers… training a plurality of predictors using the extracted medical information… and training the classification algorithm based on an ensemble of the plurality of predictors” (claim 81); and
e) “categorizing the feline based on the probability score, wherein if the probability score is a high probability score, the feline is assigned to a Prediction of Disease category… and determin[ing] a” (claim 81), wherein “the high probability score indicates that the feline will develop CKD” (claim 82) and “the probability score ranges from 0 to 100, and… the high probability score has a value of between 51 and 100” (83).
‘389 further claims determining a customized recommendation for the feline based on the categorizing. ‘389 does not claim administering a treatment diet as recited; training the classification algorithm using a mixture-of-experts learning technique; or determining the probability score thresholds by applying a linear discriminant analysis.
Al-Murrani discloses “a method for… prognosis of kidney disorder in a feline” (pg. 16, para. 0204), and teaches “utilizing at least one relevant biomarker isolated and measured from a biological test sample taken from such feline” (pg. 2, para. 0021), wherein “the level of expression of the one or more biomarkers… is normalized” (pg. 6, para. 0063). Al-Murrani exemplifies “acute or chronic abnormal loss of kidney function, such as renal failure” (pg. 10, para. 0157), i.e., CKD, and “measurements… including… urinary protein levels” (pg. 3, para. 0022).
Al-Murrani further teaches application comprising “predict[ing], detect[ing], and diagnos[ing] in such feline a change from a normal state to an abnormal state leading to a kidney disorder” (pg. 13, para. 0189), and teaches that their method is “predictive of a decline in renal function, as might otherwise be diagnosed at a later time” (pg. 3, para. 0023).
Additionally, Al-Murrani teaches “providing a kidney protective diet… to a feline having a kidney disorder as diagnosed or diagnosable by the method[s]” (pg. 6, paras. 0079-0080), wherein “the kidney protective diet comprises one or more of the following modifications relative to a standard feline diet: Reduced phosphorus, Reduced levels of protein, Reduced sodium, Increased levels of omega-3 fatty acids”, i.e., polyunsaturated fatty acids, “Increased levels of B-complex vitamins, [and] Increased antioxidants” (pg. 6, paras. 0088-0094).
In particular, Al-Murrani exemplifies a treatment diet providing “from about 3.6 to about 7.9 g/100 kcal ME protein” (pg. 6, para. 0096), i.e., a protein level between about 36 to about 79g/1000 kcal. This disclosed range overlaps the claimed specific range of a protein level less than 70g/1000 kcal. Where prior art discloses a range which overlaps or approaches a claimed range, the prior art disclosure renders the claimed range as obvious (MPEP 2144.05 § I).
Additionally, Al-Murrani exemplifies a diet providing “from about 0.04 to about 0.17 g/100 kcal ME phosphorus” (pg. 6, para. 0096), i.e., a phosphorus level between about 0.4 to about 1.7 g/1000 kcal. This disclosed range falls within (i.e., overlaps) the claimed specific range of a phosphorus level between 0.1 g/1000 kcal and 20 g/1000 kcal.
Additionally, Al-Murrani exemplifies a diet providing “from about 0.008 to about 0.07 g/100 kcal ME sodium” (pg. 6, para. 0096), i.e., a sodium level between about 8 mg to about 70 mg/1000 kcal. This disclosed range falls within (i.e., overlaps) the claimed specific range of a sodium level between 0.1 mg/1000 kcal and 50 mg/1000 kcal.
Al-Murrani does not describe a treatment diet providing a potassium level between 0.1 mg/1000 kcal and 50 mg/1000 kcal; and does not teach training an algorithm using a mixture-of-experts learning technique; or applying a linear discriminant analysis.
Friesen discusses administration of food compositions for preventing or treating kidney disease (Abstract), wherein the patient is preferably a canine or feline (para. 0013), and exemplifies food compositions comprising one or more food ingredients containing protein in amounts of from about 5% to about 30%, sodium in amounts of from about 0.01% to about 2%, potassium in amounts of from about 0.01% to about 2%, and phosphorus in amounts of from about 0.2 to about 1% dry matter weight (para. 0025). The low end of the disclosed range of potassium, 0.01% potassium, is equivalent to 100 mg potassium per 1 kg dry matter.
Friesen also mentions a utilized standard of 3.3 kcal/g dry food (para. 0054, Table caption), which equals 1000 kcal per about 0.303 kg dry matter. According to this standard, 0.01% potassium by dry weight would be equivalent to a potassium level of about 30.3 mg/1000 kcal. Friesen thus teaches administering a treatment diet having a potassium level of about 30.3 mg/1000 kcal, which falls within the claimed range of 0.1 mg/1000 kcal to 50 mg/1000 kcal.
Friesen teaches combination of the discussed food compositions with one or more renal diagnostic devices and methods for evaluating the presence and severity of renal dysfunction, including assessment of urine specific gravity, creatinine levels, albuminuria, proteinuria, blood urea nitrogen (BUN) concentration and/or urine pH (pg. 5, para. 0043). Friesen also presents study results indicating that administering food compositions containing relatively low levels of protein, sodium and potassium as disclosed is beneficial for preventing and/or treating feline kidney disease (pg. 7, para. 0059; pg. 8, para. 0062; pg. 9, para. 0068).
Friesen does not teach training an algorithm using a mixture-of-experts learning technique; or applying a linear discriminant analysis.
Maathuis evaluates “neural network approaches to perform a regression prediction task on chaotic time series”, and presents a modeling framework where “Different NNs”, neural networks, are “MLPs”, multilayer perceptrons, “Residual MLPs and Long Short-Term Memory (LSTM), are… embedded within a larger architecture, the Hierarchical Mixture of Experts (HME)” (pg. 326, Abstract; pg. 327, para. 3).
Maathuis discusses the difficulty and cross-field utility of chaotic time-series prediction, stating that “Many disciplines are concerned with time series prediction… Most problems in nature deal with nonlinear or chaotic data, which poses a challenge” (pg. 326, para. 1). Maathuis presents findings that their HME architecture outperforms considered baseline models on sparse time series data with high predictive accuracy to ground truth (pg. 337, Fig. 3c; pg. 338, Fig. 4).
Maathuis does not teach applying a linear discriminant analysis.
Martin describes “methods for correcting a data set and classifying a data set” (Abstract) via machine learning algorithm(s) trained on biomarker datasets to classify presence or risk of a disease (pg. 1, para. 0008; pg. 4, paras. 0026 and 0028), and teaches that “The classification engine may use… linear discriminant analysis techniques” (pg. 5, para. 0032). In this way, Martin teaches that classifications may be discriminated based on a threshold value determined by applying a linear discriminant analysis to the received data. Martin further teaches that the linear discriminant analysis technique has been implemented as an available R package, called ‘lda’ (pg. 5, para. 0032).
With respect to instant claim 11, ‘389 claims system functions of displaying the categorization and customized recommendation on a graphical user interface (claim 85).
With respect to instant claim 12, ‘389 claims embodiments wherein the one or more input levels are received from a remote second system via a communication device, and further claims system function of transmitting the categorization to the remote second system via the communication device (claim 86).
With respect to instant claim 16, ‘389 claims embodiments wherein the machine learning algorithm is trained on a dataset of biomarker levels extracted from medical information associated with, for each sample feline, a temporally ordered subset of clinical visit records (claim 81). In other words, biomarker levels comprising sequential measurements of the biomarkers measured at different time points.
An invention would have been obvious to one of ordinary skill in the art if some teaching in the prior art would have led that person to combine prior art reference teachings to arrive at the claimed invention. Before the effective filing date of the claimed invention, said practitioner would have provided a treatment diet to a feline, as taught by Al-Murrani, according to kidney protective dietary recommendations determined by the feline biomarker analysis method of ‘389, because Al-Murrani teaches that teaches that these dietary recommendations have utility for treating felines at risk of kidney disease (pg. 6, paras. 0079-0080). Said practitioner would have had a reasonable expectation of success because ‘389 and Al-Murrani both discuss analyzing biomarker levels (e.g., urine protein levels) to prognose feline chronic kidney disease risk and determine dietary recommendations.
An invention would have been obvious to one of ordinary skill in the art if some teaching in the prior art would have led that person to combine prior art reference teachings to arrive at the claimed invention. Before the effective filing date of the claimed invention, said practitioner would have enhanced the feline biomarker analysis method of ‘389, in view of Al-Murrani, by modifying the dietary recommendations of ‘389 (and Al-Murrani) to include the relatively low potassium level disclosed by Friesen, because Friesen teaches that administering food compositions containing relatively low levels of protein, sodium and potassium as disclosed is beneficial for preventing and/or treating feline kidney disease (paras. 0059, 0062 and 0068). Said practitioner would have had a reasonable expectation of success because ‘389, Al-Murrani and Friesen all discuss analyzing biomarker levels (e.g., urine protein levels) to prognose feline kidney disease risk and dietary compositions for treating said risk.
An invention would have been obvious to one of ordinary skill in the art if some teaching in the prior art would have led that person to combine prior art reference teachings to arrive at the claimed invention. Before the effective filing date of the claimed invention, said practitioner would have implemented use of a mixture of experts learning technique, as taught by Maathuis, within the feline biomarker analysis method of ‘389, in view of Al-Murrani and Friesen, because ‘389 discloses training a classification algorithm on time series data via an ensemble technique while Maathuis teaches that mixture of experts is an ensemble technique with broad cross-domain applicability (pg. 326, para. 1; pg. 332, para. 3) which outperforms considered baseline models on sparse time series data with high predictive accuracy to ground truth (pg. 337, Fig. 3c; pg. 338, Fig. 4). Said practitioner would have had a reasonable expectation of success in combining these teachings because ‘389 and Maathuis both discuss predictive modeling of time series data via ensemble techniques.
An invention would have been obvious to one of ordinary skill in the art if some teaching in the prior art would have led that person to combine prior art reference teachings to arrive at the claimed invention. Before the effective filing date of the claimed invention, said practitioner would have implemented linear discriminant analysis, as taught by Martin, within the feline biomarker analysis method of ‘389, in view of Al-Murrani, Friesen and Maathuis, because Martin teaches that linear discriminant analysis is a known means of determining classification thresholds that has been implemented as an available R package (pg. 5, para. 0032). Said practitioner would have had a reasonable expectation of success in combining these teachings because ‘389 and Martin both discuss classifying disease risk, based on analyzed biomarker levels, via trained machine learning algorithms.
In this way the claims of ‘389, in view of Al-Murrani, Friesen, Maathuis and Martin, make obvious the limitations of instant claims 10-12 and 16. Thus, the invention is prima facie obvious. This is a provisional nonstatutory double patenting rejection because the patentably indistinct claims have not in fact been patented.
Response to Arguments - Double Patenting
In the Remarks, Applicant alleges that there is no factual basis to conclude that claims 10-12 and 16 are an obvious variation of claims of co-pending Application No. 17/250,389 because Martin, Al-Murrani and Friesen fail to teach or suggest the amended claim 10 as detailed (pg. 9, paras. 3-4).
This argument is found persuasive insofar as the cited references are considered deficient with respect to features of amended claim 10. Consequently, the previous provisional rejection on the ground of nonstatutory double patenting has been withdrawn. Applicant’s amendment of the claims, to incorporate new limitations, necessitated search and application of additional prior art. A new rejection on the ground of nonstatutory double patenting has been issued in this Office action, applying an additional reference (Maathuis) that is considered to remedy the deficiencies of the previously-applied art with respect to these features.
Conclusion
At this point in prosecution, no claim is allowed.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Theodore C. Striegel whose telephone number is (571)272-1860. The examiner can normally be reached Mon-Fri 9am-5pm ET.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Olivia M. Wise can be reached at (571)272-2249. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/T.C.S./Examiner, Art Unit 1685
/JESSE P FRUMKIN/Primary Examiner, Art Unit 1685 May 29, 2026
Prosecution Insights