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 .
Response to Applicant’s Arguments
35 U.S.C. § 101
Applicant’s remarks filed on 11 June 2026, on pages 5-10, regarding the prior-art rejections under 35 U.S.C. § 101 have been fully considered. The arguments are not persuasive for the reasons set forth below.
Applicant argues that the ordered combination defines a technical workflow because the claims acquire biological datasets, compute similarity, and generate a new tissue-level representation through similarity-derived weights.
Examiner respectfully disagrees. Under MPEP § 2106.04(d), the additional elements must integrate the exception into a practical application. Here, “by analyzing a sample” and “cultured in an in vitro environment” supply input data for the similarity-weighted calculation; Spec. [0047] states target-tissue data may be acquired by sample analysis “but is not limited thereto.” The claimed improvement comes from the weighted synthesis itself, not from a claimed technological improvement in sample analysis, cell culture, or computer operation.
Applicant argues that claims 1 and 7-8 do not recite an abstract idea because they require “a specific computational technique in bioinformatics” using omics datasets from distinct biological sources that cannot practically be performed in the human mind.
Examiner respectfully disagrees. Claims 1 and 7-8 recite a mental process because, under BRI, the claims require comparing first and second omics datasets to determine relative similarity, assigning greater or lesser weight based on that similarity, and forming an estimated tissue-level result from the weighted information. These are observation, evaluation, judgment, and mental comparison steps that fall within MPEP § 2106.04(a)(2). The biological source of the data does not change the nature of the recited analysis because the claims do not require a specific assay improvement, computer improvement, or non-mental calculation technique beyond “calculating a similarity,” applying that similarity “as a weight,” and performing “weighted synthesis.”
Applicant argues that claims 1 and 7-8 cannot be practically performed in the human mind because the omics data involve gene-expression information across multiple genes, similarity calculations across plural cells, and weighted synthesis requiring “systematic numerical processing.”
Examiner respectfully disagrees. The argument relies on features not recited at the claimed level of detail. Claim 1 recites “calculating a similarity” and applying that similarity “as a weight” to perform “weighted synthesis,” but it does not require any particular number of genes, dataset size, numerical precision, matrix operation, algorithm, or computational burden that would make the comparison incapable of mental performance. Under BRI, the claim covers comparing biological datasets, judging relative similarity, assigning weights, and combining weighted information, which are mental evaluation and calculation steps under MPEP § 2106.04(a)(2).
Applicant argues that claims 1 and 7-8 do not recite a “method of organizing human activity” because the data-acquisition steps involve biological samples and experiments, not commercial, social, economic, or interpersonal activity.
The argument is moot to that extent. The § 101 rejection does not need to rely on certain methods of organizing human activity because claims 1 and 7-8 recite mathematical concepts and, secondarily, mental-process-type comparison/evaluation through “calculating a similarity,” applying the similarity “as a weight,” and performing “weighted synthesis.” The limitations “by analyzing a sample of the target tissue” and “cultured in an in vitro environment” are evaluated as additional elements under Prong Two. They identify the source of input data for the calculation, but they do not integrate the exception into a practical application because the claims do not recite a particular assay improvement, cell-culture improvement, transformation, treatment step, or improvement to computer functionality.
Applicant argues that claims 1 and 7-8 merely use mathematical operations within computational biology and drug-response modeling, and therefore do not recite a judicial exception.
Examiner respectfully disagrees. A claim may recite a judicial exception even when the mathematical operation is used in a scientific or bioinformatics context. Here, claims 1 and 7-8 expressly require “calculating a similarity,” applying the calculated similarity “as a weight,” and performing “weighted synthesis.” Those limitations define the operative claim mechanism as comparison, weighting, and mathematical combination of information. The biological field of use does not remove the mathematical concept from Prong One.
Applicant argues that the claims as a whole are directed to a specific computational bioinformatics technique that cannot practically be performed in the human mind.
Examiner respectfully disagrees. Under BRI, claims 1 and 7-8 do not recite a specific non-mental algorithm, dataset scale, numerical precision, model architecture, or computer-function improvement. The claims recite acquiring omics data, calculating similarity, and using similarity as a weight for synthesis. That claim language still covers mental-process-type comparison and evaluation at the claimed level, and any biological data-source limitations are addressed as additional elements under Prong Two rather than as removing the Prong One exception.
Applicant argues that claims 1 and 7-8 transform biological data into a unified and functionally meaningful output.
Examiner respectfully disagrees. The claims do not recite a transformation step. Claim 1 recites “estimating information on the target tissue” by “differentially synthesizing information on the plurality of cells,” where similarity is applied “as a weight” to perform “weighted synthesis.” Under BRI, this is generating estimated information from input information by comparison, weighting, and synthesis. The claim does not require changing a physical sample, improving an assay, improving cell culture, or improving computer operation. Thus, the alleged transformation is attorney characterization, not a claimed practical application.
Applicant argues that claims 1 and 7-8 provide a concrete technical benefit because Spec. [0004]-[0007] identifies the in vitro/in vivo accuracy problem, and the claimed similarity-weighted synthesis improves tissue-level predictions by accounting for similarity between tissue and cells.
Examiner respectfully disagrees. Under Prong Two, the claim must integrate the judicial exception into a practical application; it is not enough that the exception produces a useful or more accurate result. The cited benefit arises from the same abstract analysis recited in the claims: “calculating a similarity,” applying that similarity “as a weight,” and performing “weighted synthesis.” Spec. [0014] states the accuracy benefit comes from “differentially synthesizing cell-level information based on the similarity,” which is the identified exception itself, not an additional technological element that improves sample analysis, cell culture, drug testing, or computer operation.
Applicant argues that claims 1 and 7-8 do not merely recite data gathering or field-of-use limitations because the target-tissue omics data and in vitro-cell omics data define distinct data domains that are central to the computational problem.
Examiner respectfully disagrees. Under MPEP § 2106.04(d)(2), Prong Two asks whether the additional elements integrate the exception into a practical application, and MPEP § 2106.05(g) identifies insignificant extra-solution activity, including data gathering, as not sufficient for integration. Claims 1 and 7-8 recite “acquiring first omics data” and “acquiring second omics data” before “calculating a similarity” and performing “weighted synthesis.” Spec. [0047] states target-tissue data may be acquired by sample analysis “but is not limited thereto,” and Spec. [0039] states cell-line information may be obtained from a database or at low experimental cost. Thus, the acquisition limits supply input data for the abstract calculation, but do not recite an improved assay, improved cell-culture process, or improved computer operation.
Applicant argues that the claims meaningfully constrain the abstract idea because the datasets interact and their differences are reconciled through specifically defined processing.
Examiner respectfully disagrees. MPEP § 2106.05(h) explains that generally linking a judicial exception to a particular technological environment or field of use does not integrate the exception into a practical application. The alleged reconciliation is the identified exception: “calculating a similarity,” applying that similarity “as a weight,” and performing “weighted synthesis.” The claims use biological datasets as the field-specific inputs and output estimated “tissue-level information,” but they do not apply the calculation in a treatment step, transform a physical article under MPEP § 2106.05(c), or improve another technology under MPEP § 2106.05(a).
Applicant argues that claims 1 and 7-8 do not merely apply an abstract idea on a generic computer because the improvement lies in the structured sequence of similarity calculation and weighted synthesis.
Examiner respectfully disagrees. Under MPEP § 2106.05(f), merely using a computer to execute an abstract idea does not integrate the exception into a practical application. Claims 1 and 7-8 recite generic computing implementation while the alleged improvement is “calculating a similarity,” applying the similarity “as a weight,” and performing “weighted synthesis.” Under MPEP § 2106.05(a), the claim must improve computer functionality or another technology; here, the asserted improved output results from the abstract comparison-and-weighting process itself.
Applicant argues that the claims recite a particular data-processing pipeline that transforms biological datasets into a new predictive output with improved accuracy.
Examiner respectfully disagrees. The claims do not recite transforming biological datasets. They recite “estimating information on the target tissue” by “differentially synthesizing information” using similarity-based weights. MPEP § 2106.05(c) concerns transformation of a particular article to a different state or thing, not generation of new information from input information. The improved accuracy described in Spec. [0014] arises from “differentially synthesizing cell-level information based on the similarity,” which is the identified judicial exception, not an additional integrating element under MPEP § 2106.04(d)(2).
Applicant argues that claims 1 and 7-8 recite significantly more because the ordered combination calculates similarity, converts similarity into weights, and performs weighted synthesis to generate a tissue-level prediction.
Examiner respectfully disagrees. Those recited steps are the identified judicial exception, not additional elements beyond the exception. Claims 1 and 7-8 recite “calculating a similarity,” applying the similarity “as a weight,” and performing “weighted synthesis.” Under MPEP § 2106.05, Step 2B asks whether additional elements, individually or as an ordered combination, add significantly more than the exception. Repeating the comparison, weighting, and synthesis steps does not supply an inventive concept separate from the abstract analysis itself.
Applicant argues that the ordered combination is not routine or conventional because the cited § 102 and § 103 references allegedly do not perform the same similarity-driven weighted synthesis.
Examiner respectfully disagrees. Novelty or nonobviousness of the alleged synthesis mechanism does not, by itself, establish eligibility under § 101. The Step 2B inquiry focuses on whether the claim adds significantly more than the judicial exception. The additional elements here are generic computing components and data-source limitations: “computing device,” “memory,” “processor,” sample-based omics acquisition, and in vitro-cell omics acquisition. Spec. [0037] allows “any type of device equipped with a computing function,” Spec. [0084] identifies processors “well known in the art,” Spec. [0039] states cell-line information may be obtained from a database or at low experimental cost, and Spec. [0047] states sample analysis is “not limited thereto.”
Applicant argues that the Office improperly considered elements individually and ignored the ordered combination.
Examiner respectfully disagrees. The ordered combination has been considered. In combination, the additional elements obtain biological input data and execute the similarity-weighted synthesis on ordinary computing components. MPEP § 2106.05(d) and § 2106.07(a) permit reliance on specification admissions for well-understood, routine, conventional activity, and MPEP § 2106.05(f) addresses generic computer implementation. The tissue-level prediction results from the abstract calculation itself, not from an unconventional additional technical arrangement.
Applicant argues that the Office improperly considered elements individually and ignored the ordered combination.
Examiner respectfully disagrees. The ordered combination has been considered under MPEP § 2106.05. The combination of “calculating a similarity,” applying the similarity “as a weight,” and performing “weighted synthesis” is the identified abstract analysis itself, not an additional element that supplies significantly more. The additional elements, considered together, acquire biological input data and execute the similarity-weighted calculation on generic computing components. Spec. [0037] states the device may be “any type of device equipped with a computing function,” and Spec. [0084] identifies processors “well known in the art.” Thus, the ordered combination does not add an inventive concept beyond using ordinary computing components and biological input data to perform the claimed comparison, weighting, and synthesis.
Applicant argues that claims 2-6 are eligible for the same reasons.
Examiner respectfully disagrees. Claims 2-6 add cell-line context, feature vectors, vector distance, classification confidence scores, and drug-effect information. These limitations specify inputs, outputs, or mathematical paths for the same similarity-weighted estimation. They do not add a separate inventive concept beyond the identified judicial exception and the generic/data-source elements already evaluated.
35 U.S.C. § 102/103
Applicant’s remarks filed on 11 June 2026, on pages 10-12, regarding the prior-art rejections under 35 U.S.C. § 102/103 have been fully considered. The arguments are moot for the reasons set forth below.
Applicant argues that Szeto does not anticipate amended claims 1 and 7-8 because Szeto does not apply each calculated similarity as a weight to corresponding cell information and traversed 35 USC 103.
Examiner respectfully agrees. Szeto does not disclose the amended weighted-synthesis limitation in a single-reference § 102 analysis. However, the argument is moot under the current § 103 rejection because Alemi teaches similarity-weighted aggregation: multiplying similarity score by outcome, summing weighted outcomes, and weighing each case by similarity (Alemi [0032], [0042], [0075], [0077]). Combined with Szeto’s cell-line omics/drug-response prediction system, the references render amended claim 1 obvious.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-8 satisfy Step 1 but remain ineligible: they fall within a statutory category, recite abstract mathematical and mental-process concepts, lack a practical application, and do not include an inventive concept.
Step 1.
Claim 1 recites a process as “a method.” Claim 7 recites a machine as “a device” with “memory” and a “processor.” Claim 8 recites a manufacture because it is stored on a “non-transitory computer-readable recording medium.”
The claims therefore pass Step 1 and proceed to Step 2A.
Prong One:
Under Prong One, the analysis identifies whether the claim recites a judicial exception, including mathematical concepts, mental processes, or certain methods of organizing human activity, before considering any additional elements.
Independent Claims Analysis:
Claim 1 is representative because claims 7 and 8 perform the same estimating operations. Claim 7 recasts the invention as a device, and claim 8 recasts it as a non-transitory medium, but the judicial-exception analysis is unchanged.
(a) A method for estimating tissue-level information which is performed in a computing device;
(b) acquiring first omics data for a target tissue by analyzing a sample of the target tissue;
(c) acquiring second omics data for a plurality of cells associated with the target tissue and cultured in an in vitro environment;
(d) calculating a similarity between the target tissue and each of the plurality of cells based on the first omics data and the second omics data; and
(e) estimating information on the target tissue by differentially synthesizing information on the plurality of cells, wherein the similarity calculated for each cell is applied as a weight to the corresponding cell information to perform a weighted synthesis.
Under BRI, the non-bold language covers comparing omics data, determining similarity values, applying those values as weights, and combining cell-level information to estimate tissue-level information.
Limitations (d)-(e) recite a mathematical concept because they require “calculating a similarity,” applying it “as a weight,” and performing “weighted synthesis.” Spec. 0053 confirms a “weight sum” using similarities as weights. Refer to MPEP § 2106.04(a)(2), Mathematical Concepts, which covers mathematical calculations and textual mathematical relationships.
Limitations (d)-(e) also reasonably recite a mental-process-type evaluation at the claimed level because the claim does not specify how similarity is compared; it broadly covers comparing information, judging relative closeness, and forming an estimate. Refer to MPEP § 2106.04(a)(2), Mental Processes.
For example an analyst using tissue and cell data tables could compare profiles, identify closer cells, assign larger weights, and combine listed values into an estimate. The operative rule is comparison → weighting → weighted estimate.
Dependent Claims Analysis:
Claim 2 narrows the data source to cell-line information and inherits claim 1’s abstract idea. Claims 3-4 add feature vectors, vector similarity, and vector-space distance, all mathematical calculations. Claim 5 adds model classes and confidence scores, another mathematical scoring approach. Claim 6 applies the same weighted synthesis to drug-effect information. Refer to MPEP § 2106.04(a)(2).
Prong One is satisfied for claims 1-8.
Prong Two:
Under Prong Two, the analysis determines whether the additional elements integrate the identified judicial exception into a practical application, considering the claim as a whole rather than the exception alone.
Independent Claims Analysis:
The additional elements are the computing device, sample-based first omics acquisition, in vitro-cell second omics acquisition, claim 7’s memory and processor, and claim 8’s non-transitory recording medium.
Individual Additional Elements Evaluation:
The computing device, memory, processor, and medium do not integrate the exception because they merely execute or store the similarity-weighting process. Refer to MPEP § 2106.04(d) and § 2106.05(f), mere instructions to implement an abstract idea on a computer.
The sample-analysis and in vitro-cell limitations gather input data for the calculation. They do not recite a particular assay, an integral machine, transformation of an article, treatment or prophylaxis, or an improvement to computer function. Refer to MPEP §§ 2106.05(b), 2106.05(c), 2106.05(g), and 2106.04(d)(2).
Spec. 0036 describes improved accuracy, but that improvement arises from “calculating a similarity” and “synthesizing cell-level information” the judicial exception itself. Refer to MPEP § 2106.05(a), which requires an improvement to computer functionality or another technology, not merely a better abstract result.
Viewed as a whole, the additional elements collect biological data and perform weighted synthesis on ordinary computing components. This is a field-limited use of the exception, not a practical application. Refer to MPEP §§ 2106.04(d), 2106.05(e), and 2106.05(h).
Dependent Claims Analysis:
Claim 2 adds cell-line context, and claim 6 adds drug-effect information. Claims 3-5 add vector, distance, model, and confidence-score details. These limitations narrow the data or deepen the mathematical analysis, so they do not satisfy Prong Two.
Prong Two is not satisfied; Step 2B is reached.
Step 2B:
Under Step 2B, the analysis determines whether the additional elements, individually or as an ordered combination, add significantly more than the judicial exception and amount to an inventive concept.
Independent Claims Analysis:
The same additional elements are evaluated: the computing device, sample-based omics acquisition, in vitro-cell omics acquisition, memory and processor, and non-transitory recording medium.
Individual Additional Elements Evaluation:
The computing elements do not add an inventive concept. Spec. 0037 states that the device may be a notebook, desktop, laptop, or “any type of device equipped with a computing function.” Spec. 0084 identifies a CPU, MPU, MCU, GPU, or “any type of processor well known in the art.” Spec. 0085 identifies RAM-type memory, and 0088 identifies storage including “any computer-readable recording medium well-known in the art.” Refer to MPEP §§ 2106.05(d) and 2106.07(a).
The data-source elements likewise do not add significantly more. Spec. 0039 states that cell-line information may be “easily obtained from a disclosed database” or at “low experimental cost,” and 0047 states that target-tissue data may be acquired by sample analysis “but is not limited thereto.” Refer to MPEP §§ 2106.05(g), 2106.05(h), and 2106.05(d).
Claim 5’s model limitation does not add an inventive concept because Spec. 0072 permits traditional models, including a decision tree, support vector machine, or logistic regression. Refer to MPEP §§ 2106.05(d) and 2106.07(a).
The ordered combination does not add significantly more because the claimed sequence uses the additional elements only to obtain data and execute similarity-weighted synthesis. The tissue-level estimate results from the abstract calculation itself, not from an unconventional technical arrangement. Refer to MPEP §§ 2106.05(d), 2106.05(f), and 2106.05(g).
Dependent Claims Analysis:
Claim 2’s cell-line data, claims 3-4’s vector and distance calculations, claim 5’s classification and confidence score, and claim 6’s drug-effect information do not add an inventive concept. They specify inputs, outputs, or mathematical paths for the same weighted estimation. Refer to MPEP §§ 2106.04(a)(2), 2106.05(d), and 2106.07(a).
Claims 1-8 satisfy Step 1. Prong One is satisfied because limitations (d)-(e) recite mathematical concepts and, secondarily, a mental-process-type comparison or evaluation at the claimed level. The additional elements neither integrate the exception nor add significantly more. The § 101 rejection is maintained.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 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.
Claims 1-2, 6-8 are rejected under 35 U.S.C. 103 as being obvious by Szeto (US 2018/0190381 A1) and in further view of Alemi US20090132460.
Claim 1. Szeto teaches, A method for estimating tissue-level information which is performed in a computing device, the method comprising: (Szeto, par. 0002-0005)
acquiring first omics data for a target tissue by analyzing a sample of the target tissue; (Szeto, par. 0011-0012, 0043, 0026-0027 “ biopsy samples to obtain the omics data“, 0041, “Collected from individual cancer samples “)
Szeto describes a system and method that uses known cell line genomics and drug-response data. Szeto teaches collecting patient genomic-scale data from individual cancer samples (i.e., analyzing a sample of the cancer/tumor tissue using microarray or sequencing) and using that patient omics/pathway model data as input to the drug-response prediction pipeline.
acquiring second omics data for a plurality of cells associated with the target tissue and cultured in an in vitro environment; (Szeto, par. 0011-0012, 0043, 0026-0027, 0036, 0041, 0044, abstract “cell line genomics”)
Szeto describes a system and method that uses known cell line genomics and drug-response data. Szeto expressly treats the cell-culture/tissue omics/pathway models as the inputs used to build the predictor library and then evaluate patient data (e.g., cell-line pathway models → response predictors; patient pathway model → test models).
calculating a similarity between the target tissue and ; (Szeto, par. 0011, 0026-0027,0037- 0039, 0043)
Szeto compares patient tissue data against cell-derived prediction models and treats a better match as a higher score. Szeto explains that actual patient data are run through response predictors built from cell/tissue omics data, producing prediction scores. It then states that the standardized score shows the “conformance of the patient data set with the performance of the response predictor,” and that when the patient dataset is “more similar to the original dataset used in the calculation of a prediction model, a higher prediction score is observed”.
and estimating information on the target tissue . (Szeto, par. 0011-0012, 0026-0027, 0036-0039, 0042, Figure 1A-1c, Figure 2A)
Szeto describes, drug response in patients for prediction using omics data and biological processes derived from omics data.
Szeto discloses estimating information on the target tissue because patient omics/pathway data produce prediction scores that predict the patient’s drug response.
Szeto describe the majority of the limitations above, except by the strikethrough parts.
Alemi teaches the missing per-item weighting relationship. Alemi determines similarities between an individual and other cases, assigns a similarity score, and calculates the expected outcome by using the product of the similarity score and the outcome (Alemi, [0032], [0042]). Alemi then makes the “each” relationship express: Each case in the database can be weighed based on its similarity (Alemi, [0077]); for medication outcomes, Alemi calculates the predicted outcome over database records using the similarity-weighted outcomes for case j (Alemi, [0075]).
It would have been obvious to a POSITA merely applies Alemi’s known similarity-weighted aggregation at Szeto’s scoring stage, so each cell-line response contributes in proportion to its similarity to the target tissue. The predictable result is an estimated tissue-level drug-response output produced by weighted synthesis of the plurality of corresponding cell-line information.
Claim 2. Szeto in further view of Alemi teaches, The method of claim 1, wherein the second omics data include omics data for cell lines cultured in an in vitro environment, and the information on the plurality of cells includes information on the cell line. (Szeto, par. 0036, 0043)
Claim 6. Szeto in further view of Alemi teaches, The method of claim 1, wherein the estimating of the information on the target tissue includes: estimating a drug effect on the target tissue by synthesizing drug effect information on the plurality of cells. (Szeto, See at least, fig, 5-6, par. 0011-0012, 0036, 0039, 0042, Abstract)
Szeto discloses estimating a drug effect on the patient's target tissue by synthesizing drug effect information from multiple cell lines.
Note: Claims 7-8 are rejected with the same analysis above to being very similar to claim 1.
Claim(s) 3-4 is/are rejected under 35 U.S.C. 103 as being unpatentable over Szeto (US 2018/0190381 A1) and in further view of Alemi US20090132460 with Newman - WO2016118860A1.
Claim 3. Szeto in further view of Alemi teaches, The method of claim 1, wherein the calculating of the similarity includes: generating a first feature vector from the first omics data; par. 0027, 0013, 0038, 0040, 0043, Figure 2A)
generating a second feature vector from the second omics data; par. 0027, 0013, 0038, 0040, 0043, Figure 2A)
and
calculating the similarity based on a vector similarity between the first feature vector and the second feature vector. (par. 0016, 0020-0021, 0027, 0013-0014, 0038-0039, 0040, 0043, Figure 2A)
Szeto teaches calculating the similarity (under BRI, a patient-specific score indicative of closeness) between patient omics/pathway data and pre-built response predictors by computing null-model standardized scores and ranking the predictors: Generate large library of response predictors … using multiple distinct machine learning algorithms … Generate null models … Standardize raw data … obtain normalized results … rank results…Szeto further explains that the system “uses a patient pathway model to generate respective test models … [and] ranks the respective test models by their respective gain” and that “the difference in standardized score is then used for ranking … where the original patient dataset is more similar … a higher prediction score is observed”.
However, Szeto fails to disclose generating a first feature vector from the first omics data; generating a second feature vector from the second omics data; and calculating the similarity based on the vector similarity between the first feature vector and the second feature vector.
Newman (WO 2016/118860 A1) teaches the Missing Element in bold, describing explicit vector-to-vector similarity/distance between omics-derived feature vectors: Newman defines a feature profile m and compares profiles using a difference measurement that “may be a correlation coefficient … Euclidean distance” (para. [00150]); related claims teach that the “lowest error is obtained using a Pearson … Spearman … Euclidean distance” (para. [0008]/claim 9). Newman also normalizes vector norms when constructing nulls—“wherein m and m have the same Euclidean norm (|m|=|m*|)”* (para. [0009])—confirming standard vector-space operations on omics feature vectors.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention to combine the teachings of Szeto with Newman because both references address the shared purpose of quantifying how a sample’s omics profile relates to known cellular signatures/models to drive downstream decisions. Szeto (paras. [0038]–[0039], Newman (para. [0008-0009],[00150]), a POSITA would obtain the claimed calculating the similarity … based on the vector similarity between the first feature vector and the second feature vector, using well-specified correlation or Euclidean metrics that directly quantify vector closeness while retaining Szeto’s overall scoring/ranking pipeline. (Newman paras. [00150], [0008]).
A person of ordinary skill in the art would have been motivated to integrate the vector-similarity between first and second feature vectors from Newman into the system of Szeto to achieve the benefit of “ lowest error is obtained using a Pearson … Spearman … Euclidean distance” (para. [0008]), and identifies correlation/Euclidean difference measurements for profile comparison (para. [00150]).
Furthermore, the proposed combination is obvious because use of known technique to improve similar methods in the same way. The technique of vector-space similarity (e.g., Pearson correlation, Euclidean distance) between omics feature vectors is known in the art (as evidenced by Newman, paras. [0008], [0009], [00150]) for improving omics-based inference by quantifying profile closeness with standard distance/similarity metrics. Applying this known technique to the analogous Szeto framework (patient omics → scoring → ranking) predictably improves it in the same manner to achieve the claimed vector-similarity computation between the first and second feature vectors a direct, quantitative similarity measure feeding Szeto’s selection/ranking step.
A PHOSITA would have had a reasonable expectation of success in combining the references because the modification requires only routine data-processing: Szeto already ingests patient omics/pathway data and computes scores against a library (Fig. 2A) (para. [0021]); swapping-in or adding Newman’s vector similarity (correlation/Euclidean) merely requires forming two feature vectors (tissue, cell) and computing a standard metric that Newman specifies and norm-controls (paras. [0009], [00150]). Both references provide enabling detail for their respective steps, and their data types (gene-expression/omics) and computational context align.
Claim 4. Szeto in further view of Alemi with Newman teaches, The method of claim 3, wherein the vector similarity is calculated based on a distance between the first feature vector and the second feature vector in a vector space. (Newman, par. 0008)
Claim(s) 5 is rejected under 35 U.S.C. 103 as being unpatentable over Szeto (US 2018/0190381 A1) and in further view of Alemi US20090132460 with Tan, Y., & Cahan, P. (2018). SingleCellNet: a computational tool to classify single cell RNA-Seq data across platforms and across species. Posted on bioRxiv. https://doi.org/10.1101/508085, PTO-892-U.
Claim 5. Szeto in further view of Alemi teaches, The method of claim 1, wherein the calculating of the similarity includes: inputting the first omics data into a classification model that receives omics data (Szeto, par. 0027 “machine learning algorithm that uses omics data and/or pathway models generated from a cell culture or tissue” par. 0038 “actual patient data using null models for each of the response predictors in the database”, par. 0031 “patient dataset (omics data or pathway model)” par. 0013 “ machine learning system may uses various classifiers”)
and Szeto, par. 0024 “representation of dasatinib sensitivity sorted by cell line type”, par. 0025“representation of dasatinib sensitivity sorted by human TCGA tumor type”, fig.5-fig.6)
and calculating the similarity .(Szeto, par. 0039 “here the original patient dataset is more similar to the original dataset used in the calculation of a prediction model, a higher prediction score is observed (as the prediction model is optimized for predicting a response to a specific drug).”)
Szeto teaches inputting the first omics data into a classification model that receives omics data and obtaining a score, of the Claim 5, the method of claim 1, wherein the calculating of the similarity includes: inputting the first omics data into a classification model that receives omics data and outputs classes of cells to obtain a confidence score for each class; and calculating the similarity based on the obtained confidence score, Szeto describes inputting patient omics data into machine learning classifiers built from cell-line data to generate prediction scores, and teaches that higher scores indicate greater similarity between the patient's data and the cell-line data underlying the model. However, Szeto does not describe a classification model that outputs classes of cells to obtain a confidence score for each class or calculating the similarity based on the obtained confidence score.
Tan & Cahan teaches a classification model that receives omics data and outputs classes of cells to obtain a confidence score for each class and calculating the similarity based on the obtained confidence score of the Claim 5. Under MPEP 2111, a confidence score for each class is reasonably interpreted as any numerical value a classifier produces for each output category indicating the degree of match between the input and that category. Tan & Cahan describes a multi-class Random Forest classifier trained on cell-type-annotated single cell RNA-Seq data that takes query omics data as input, classifies each cell against all reference cell types, and outputs a per-class classification score summing to one quantitative value for each cell-type class that directly measures the degree of match. These per-class scores then serve as the quantitative measure of similarity, which the authors expressly describe as more informative than binary categorical assignment. (Tan & Cahan, See at least, Summary, page 2: SingleCellNet, which addresses these issues and enables the classification of query single cell RNA-Seq data in comparison to reference single cell RNA-Seq data; Discussion, pag. 7: a multi-class Random Forest classifier is then trained with the transformed training data; a classification score is generated for each query cell; Figure 2, page 8 description: The classification score for each column/cell sums up to 1, with a range from 0 black to 1 yellow; Discussion, page 7: a quantitative, rather than a binary, metric of identity is informative).
The combination of Szeto + Tan & Cahan makes obvious the full limitation under inputting the first omics data into a classification model that receives omics data and outputs classes of cells to obtain a confidence score for each class; and calculating the similarity based on the obtained confidence score because Szeto seeks to quantify how similar a patient's omics profile is to reference cell-line data and already inputs patient omics data into trained classifiers for this purpose. Tan & Cahan teaches a multi-class Random Forest cell-type classifier that takes omics data and produces per-class confidence scores a known technique that directly measures the tissue-to-cell similarity Szeto's system requires. A person of ordinary skill would implement Tan & Cahan's classifier within Szeto's pipeline as the similarity calculation step, and the predictable result is per-class confidence scores quantifying how closely the patient tissue matches each reference cell type. (See at least, Szeto, par. 0039: where the original patient dataset is more similar to the original dataset used in the calculation of a prediction model, a higher prediction score is observed; par. 0013: contemplated machine learning system may use various classifiers, including...random forest algorithms; Tan & Cahan, Discussion page 7: a multi-class Random Forest classifier is then trained with the transformed training data; a classification score is generated for each query cell; Discussion, page 7: a quantitative, rather than a binary, metric of identity is informative).
A skilled artisan who read Szeto's application would combine Tan & Cahan with Szeto because both references are in the same field of endeavor under MPEP § 2141.01(a), computational analysis of omics data from cells and tissues using machine learning classifiers and Szeto expressly relies on similarity between patient omics data and cell-line reference data as the mechanism driving accurate drug prediction, directly pointing toward the need for a quantitative cell-type classification method that Tan & Cahan provides. (See at least, Szeto, par. 0002: the field of the invention is systems and methods of predicting drug responses using omics information; par. 0039: where the original patient dataset is more similar to the original dataset used in the calculation of a prediction model, a higher prediction score is observed; Tan & Cahan, Summary, page 2: SingleCellNet, which addresses these issues and enables the classification of query single cell RNA-Seq data in comparison to reference single cell RNA-Seq data).
Integrating Tan & Cahan's cell-type classification into Szeto's pipeline solves the problem of quantifying which cell types in the reference library are most similar to the patient's tissue, replacing Szeto's indirect similarity measurement, based on drug-response prediction score magnitude with a direct, multi-class cell-type classification that produces explicit per-class confidence scores. The benefit is a more precise and interpretable similarity assessment, and the result is predictable because both systems already process gene-level omics data through machine learning classifiers to produce quantitative scores. (See at least, Szeto, par. 0039: a higher prediction score for a response predictor using a patient dataset pathway model or omics data indicates that the patient's response to treatment with the drug used in the response predictor may also be accurately predicted; Tan & Cahan, Discussion, page. 7: a quantitative, rather than a binary, metric of identity is informative, or when the presence of shared cell types across datasets is unclear).
A person of ordinary skill would have a reasonable expectation of success because both references process gene-level omics data through machine learning classifiers. Szeto explicitly teaches using expression data and copy number data as omics inputs, and Tan & Cahan's classifier operates on gene expression data the same category of input.
Relevant Prior Art:
US20150294062:
A method for identifying a target molecular profile associated with a target cell population. A set of reference molecular profiles and a set of sample molecular profiles are received. Each sample molecular profile is associated with a sample cell from a sample cell population, which includes a mixture of target cells and reference cells. Each of the sample molecular profiles is indicative of a respective target molecular profile. An average target molecular profile is calculated. A proportion value is calculated for each sample molecular profile. A respective target molecular profile is calculated for each sample molecular profile based on the respective calculated proportion value and a closest similarity to the average target molecular profile. Refer to abstract
Conclusion
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 JOSHUA DAMIAN RUIZ whose telephone number is (571)272-0409. The examiner can normally be reached 0800-1800.
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, Shahid Merchant can be reached at (571) 270-1360. 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.
/JOSHUA DAMIAN RUIZ/Examiner, Art Unit 3684
/Shahid Merchant/Supervisory Patent Examiner, Art Unit 3684
Prosecution Insights