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 .
Election/Restrictions
Claims 17-21 and 24-27 are withdrawn from further consideration pursuant to 37 CFR 1.142(b) as being drawn to a nonelected invention (claims 24-27) and species (claims 17-21), there being no allowable generic or linking claim. Election was made without traverse in the reply filed on 22 May 2026.
Claim Status
Claims 17-21 and 24-27 are withdrawn.
Claims 1, 4, 6-8, 11, 13, 15, and 22-23 are pending.
Priority
This application is a 371 of PCT/US2021/015848, filed 01/29/2021, which claims benefit of application no. 62/967,941, filed 01/30/2020. The instant application has the effective filing date of 30 January 2020.
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 09/16/2022, 08/07/2024, 05/23/2025, 09/09/2025, and 01/08/2026 are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statements have been considered by the examiner.
Drawings
The drawings, submitted on 07/29/2022, have been accepted by the examiner.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1, 13, 15, and 22-23 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention, as detailed below.
Claims 1, 13, 15, and 22-23 recite the limitation “optimized”, “optimizing”, or “optimal” as relating to “combinations of biopolymer sequences”, “linear combinations of conformal inference intervals”, “number of experiments”, and “wet-lab resources”, wherein the optimization terms are undefined. Please define the metes and bounds of the terms as it relates to the disclosed variables.
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, 4, 6-8, 11, 13, 15, and 22-23 are rejected under U.S.C 101 because the claimed invention is directed to abstract ideas without significantly more as detailed in the analysis below.
Eligibility Step 1: Subject matter eligibility evaluation in accordance with MPEP § 2106:
Claims 1, 4, 6-8, 11, 13 and 15 are directed to a statutory category (method).
Claims 22 is directed to a statutory category (product).
Claims 23 is directed to a statutory category (system).
[Eligibility Step 1: YES]
Eligibility Step 2A: This step determines whether a claim is directed to a judicial exception in accordance with MPEP § 2106.
Eligibility Step 2A -- Prong One: Limitations are analyzed to determine if the claims recite any concepts that could equate to a judicial exception (i.e. abstract idea, law of nature, or natural phenomenon).
Recitations of judicial exceptions:
Claims 1, 22, and 23: training a machine learning model using a plurality of observed biopolymer sequences and labeled biopolymer sequences corresponding to each observed biopolymer sequence (mathematical concept, mental process)
determining a plurality of candidate biopolymer sequences to observe having a highest predicted value of the labeled biopolymer sequences based on the machine learning model; (mental process)
for each candidate biopolymer sequence, determining a conformal inference interval representing a likelihood that the candidate biopolymer sequence has the predicted value of the labeled biopolymer sequences; (mental process, mathematical concept)
selecting at least one candidate biopolymer sequence having an optimized linear combination of the conformal inference interval and the predicted value of the labeled biopolymer sequences. (mental process)
Claim 4: wherein determining the conformal inference interval is based on a second set of observed biopolymer sequences. (mathematical concept, mental process)
Claim 6: The computer-implemented method of Claim 4, wherein determining the conformal inference interval further includes: calculating a residual interval based on each output of the machine learning model for the second set of observed biopolymer sequences and corresponding labeled biopolymer sequences corresponding to each of the second set of biopolymer sequences; (mathematical concept)
for each output of the machine learning model, calculating an average distance to a plurality of nearest neighbors of the observed biopolymer sequences within a metric space; (mathematical concept)
and calculating a conformal score based on a ratio of the residual to a sum of the average distance and a constant. (mathematical concept)
Claim 7: The computer-implemented method of Claim 4, wherein selecting the at least one candidate biopolymer sequence includes:
calculating an average distance in a metric space to a plurality of nearest neighbors in the metric space; (mathematical concept)
generating a confidence interval based on the at least one candidate biopolymer sequence and the average distance; (mathematical concept)
and selecting at least one candidate biopolymer sequence based on the confidence interval. (mental process)
Claim 8: wherein the conformal interval is at least 50% and at most 99%. (mathematical concept, mental process)
Claim 11: The method of Claim 1, wherein the predicted value is a function value of the biopolymer sequences, wherein the function is one or more of binding affinity, binding specificity, catalytic activity, enzymatic activity, fluorescence, solubility, thermal stability, conformation, immunogenicity, and any functional property of biopolymer sequences. (mathematical concept, mental process)
Claim 13: training a model to approximate labeled biopolymer sequences of initial sampled from a plurality of observed sequences; (mathematical concept, mental process)
For a particular batch of the plurality of sequences, having labeled biopolymer sequences generated by a trained model and conformal interval for each observed sequence, choosing at least one sequence from the plurality of observed sequences that optimized a combination of the labeled biopolymer sequences generated by the trained model and the conformal interval; (mental process)
recalculating the conformal interval for the remaining sequences (mathematical concept)
Claim 15: The method of Claim 13, further comprising identifying an optimal number of batch experiments to run in parallel based on optimizing wet-lab resources (mathematical concept, mental process)
Step 2A - Prong One Analysis:
Determining, selecting, or identifying secondary data based on observables equate to analysis techniques that require no more than mere observations of data that can be completed with only the human mind and pen/paper. As such, limitations that recite these processes fall under the mental process grouping of abstract ideas. Limitations that merely recite additional information to the mental processes are similarly categorized (claims 4, 8, and 11).
Furthermore, limitations that reciting calculating scores and metrics such as averages, sums, and likelihoods equate to performing mathematical calculations on information to derive secondary data. Equations that recite such techniques fall under the mathematical concepts grouping of abstract ideas. Limitations that merely recite additional information to the mathematical concepts are similarly categorized (claims 4, 8, and 11).
Therefore, the claims appear to recite judicial exceptions.
[Eligibility Step 2A – Prong One: YES]
Eligibility Step 2A – Prong Two: A claim that integrates a judicial exception into a practical application will apply, rely on, or use the judicial exception in a manner that imposes a meaningful limit on the judicial exception. If the claim contains no additional claim elements beyond the abstract idea, the claim fails to integrate the abstract idea into a practical application (MPEP 2106.04(d)). This step analyzes limitations that are considered additional elements to determine if they integrate the judicial exceptions into practical application. Additional elements are recited, categorized, and analyzed below.
Computer Elements:
Claims 1 and 13: computer-implemented method
Claim 22: non-transitory computer readable medium storing instructions for optimizing design of biopolymer sequences thereon, wherein the instructions, when executed by the processor, cause the processor to:
Neural Network Elements:
Claim 4: wherein the machine learning model is a neural network fine-tuned using the observed biopolymer sequences and their labels
Step 2A - Prong Two Analysis:
Generic computer components and implementations provide mere instructions to implement the abstract ideas per Alice Corp., 573 U.S. at 223, 110 USPQ2d at 1983. See also 573 U.S. at 224, 110 USPQ2d at 1984.
As the neural network components are recited at a high level of generality, it appears to merely act as a tool to enact the judicial exceptions per MPEP 2106.05 (f).
As such, when evaluated separately or in the context of a whole claimed invention, the limitations do not integrate the judicial exceptions into practical application.
[Eligibility Step 2A – Prong Two: NO]
Eligibility Step 2B: Claim elements are probed for inventive concept equating to significantly more than the judicial exception (MPEP 2106.04(II)).
Step 2B Analysis:
The computer components are further found to be well-understood, routine, and conventional per Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 for receiving or transmitting data over a network; and Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); OIP Techs., 788 F.3d at 1363, 115 USPQ2d at 1092-93 for storing and retrieving information in memory.
The neural network components are found to be well-understood, routine, and conventional per Ching et al. (J Royal Society Interface; Vol. 15: 141; 2018), which reviews obstacles for deep learning in biology and medicine; and teaches fine-tuning neural networks using a small set of positive examples of SMILES molecules and their targets (page 25, column 2) (page 32, column 1).
[Eligibility Step 2B: NO]
As such, claims 1, 4, 6-8, 11, 13, 15, and 22-23 are directed to judicial exceptions and rejected under 35 U.S.C 101.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claim 13 is rejected under 35 U.S.C. 102(a)(1) as being anticipated by Giblin et al. (J Chem Inf Model; Vol. 58: 1870-1888; 2018).
Claim 13 is directed to a computer-implemented method for optimizing design of biopolymer sequences comprising that includes: training a model to approximate labeled biopolymer sequences of initial samples from a plurality of observed sequences; choosing at least one sequence from the plurality of observed sequences that optimizes a combination of the labeled biopolymer sequences generated by the trained model and the conformal interval; and recalculating the conformal interval for the remaining sequences.
Giblin et al. describes prospectively validated proteochemometric models for the prediction of small-molecule binding to Bromodomain Proteins.
Giblin et al. teaches screening compounds in the liquid sample library at AstraZeneca against the model to predict activity against four bromodomains (page 5, column 1); applying conformal prediction using the confidence levels of 0.7, 0.8, and 0.9 and the corresponding significance values of 0.3, 0.2, and 0.1 to compare new compound-target pair p-values to the model training set values to shortlist compounds for experimental testing (page 5, column 1); and conducting Leave-one-target-out (LOTO) validation by sequentially removing all compound-target pairs associated with one bromodomain target, training a model based on the remaining data, and predicting for the hold-out target data points (page 4, column 2).
Giblin et al. further teaches such proteochemometric (PCM) classification models can be applied to future virtual screening and compound design, including off-target prediction for bromodomains (page 1, column 1).
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The 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.
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 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claim 1, 4, 11, and 22-23 are rejected under 35 U.S.C. 103 as being unpatentable over Phlyophisut et al. (BMC Bioinformatics; Vol. 20: 270; 2019) in view of Toccaceli et al. (Machine Learning; Vol. 108: 489-510; 2019).
Phlyophisut et al. describes a deep neural network model, MHCSeqNet, that performs universal MHC binding prediction.
Claims 1, 22, and 23 are directed to computer-implemented methods, mediums, and systems that train machine learning models using observed biopolymer sequences and corresponding labeled sequences; and determining which observed sequence has the highest predicted value of the labeled sequence.
Phlyophisut et al. teaches training models using an amino acid sequence-based representation that permits inference across alleles (page 5, column 1) and deep learning to predict the probability of binding between peptide and MHC allele, where a prediction of 0.0 indicates no binding and 1.0 indicates a strong binding (page 2, column 2).
Phylophisut et al. further teaches implementing MHCSeqNet using Python 3 and several other software tool packages (page 2, column 2), which inherently require the presence of a processer, memory, and computer-readable medium to implement the method described.
Claims 1, 22, and 23 are further directed to determining a conformal inference interval by calculating a measure of likelihood that each observed sequence has the same value as its label; and selecting at least one observed sequence with an optimized linear combination of its conformal inference and predicted label value of the candidate sequence.
Phylophisut et al. teaches using Adam optimization algorithm (page 4, column 1), false discovery rate (page 6, fig. 2), and F1 scores (page 4, column 2) as metrics to ensure the model predicts peptide-MHC binding with high accuracy (page 2, column 1); and selecting potent, immunogenic neoepitopes based on the premise that effective neoepitope should bind with high affinity MHCs (page 1, column 1).
Claim 4 is directed to the machine learning model being a neural network that is fine-tuned using the observed biopolymer sequences and their label and calculating the conformal inference based on a second set of observed biopolymer sequences.
Phylophisut et al. teaches MHCSeqNet employs neural network architectures (page 1, column 1); pre-trains the one-hot, 1-gram, and 3-gram peptide representations on three different datasets (page 5, column 2); trains the 1-gram model using two combined sequence datasets, which slightly improve performance (page 5, column 2); and further improvements are possible through the use of additional model fine-tunings (page 2, column 1).
Phylophisut et al. further teaches training the model on the combination of Swiss-Prot proteins and simulated human proteasome-cleaved 9-mers (page 6, column 1); evaluating performance of the models via F1 and AUC using a different MHC class I binding affinity dataset (page 6, column 2); and additionally, to ensure that this test is independent from the evaluation using binding affinity data, all peptidome entries that overlap with our training dataset were removed from consideration (page 7, column 2).
Claim 11 is directed to the predicted value being a function of the observed biopolymer sequence and one or more of the following properties: binding affinity, binding specificity, catalytic activity, enzymatic activity, fluorescence, solubility, thermal stability, conformation, or immunogenicity.
Phylophisut et al. teaches MHCSeqNet outperforms NetMHCPan and MHCflurry on both MHC binding affinity and ligand peptidome datasets (page 2, column 1).
Phylophisut et al. does not explicitly teach the conformal inference is a measure of likelihood that each observed sequence has the same value as its label; nor that the selected neoepitopes have an optimized linear combination of its conformal inference and predicted label value of the candidate sequence (claims 1 and 22-23).
Toccaceli et al. describes a method of combining inductive Mondrian conformal predictors.
Toccaceli et al. teaches Conformal Predictors (CP) can output a confidence measure, i.e. a p value for the hypothesis that the label of a test object x is y (page 1, column 1); the CP values can be assigned a weight by (b) computing a linear combination of the base CP p values using the predicted weights (page 11, column 1); and in order to have just one CP metric, combining Precision and Recall into the F1 score (page 14, column 1).
Toccaceli et al. further teaches applying CP combination methods for the machine learning prediction of the activity of chemical compounds towards biological targets of interest (page 11, column 1); and ranking test objects according to their p values in search of those that are most likely to be Active (page 18, column 1).
Therefore Phylophisut et al. teaches a method of using a neural network to predict binding affinity of amino acid sequences to MHC alleles, confirming accuracy using F1 scores, and selecting neoepitopes based on the results. Toccaceli et al. teaches using F1 scores as a conformal predictor, computing a linear combination of the CP and predicted label value, and using the results to rank predicted properties of pharmaceutical compounds. As such, it would be obvious to one of ordinary skill in the art to apply technique of using the linear combination of F1 score, as the conformal predictor and predicted label values to the technique of selected sequences with the highest ranked accuracy with a reasonable expectation of success.
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Phylophisut et al. (BMC Bioinformatics; Vol. 20: 270; 2019) in view of Toccaceli et al. (Machine Learning; Vol. 108: 489-510; 2019), as applied to claims 1, 4, 11, 22, and 23, and in further view of Romano et al. (IDS, filed 09/16/2022; NPL; cite no. 7; 2019) and Papadopoulos et al. (J Artificial Intelligence Research; Vol. 40; 2011).
Phylophisut et al. in view of Toccaceli et al. teach machine learning models that use the linear combination of predicted values and F1 scores, as conformal predictors, to ensure the selection of sequences with accurately predicted functional properties.
Claim 6 is directed to calculating a residual interval based on each output of the machine learning model for the second set of observed biopolymer sequences and corresponding labeled biopolymer sequences corresponding to each of the second set of biopolymer sequences.
Phylophisut et al. and Toccaceli et al. do not teach the method according to claim 6.
Romano et al. describes conformalized quantile regression.
Romano et al. teaches the conformal method begins by splitting the training data into two disjoint subsets: a proper training set and calibration set (page 4, column 1); and computing the absolute residuals on the calibration set, according to equation 7, below (page 4, column 1):
Equation 7: Ri = | Yi – u (Xi) |
Claim 6 is further directed to for each output of the machine learning model, calculating an average distance to a plurality of nearest neighbors of the observed biopolymer sequences within a metric space; and calculating a conformal score based on a ratio of the residual to a sum of the average distance and a constant.
Romano et al. teaches locally adaptive split conformal prediction, is an approach to making conformal prediction adaptive to heteroskedascity that starts from the observation that one can replace the absolute residuals in equation (7) by any other loss function that treats the data exchangeably.; and in this case, the absolute residuals Ri are replaced by the scaled residuals:
PNG
media_image1.png
64
341
media_image1.png
Greyscale
where ˆσ(Xi) is a measure of the dispersion of the residuals at Xi (page 7, column 1); and adding a hyper-parameter γ > 0 as a constant offset to the scale estimator ˆσ(x), such that the scaled residuals become equation 17 (page 8, column 1):
PNG
media_image2.png
59
148
media_image2.png
Greyscale
Romano et al. further teaches evaluating locally adaptive conformal prediction (Section 5) using the same three underlying regression algorithms (page 9, column 1), such as ridge local, where the mean absolute deviation (MAD) estimator ˆσ is k-nearest neighbors with k = 11 (page 9, column 1); and setting the hyper-parameter γ in equation 17 to 1, which improves performance considerably compared to γ = 0 (page 9, column 1).
Romano et al. further teaches conformal quantile regression is a new way of constructing prediction intervals that combines the advantages of conformal prediction and quantile regression; provably controls the miscoverage rate in finite samples, under the mild distributional assumption of exchangeability, while adapting the interval lengths to heteroskedasticity in the data; and expecting the ideas behind conformal quantile regression to be applicable in the related setting of conformal predictive distributions (page 11, column 1).
Therefore Romano et al. provides sufficient motivation for one of ordinary skill in the art to calculate a ratio of the residual to a distance of at least two nearest neighbors and a constant when in order to adapt interval lengths to heteroskedasticity in the data when making conformal predictions.
Romano et al. does not teach using the average distance of the k-nearest neighbors.
Papadopoulos et al. describes regression conformal prediction with nearest neighbors.
Papadopoulos et al. teaches in order to create any CP we need to define a nonconformity measure based on the underlying algorithm in question (page 6, column 1), then improve the typical regression nonconformity measure by normalizing it with the expected accuracy of the underlying method (page 10, column 1); in which we could use λki we as a measure of accuracy, that is consistent across different data sets and compares the sum (page 11, column 1) of the distances between xi and its k nearest neighbours with the median of the distances of all training examples from their k nearest neighbours (page 10, column 1); and defining ai as:
PNG
media_image3.png
69
131
media_image3.png
Greyscale
in which the parameter γ ≥ 0 controls the sensitivity of each measure to changes of λki. (page 11, column 1).
Papadopoulos et al. further teaches the k-Nearest Neighbours algorithms base their predictions on the k training examples that are nearest to the unlabeled example in question according to some distance measure, such as the Euclidean distance; specifically, for an input vector xl+1 the k-Nearest Neighbours Regression (k-NNR) algorithm finds the k nearest training examples to xl+1 and outputs the average or in some cases the median is also used of their labels as its prediction (page 6, column 1).
Therefore Papadopoulos et al. teaches a similar equation in which a known technique is to use the sum of the average distances to k-nearest neighbors. As such, it is obvious to one of ordinary skill in the art to calculating a conformal score based on a ratio of the residual to a sum of the average distance and a constant, based on the teaching of Romono et al. in view of Papadopoulos et al.
Claims 7-8 are rejected under 35 U.S.C. 103 as being unpatentable over Phylophisut et al. (BMC Bioinformatics; Vol. 20: 270; 2019) in view of Toccaceli et al. (Machine Learning; Vol. 108: 489-510; 2019), as applied to claims 1, 4, 11, 22, and 23, and in further view of Cortes-Ciriano et al. (IDS, filed 09/16/2022; NPL; cite no. 5; 2019).
Phylophisut et al. and Toccaceli et al. teach the method of machine learning models that use the linear combination of predicted values and F1 scores, as conformal predictors, to ensure the selection of sequences with accurately predicted functional properties.
Claim 7 is directed to determining which candidate sequences to select by a method that includes: calculating an average distance to at least two nearest neighbors; calculating a confidence interval based on the candidate sequence and average distance; and selecting candidate sequence(s), based on the confidence interval.
Cortes-Ciriano et al. describes a conformal prediction (CP) process applied to QSAR models.
Cortes-Ciriano et al. teaches choosing an appropriate non-conformity measure is essential in regression to maximize efficiency, defined as the average size of the confidence regions; Svensson et al. benchmarked the efficiency of 6 approaches to scale the errors in prediction in the non-conformity function using RF models and 29 public bioactivity data sets including: (v) the average distance to the five nearest neighbours in two-dimensional incremental PCA; and using the non-conformity functions to generate well-calibrated confidence intervals.
Claim 8 is directed to the conformal interval being at or within the range of 50-99%.
Cortes-Ciriano et al. teaches that the confidence level is commonly set at 0.80, as this confidence level represents a generally suitable trade-off between efficiency and validity; and CL of 80% is also reported as 0.80 in the CP literature (page 5, column 1).
Cortes-Ciriano et al. further teaches the generation of a predictive model, such as a Quantitative-Structure Activity Relationship (QSAR) model, or any other property model, consists of encoding the set of molecules for which the variable of interest has been experimentally measured using numerical descriptors (page 3, column 1); relating the covariates to the response variable using a mathematical model such as multiple linear regression, Random Forests (RF), Support Vector Machines (SVM) or Deep Neural Networks; an using these models to predict the activity or property of interest for untested molecules in silico to prioritize for further experimental testing those with higher chances of being active (page 3, column 1).
Therefore Cortes-Ciriano et al. teaches calculating a confidence interval of 80% based on a bioactivity data set, using the average distance to the five nearest neighbors, of a prediction framework analogous to that of Phylophisut et al. in view of Toccaceli et al. As such, it would be obvious to one of ordinary skill in the art to combine the techniques, with each element merely performing the same task as they do separately, with a reasonable expectation of success and improvement to the assurance in accuracy of the system.
Claim 15 is rejected under 35 U.S.C. 103 as being unpatentable over Giblin et al. (J Chem Inf Model; Vol. 58: 1870-1888; 2018), as applied to claim 13, previously, and in further view of Gonzalez et al (Proceedings 19th Inter Conf on AI STATS; Vol. 41; 2016).
Giblin et al. teaches using conformal prediction to select biopolymer sequences with accurately predicted characteristics.
Claim 15 is directed to determining an optimal number of batch experiments to perform in parallel to training the model, based on the amount of available wet-lab resources.
Giblin et al. further teaches selecting 1,139 compounds for prospective experimental testing (page 6, column 1).
Giblin et al. does not teach determining an optimal number of batch experiments to perform in parallel to training the model.
Gonzalez et al. describes Batch Bayesian Optimization via Local Penalization.
Gonzalez et al. teaches often, it is desirable to simultaneously propose batches of parameter values to explore; which is particularly the case when large parallel processing facilities are available, such as computational or physical facets of the process being optimized (page 1, column 1).
Gonzalez et al. teaches many problems, such as the configuration of machine learning algorithms or the experimental design of biological experiments require the optimization of an unknown, possibly noisy, function f (page 1, column 1); Bayesian optimization (BO) has emerged in this scenario as an efficient heuristic to optimize f if function evaluations are costly and the overall number of evaluations must be kept low (page 1, column 2); parallel approaches arise as the natural solution to circumvent the computational bottleneck around these evaluations of f (page 2, column 1); and focusing on cases in which the cost of evaluating f in a batch of points of size nb is the same as evaluating f in a single point (page 2, column 1).
Gonzalez et al. teaches such scenarios appear, for instance, in the optimization of computer models where several cores are available to run in parallel, or in wet-lab experiments when the cost of testing one experimental design is the same as testing a batch of them (page 2, column 1); and confirmation of the effectiveness of the approach is demonstrated through several simulated experiments, an algorithm configuration problem, and a real wet-lab experimental design (page 3, column 1).
Therefore Gonzalez et al. teaches a method of determining an optimal number of batch experiments to perform in parallel to training a computer model where several cores are available. It would be obvious to one of ordinary skill in the art to combine this technique with the method of Giblin et al., with each element merely performing the same function as they do separately, with a reasonable expectation of success and improvement to the system in the form of saved wet-lab resources.
Conclusion
No claims are currently allowed.
Correspondence
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Milana Thompson whose telephone number is (571)272-8740. The examiner can normally be reached Monday - Friday, 9:00-6:00 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, Karlheinz Skowronek can be reached at (571) 272-1113. 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.
/M.K.T./Examiner, Art Unit 1687
Prosecution Insights