Denoising ATAC-Seq Data With Deep Learning

§101 §103 §DOUBLEPATENT

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 . Applicant’s Amendment/Request for Reconsideration-After Non-Final Rejection, filed 5/27/2026, has been entered and fully considered. Status of Claims Claims 1-18 are pending and examined on the merits. Priority Applicant’s claim for the benefit of a prior-filed application under 35 U.S.C. 119(e) or under 35 U.S.C. 120, 121, or 365(c) is acknowledged. Priority of US application 16/236,797 filed 12/31/2018 is acknowledged. Claim Objections Claims 3 and 12 are objected to because of the following informalities: claims 3 and 12 both recite “an ReLU layer”, which should read as “a ReLU layer”. Appropriate correction is required. Claim Rejections - 35 USC § 101 This rejection is maintained from the previous Office Action. Modifications are necessitated by claim amendments. 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. Step 1: Process, Machine, Manufacture or Composition Claims 1-9 are to a “processor”. So a machine or manufacturer, with structural components like "a processor”. Claims 10-16 are to a "system". So a machine or manufacturer, with structural components like "a processor and a memory device”. Claims 17-18 are to a "non-transitory computer-readable medium". So a machine or manufacturer, with structural components like a "non-transitory computer-readable medium". Step 2A Prong One: Identification of an Abstract Idea The claims recite: One or more circuits to provide to one or more circuit to use one or more computer-implemented neural networks to denoise ATAC-seq information comprising a plurality of ATAC-seq counts at each base pair position of a genome that exclude base identity, and to use the one or more computer-implemented neural networks to generate one or more inferences corresponding to the denoised ATAC-seq information based, at least in part, on the plurality of ATAC-seq counts (Claims 1, 10 and 17). ----This step recites “to generate one or more inferences corresponding to the denoised ATAC-seq information” in a general, simple way. This limitation, as drafted, is a process that, under its broadest reasonable interpretation (BRI), covers the performance of the limitation in the mind but for the recitation of generic computing steps. That is, other than reciting “one or more circuits to provide to one or more computer-implemented neural networks,” nothing in the claim element precludes the step from practically being performed in the mind. For example, but for the “one or more circuits to provide to one or more computer-implemented neural networks” language, the claim encompasses a user simply reading the ATAC-seq multiple alignment data and determining inferences in his/her mind. The mere nominal recitation of one or more generic processors/circuits/neural networks do not take the claim limitation out of the mental processes grouping. Thus, the claims recite a mental process. Claim 7 recites: Wherein the neural networks are trained with suboptimal ATAC-seq information using a loss function to compare neural network output to model ATAC-seq information, and adjusting the neural network until the loss function is minimized. ----Training a neural network involves adjustment of many parameters according to mathematical formulas (such as minimizing the loss function). Therefore this claim equates to an abstract idea of mathematical concepts. Claim 8 recites: Wherein the loss function is a mean square error loss function. ----This claim further limit the loss function used in training a neural network. Therefore this claim equates to an abstract idea of mathematical concepts. Claim 16 recites: Wherein the system further comprises: a training engine executable on the processor according to software instructions stored in the memory device, wherein the training engine is configured to: train the neural networks with suboptimal ATAC-seq information using a loss function to compare neural network output to model ATAC-seq information, and adjust the neural network until the loss function is minimized. ----Training a neural network involves adjustment of many parameters according mathematical formulas (such as minimizing the loss function). Therefore this claim equates to an abstract idea of mathematical concepts. Claim 18 recites: wherein the medium further comprises instructions for a training engine executable on the processor, wherein the training engine is configured to train the neural networks with suboptimal ATAC-seq information using a loss function to compare neural network output to model ATAC-seq information, and adjust the neural network until the loss function is minimized. ----Training a neural network involves adjustment of many parameters according to mathematical formulas (such as minimizing the loss function). Therefore this claim equates to an abstract idea of mathematical concepts. Hence the claims individually and in combination, recites elements that classified as abstract ideas in the form of mental processes and mathematical concepts. An abstract idea performed in a generic computer environment does not change the fact the steps are directed to mathematical concepts. The claims must therefore be examined further to determine whether the claims integrate the above-identified abstract ideas into a practical application (MPEP 2106.04(d)). Step 2A Prong Two: Consideration of Practical Application The claims result in a process “to generate one or more inferences corresponding to the ATAC-seq information”, which reads on generating new information from existing data. The claims do not recite any additional elements that integrate the abstract idea/judicial exception into a practical application. This judicial exception is not integrated into a practical application because the claims do not meet any of the following criteria: An additional element reflects an improvement in the functioning of a computer, or an improvement to other technology or technical field; an additional element that applies or uses a judicial exception to effect a particular treatment or prophylaxis for a disease or medical condition; an additional element implements a judicial exception with, or uses a judicial exception in conjunction with, a particular machine or manufacture that is integral to the claim; an additional element effects a transformation or reduction of a particular article to a different state or thing; and an additional element applies or uses the judicial exception in some other meaningful way beyond generally linking the use of the judicial exception to a particular technological environment, such that the claim as a whole is more than a drafting effort designed to monopolize the exception. Step 2B: Consideration of Additional Elements and Significantly More The claimed method also recites "additional elements" that are not limitations drawn to an abstract idea. The recited additional elements are drawn to: Claims 1, 10 and 17 recites: A processor comprising: One or more circuits to provide to one or more computer-implemented neural networks a plurality of ATAC-seq counts at each base pair position of a genome that exclude base identity, and to use the one or more computer-implemented neural networks to generate one or more inferences corresponding to the ATAC-seq information based, at least in part, on the plurality of ATAC-seq counts. Claim 2 recites: wherein the neural network comprises at least three successive convolutional layers followed by a residual connection that sums the ATAC-seq counts at each base pair position input with the final convolutional layer output. Claim 3 recites: wherein each convolutional layer and the residual connection is followed by an ReLU layer. Claim 4 recites : wherein the convolutional layers have a receptive field size of 100 to 10,000. Claim 5 recites: wherein the convolutional layers use a number of filters selected from 1 to 100. Claim 6 recites: wherein the neural network comprises 3 successive convolutional layers of receptive field size 100, 100, and 300, respectively, and using 15, 15, and 1 filters, respectively. Claim 10 recites: A system comprising: a processor and a memory device, Claim 11 recites: The system of claim 10, wherein the neural network comprises at least three successive convolutional layers followed by a residual connection that sums the ATAC-seq counts of each segment input with the final convolutional layer output. Claim 12 recites: The system of claim 11, wherein each convolutional layer and the residual connection is followed by an ReLU layer. Claim 13 recites: The system of claim 11, wherein the convolutional layers have a receptive field size of 100 to 10,000. Claim 14 recites: The system of claim 11, wherein the convolutional layers use a number of filters selected from 1 to 100. Claim 15 recites: The system of claim 11, wherein the neural network comprises 3 successive convolutional layers of receptive field size 100, 100, and 300, respectively, and using 15, 15, and 1 filters, respectively. Claim 17 recites: A non-transitory computer-readable medium. The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception because these additional elements can be summarized as: 1) data input/output (claims 1, 9, 17); 2) software product (“a non-transitory computer-readable medium”, claim 17); and 3) technological environment (“A system comprising: a processor and a memory device”, “one or more neural networks”, “convolutional layer” and “ReLU layer”; claims 1-6, 10-15 ). Group 1) are insignificant extra-solution activities because they are necessary for data analysis; group 2) and 3) are merely providing technical environment to execute the abstract idea. The claims do not include additional elements that are sufficient to amount of significantly more than the judicial exception because it is routine and conventional to perform the acts of modeling nucleic acid data using neural networks. Other elements of the method include computing components which are recitation of generic computer structures that serves to perform generic computer functions that are well-understood, routine, and conventional activities previously known to the pertinent industry. Additionally, the conventionality of the additional elements like “convolutional neural networks” and a generic computer system can further be evidenced in the following references: Koh (“Denoising genome-wide histone ChIP-seq with convolutional neural network” Bioinformatics, 33, 2017, i225–i233). Cited on the 4/11/2023 IDS. Hiranuma (“DeepATAC: A deep-learning method to predict regulatory factor binding activity from ATAC-seq signals”, bioRxiv preprint doi: https://doi.org/ 10.1101/172767; this version posted August 6, 2017). Cited on the 4/11/2023 IDS. Thibodeau, A., Uyar, A., Khetan, S. et al. A neural network based model effectively predicts enhancers from clinical ATAC-seq samples. Sci Rep 8, 16048 (2018). Cited on the 4/11/2023 IDS. Koh discloses a method (Coda) to denoise the ChIP-seq data with convolution neural network (CNN) (page i226, col 2, para 1-3, and Fig 2, page i226; page i227, col 1, para 1). However, the datasets of Koh are ChIP-seq data, not ATAC-seq data. These two data, ChIP-seq data and ATAC-seq data, are both about chromatin profiling. The ways genomic fragments captured are different but the final output are both NGS sequence reads. ChIP-seq is for detecting regulatory regions (such as transcription factor binding sites) in the genome, while ATAC-seq is slightly more general, about any open part in the genome. Koh points out that the method can be used for the ATAC-seq data (“We believe that a similar approach can be used in other biological assays, e.g. ATAC-seq and DNase-seq, where it is near impossible to analytically characterize all types of technical noise or the overall data distribution but possible to generate noisy versions of high-quality samples through experimental or computational perturbation. This can significantly reduce cost while maintaining or even improving quality, especially in high-throughput settings or when dealing with limited amounts of input material (page i230, col 2, last para through pg. i231, col 1, 1st para). Similar to Koh., Hiranuma discloses a method (DeepATAC) to predict the transcription factors binding activities from the ATAC-seq datasets, using the CNN technology. Hiranuma’s method deals with the ATAC-seq, not the ChIP-seq. but Hiranuma’s method takes the base identity into consideration by encoding the DNA sequence samples using the 1 hot-encoded 200 base pair windows (equivalent to a segment. Page 1, Section “Abstract” and page 4, para 1-2 under section “Method”). Additionally, Hiranuma points out that the ChIP-seq can be used to get high-quality subset of ATAC-seq data (“The DeepATAC model was trained on the subset of DeepSEA data. Specifically, the DNA sequence samples were 1 hot-encoded 200 base pair windows (4 by 200 matrix), where we observed (using ChIP-seq peaks) at least one of the 163 and 91 regulatory factors for the K562 and GM12878 cell types respectively. (page 4, lines 1-4 under section “4 Methods”). Hence, just to look at the data handed by CNN models, ChIP-seq looks exactly like ATAC-seq. therefore, the subject matter explored in instant claims had been explored by at least two groups. Thibodeau taught a neural network based model to predict enhancers from clinical ATAC-seq samples. Although not using the convolutional neural network model, Thibodeau’s neural network has several layers (“multiple-layer perceptron neural network”). Viewed as a whole, the additional claim elements do not provide meaningful limitation(s) to transform the abstract idea recited in the instantly presented claims into a patent eligible application of the abstract idea such that the claim(s) amounts to significantly more than the abstract idea itself. Therefore, the claim(s) are rejected under 35 U.S.C. 101 as being directed to non-statutory subject matter. Response to Applicant’s Argument Applicant's arguments filed 27 May 2026 have been fully considered but they are not persuasive. Particularly: In the Remarks filed 27 May 2026, Applicant argued (page 6, paras 1-2) that the amendment “The amendment clarifies that the neural networks are used in the practical application of an improved computer-implemented method for denoising ATAC-seq experimental information”; “The process thus requires less computer storage and retrieval of data in carrying out the denoising ATAC-seq data”; and “when considering the amended claims as a whole, any judicial exception recited in the method using neural networks integrates the exception into the practical application of improved ATAC-seq experimental data analysis”. Applicant’s argument refers to Step 2A/Prong Two in the 101 analysis, regarding whether claims are integrated into a practical application or not. In response, Applicant’s argument is not persuasive. To integrated claims into a practical application at Step 2A/Prong Two, additional elements that apply, capture and reflect the judicial exceptions recited in the method using neural networks are needed. However, we cannot identify such additional elements. The recited generic computer and computing components are not good enough. It is important to note that in order for a method claim to improve computer functionality, the broadest reasonable interpretation of the claim must be limited to computer implementation. That is, a claim whose entire scope can be performed mentally, cannot be said to improve computer technology. Synopsys, Inc. v. Mentor Graphics Corp., 839 F.3d 1138, 120 USPQ2d 1473 (Fed. Cir. 2016) (a method of translating a logic circuit into a hardware component description of a logic circuit was found to be ineligible because the method did not employ a computer and a skilled artisan could perform all the steps mentally). Similarly, a claimed process covering embodiments that can be performed on a computer, as well as embodiments that can be practiced verbally or with a telephone, cannot improve computer technology. See RecogniCorp, LLC v. Nintendo Co., 855 F.3d 1322, 1328, 122 USPQ2d 1377, 1381 (Fed. Cir. 2017) (process for encoding/decoding facial data using image codes assigned to particular facial features held ineligible because the process did not require a computer). (MPEP 2106.05(a).I). As discussed above over the 101 analysis, the claim 1 recites “to generate one or more inferences corresponding to the denoised ATAC-seq information” in a general, simple way. This limitation, as drafted, is a process that, under its broadest reasonable interpretation (BRI), covers performance of the limitation in the mind but for the recitation of generic computing steps. That is, other than reciting “one or more circuits to provide to one or more computer-implemented neural networks,” nothing in the claim element precludes the step from practically being performed in the mind. For example, but for the “one or more circuits to provide to one or more computer-implemented neural networks” language, the claim encompasses a user simply reading the ATAC-seq multiple alignment data and determining inferences in his/her mind. The mere nominal recitation of one or more generic processors/circuits/neural networks do not take the claim limitation out of the mental processes grouping. Thus, the claims recite a mental process. Therefore, claim 1 (and so are claims 10 and 17) is not purely performed in computers. The argument of improving computer functionating is not persuasive. In the Remarks, Applicant argues (page 6, 3rd para) that “this advantageous method for denoising ATAC-seq data is not described or suggested in the prior art, and therefore is significant extra solution activity. Thus, as prescribed by MPEP 2106.05(g), the correct answer under Step 2B should be ‘Yes’". In response, Applicant arguments are not persuasive. Denoising ATAC-seq data is not significant extra solution activity. Denoising ATAC-seq data is a mental activity can be achieved in the human mind, therefore part of the JEs. 35 USC 101 and 35 USC 102 or 103 are different statutes. They are examined separately. Art-free does not mean significantly more. At Step 2B additional elements are what concerned. Generic computer structures: neural network, processor, memory, individual or as a whole, the claims do not recite any non-routine additional elements, under Step 2B. The conventionality of recited additional elements are further evidenced by the reference of Koh (“Denoising genome-wide histone ChIP-seq with convolutional neural network” Bioinformatics, 33, 2017, i225–i233. Cited on the 4/11/2023 IDS), and Hiranuma (“DeepATAC: A deep-learning method to predict regulatory factor binding activity from ATAC-seq signals”, bioRxiv preprint doi: https://doi.org/10.1101/172767; this version posted August 6, 2017. Cited on the 4/11/2023 IDS)). Both Koh and Hiranuma have used computer (including processors and memory) and neural networks to model epigenetics datasets that are output from the NGS facilities in the same format of aligned short reads. Hence, the 101 rejection is maintained. Double Patenting This rejection is newly installed. Necessitated by claim amendment. 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. Claims 1, 10 and 17 are rejected on the ground of nonstatutory double patenting as being unpatentable over claims 1, 11 and 19 of U.S. Patent No. 11,657,897 B2 (US16/236,797), titled “Denoising ATAC-seq data with deep learning,” same priority date 2018-12-31 and same assignee/current assignee.. Although the claims at issue are not identical, they are not patentably distinct from each other because: The reference parent independent claim 1 recites: “1. A computer-implemented method for denoising ATAC-seq data comprising: partitioning an ATAC-seq dataset into a plurality of segments representing a genome, wherein the segments comprise ATAC-seq counts at each base pair position and exclude base identity at each base pair position; and using one or more neural networks to identify the ATAC-seq dataset based, at least in part, on the plurality of segments.” The reference parent system claim 11 recites: “11. A system for denoising ATAC-seq data comprising: a processor, a memory device, and a denoising engine executable on the processor according to software instructions stored in the memory device, wherein the denoising engine is configured to: partition an ATAC-seq dataset into a plurality of segments representing a genome, wherein the segments comprise ATAC-seq counts at each base pair position and exclude base identity at each base pair position; and using one or more neural networks to identify the ATAC-seq dataset based, at least in part, on the plurality of segments.” The reference parent CRM claim 19 recites: “19. A non-transitory computer-readable medium for denoising ATAC-seq data comprising instructions for denoising an ATAC-seq dataset that, when executed by a processor, cause the processor to perform one or more steps comprising: partitioning an ATAC-seq dataset into a plurality of segments representing a genome, wherein the segments comprise ATAC-seq counts at each base pair position and exclude base identity at each base pair position; and using one or more neural networks to identify the ATAC-seq dataset based, at least in part, on the plurality of segments.” Analysis: the pending amended processor/system/CRM claims appear patentably indistinct from parent claims directed to denoising ATAC-seq data using neural networks on ATAC-seq counts at each base-pair position excluding base identity. Differences such as phrasing as processor circuitry, system/memory, or non-transitory medium, and adding “generate one or more inferences corresponding to the denoised ATAC-seq information,” appear to be obvious variants of the same denoising engine/method disclosed and claimed in the parent, absent a terminal disclaimer or a persuasive showing of patentable distinctness. A provisional OTDP rejection is appropriate, subject to checking ownership and terminal-disclaimer status in the live file wrapper. Claim Rejections - 35 USC§ 103 This rejection is maintained from the 1/28/2026 Office Action. Modifications are necessitated by claim amendment filed 5/27/2026. 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 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 text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action. 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. Claims 1-8 and 10-18 are rejected under 35 U.S.C. 103 as being unpatentable over Koh (“Denoising genome-wide histone ChIP-seq with convolutional neural network” Bioinformatics, 33, 2017, i225–i233. Cited on the 4/11/2023 IDS), in view of Hiranuma (“DeepATAC: A deep-learning method to predict regulatory factor binding activity from ATAC-seq signals”, bioRxiv preprint doi: https://doi.org/10.1101/172767; this version posted August 6, 2017. Cited on the 4/11/2023 IDS). Claim 1 is directed to a method of inferring information out of the ATAC-seq dataset wherein the specific model input and output are not specified. “A plurality of ATAC-seq counts at each base pair position of a genome that exclude base identity” is interpreted as a further limit to “ATAC-seq information” that before denoising, and what exactly used as input for the “one or more computer-implemented neural networks” is not claimed. With respect to claim 1, Koh discloses a method (Coda) to denoise the ChIP-seq data with convolution neural network (CNN). The specific base identify in ChIP-seq data is excluded from downstream analysis, the peak signals are the information that matters. (page i226, col 2, paras 1-3 and Fig 2; page i227, col 1, para 1). Koh teaches using the CNN to denoise each bins of the ChIP-seq datasets. However, the dataset of Koh is the ChIP-seq data, not the ATAC-seq data. These two datasets, ChIP-seq data and ATAC-seq data, are similar with respect to chromatin profiling. The ways genomic fragments captured are different but the final output are both NGS sequence reads. ChIP-seq is for detecting regulatory regions (such as transcription factor binding sites) in the genome, while ATAC-seq is slightly more general, about any open part in the genome. Koh points out that the method can be used for the ATAC-seq data (“We believe that a similar approach can be used in other biological assays, e.g. ATAC-seq and DNase-seq, where it is near impossible to analytically characterize all types of technical noise or the overall data distribution but possible to generate noisy versions of high-quality samples through experimental or computational perturbation. This can significantly reduce cost while maintaining or even improving quality, especially in high-throughput settings or when dealing with limited amounts of input material (page i230, col 2, last para through page i231, col 1, 1st para). Further, Koh does not teach that his ChIP-seq comprising “a plurality of ATAC-seq counts at each base pair position of a genome” explicitly. However, this is an inherent part of ChIP-seq and ATAC-seq, because both ChIP-seq and ATAC-seq are generated by next generation sequencing (NGS), followed by peak (or count) calling. Koh bins every 25 bp and get the average signal (per base) as the peak calling for the bin (Koh, page i226, col 2, 3rd para lines 1-2: “For computational efficiency, we first bin the genome into 25 bp bins, averaging the signal in each bin”). Hiranuma also teaches that per base count is the inherent part of the ATAC-seq (Hiranuma, page 4, Fig. 3: “1 hot encoded DNA sequence”). By using bins of 25 bp, Koh essentially teaches “exclude base identity” when denoising. Similar to Koh, Hiranuma teaches a method (DeepATAC) to predict the transcription factors binding activities from the ATAC-seq datasets, using the CNN technology. Hiranuma’s method deals with the ATAC-seq, not the ChIP-seq. Hiranuma’s method takes the base identity into consideration by encoding the DNA sequence samples using the 1 hot-encoded 200 base pair windows (equivalent to a segment. Page 1, Section “Abstract” and page 4, paras 1-2 in “Method”). Since the basic processing unit is 25bp (per bin), Koh teaches “exclude base identity”. Additionally, Hiranuma points out that the ChIP-seq can be used to get high-quality subset of ATAC-seq data (“The DeepATAC model was trained on the subset of DeepSEA data. Specifically, the DNA sequence samples were 1 hot-encoded 200 base pair windows (4 by 200 matrix), where we observed (using ChIP-seq peaks) at least one of the 163 and 91 regulatory factors for the K562 and GM12878 cell types respectively (page 4, line 1-4 under section “4 Methods”). Regarding claim 2, Koh discloses a CNN configuration with two convolutional layers (Fig 2, page i226). Koh do have an output layer that omits the signal (reads on the sequence counts of each segment input in the claim limitation) for each bin in a window (page i227, col 1, para 2). Hiranuma teaches a CNN with three convolutional layers (page 4, Fig 3). Regarding claim 3, Koh discloses “It does this by feeding the noisy data through a first convolutional layer, a rectified linear unit (ReLU) layer, a second convolutional layer, and then a final ReLU or sigmoid layer (for regression or classification, respectively)” (Page i226, col 2, 3rd para in “2 Materials and methods/2.1 Model”, lines 6--9), which suggests the claim limitation “wherein each convolutional layer and the residual connection is followed by an ReLU layer”, as Koh’s convolutional layers store the intensity counts for each bin. Regarding claim 4, Hiranuma’s teaches a CNN with three convolutional layers that each has a receptive field size of 160 kernels, 240 kernels, and 480 kernels respectively (page 4, Fig 3). Hiranuma anticipate the claimed range in receptive field size (MPEP § 2131.03 I. “A SPECIFIC EXAMPLE IN THE PRIOR ART WHICH IS WITHIN A CLAIMED RANGE ANTICIPATES THE RANGE”). Regarding claim 5, Koh discloses “For the first convolutional layer, we use 6 convolutional filters, each 51 bins in length (25bp/bin); for the second convolutional layer, we use a single filter of length 1001” (Page i226, col 2, 3rd para in “2 Materials and methods/2.1 Model”), which suggests on the claim limitation “wherein the convolutional layers use a number of filters selected from 1 to 100”, as 6 and 1 are both anticipate the range of 1 to 100 (MPEP § 2131.03 I. “A SPECIFIC EXAMPLE IN THE PRIOR ART WHICH IS WITHIN A CLAIMED RANGE ANTICIPATES THE RANGE”) in the number of filters. Regarding claim 6, Koh discloses a CNN of two convolutional layers that has 6 filter and 1 filter respectively (“For the first convolutional layer, we use 6 convolutional filters, each 51 bins in length (25bp/bin); for the second convolutional layer, we use a single filter of length 1001” (page i226, col 2, 3rd para in “2 Materials and methods/2.1 Model”, lines 6-11), which suggests the claim limitation “using 15, 15, and 1 filters, respectively”. Koh is silent in the receptive field size. Hiranuma’s teaches a CNN with three convolutional layers that each has a receptive field size of 160 kernels, 240 kernels, and 480 kernels respectively. (page 4, Fig 3). Hiranuma anticipate the claimed range in receptive field size (MPEP § 2131.03 II). Regarding claim 7, Koh further discloses a method (Coda) to denoise the ChIP-seq data with convolution neural network (CNN). More specifically to use the CNN to map between “suboptimal” and “high quality” ChIP-seq data (“we introduce a convolutional denoising algorithm, called Coda, that uses convolutional neural networks (CNNs) to learn a generalizable mapping between ‘suboptimal’ and high quality ChIP-seq data (page i226, col 1, Fig. 1). Coda substantially attenuates three primary sources of noise—due to low sequencing depth, low cell input and low ChIP enrichment—enhancing signal in low-quality samples across individuals, cell types and species”. Page i225, col 2, para 2, lines 1-8). Further, Koh use a loss function to evaluate training progress (“We used the Keras package for training and AdaGrad as the optimizer, stopping training if validation loss did not improve for three consecutive epochs”. Page i227, col 2, 1st para lines 2-4). Regarding claim 8, Koh teaches the MSE function (“For the regression tasks (predicting signal), we evaluated performance by computing the Pearson correlation and mean-squared error (MSE) between the predicted and measured high-quality fold enrichment signal profiles after an inverse hyperbolic sine transformation, which reduced the dominance of outliers. We compared this with the baseline performance obtained by directly comparing the noisy and high-quality signal profiles of the target mark (after the same inverse hyperbolic sine transformation)”. Page i227, col 1, para 2 in section “2.2 Training and evaluation”) Claims 10-16 are the computer “system” version for the method version of claims 1-8. Since Koh also suggest a computer system (page i227, col 1, 2nd para), the art applied to claims 1-9 also teach claims 10-16. Claims 17-18 are the software disk version for the method version of claims 1-8. Since Koh also suggest a computer system (page i227, col 1, 2nd para) which suggests a storage media, the art applied to claims 1-8 also teach claims 17-18. An invention would have been obvious to one of ordinary skill in the art if some motivation would have led that person to modify prior art teachings to arrive at the claimed invention. Prior to the time of invention, said person would have been motivated to substitute the ChIP-seq datasets, handed by Koh’s CNN (for denoising the ChIP-seq data), with Hiranuma’s ATAC-seq datasets, to achieve a modified pipeline using CNN for denoising the ATAC-seq datasets. Because ATAC-seq is the most recent method and is rapidly gaining popularity due to its cost-efficiency and simplicity (Hiranuma: page 1, last para). The said person would have been motivated to modify Koh’s CNN to preserve Koh’s features of data partitioning and base-pair identity-free data encoding, and to introduce Hiranuma’s features of elegant three-layered CNN configuration plus the receptive field size setups, Because such a modification and recombination will have the best parts from both Koh and Hiranuma. We can reasonably expect success for the modification and recombination, as Koh’s ChIP-seq and Hiranuma’s ATAC-seq are both genome-wide chromatin accessibility assays. The raw ChIP-seq and the ATAC-seq data are both NGS sequence reads. They can both handed by the MACS2 program for peak calling, they look the same and processed by the two CNN frameworks of Koh and Hiranuma, Koh points out that his CNN can be used for Hiranuma’s ATAC-seq data (page i230, col 2, last para through page i231, col 1, 1st para) and Hiranuma points out that his CNN can hand Koh’s ChIP-seq to get high-quality subset of ATAC-seq data (page 4, line 1-4 under section “4 Methods”). Response to Applicant’s Arguments In the Remarks filed 5/27/2026, Applicant argued (page 7, 2nd para through page 8, 2nd para) that “independent claims as currently amended: ‘wherein the segments comprise ATAC-seq counts at each base pair position and exclude base identity at each base pair position’" is not taught by the applied art of Koh or Hiranuma. In response, Applicant argument is not persuasive. First, the argued “wherein the segments comprise ATAC-seq counts at each base pair position and exclude base identity at each base pair position” is not recited in the claims. “Segments” is a misleading term here. It triggers people to think the neural network model input which is not claimed. “Segments” does not exist in the claim language. Second, “a plurality of ATAC-seq counts at each base pair position of a genome that exclude base identity” is interpreted as a further limit to “ATAC-seq information” that before denoising, and what exactly used as input for the “one or more computer-implemented neural networks” is not claimed. With respect to claim 1, Koh discloses a method (Coda) to denoise the ChIP-seq data with convolution neural network (CNN). The specific base identify in ChIP-seq data is excluded from downstream analysis, the peak signals are the information that matters. (page i226, col 2, paras 1-3 and Fig 2; page i227, col 1, para 1). Koh teaches using the CNN to denoise each bins of the ChIP-seq datasets. Further, Koh does not teach that his ChIP-seq comprising “a plurality of ATAC-seq counts at each base pair position of a genome” explicitly. However, this is an inherent part of ChIP-seq and ATAC-seq, because both ChIP-seq and ATAC-seq are generated by next generation sequencing (NGS), followed by peak (or count) calling. Koh bins every 25 bp and get the average signal (per base) as the peak calling for the bin (Koh, page i226, col 2, 3rd para lines 1-2: “For computational efficiency, we first bin the genome into 25 bp bins, averaging the signal in each bin”). Hiranuma also teaches that per base count is the inherent part of the ATAC-seq (Hiranuma, page 4, Fig. 3: “1 hot encoded DNA sequence”). By using bins of 25 bp, Koh essentially teaches “exclude base identity” when denoising. Therefore, combined Koh and Hiranuma do teach “one or more computer-implemented neural networks to denoise ATAC-seq information comprising a plurality of ATAC-seq counts at each base pair position of a genome that exclude base identity, and to use the one or more computer-implemented neural networks to generate one or more inferences corresponding to the denoised ATAC-seq information based, at least in part, on the plurality of ATAC-seq counts” as claimed in claims 1, 10 and 17. Hence, the 103 rejection is maintained. Conclusion No claims are 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 GUOZHEN LIU whose telephone number is (571)272-0224. The examiner can normally be reached Monday-Friday 8-5. 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, Larry D Riggs can be reached at (571) 270-3062. 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. /GL/ Patent Examiner Art Unit 1686 /Anna Skibinsky/ Primary Examiner, AU 1635