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 summary of the Interview, March 12, 2026, is acknowledged.
The Abstract, filed April 1, 2026, has been accepted.
REMARKS
On pages 12-17, Applicant argues claims 1-15 ("the claims") are not directed to an abstract idea, have a practical application, and contain a number of particular limitations which ensure that no underlying idea would be tied up. Additionally, it is respectfully submitted that the claims recite features, which alone and in combination, provide improvements to the technological field of learning consistent representations of multiple data modalities, in particular biological entities, such as proteins and diseases, based on a knowledge graph (KG) and any additional data modality such as structured annotations and free text describing the entities using neural network systems. See Present Published Specification, paragraphs [0017]-[0020]. The claims considered as a whole recite limitations that under BRI (Broadest Reasonable Interpretation) are directed to a mental process and mathematical calculations for characterizing a natural phenomenal such as relationships among chemicals, proteins, and diseases.
Specifically on pages 12-15, Applicant argues the claims are not directed to a mental process because the claims are directed to training neural network systems to learn consistent representations of multiple data modalities. Applicant’s argument is not persuasive because the claims in view of the specification are directed to a mental process. For example, in paragraph [0009] the specification discloses the present disclosure provides a computer-implemented method of learning novel relationships among various entities, including biological entities such as chemicals, proteins, and diseases….” At high level, the claimed invention is merely directed to learning about existing relationships among biological entities. Claims do recite a mental process when they contain limitations that can practically be performed in the human mind, including for example, observations, evaluations, judgments, and opinions. Similarly, claims for "comparing BRCA sequences and determining the existence of alterations," where the claims cover any way of comparing BRCA sequences such that the comparison steps can practically be performed in the human mind, University of Utah Research Foundation v. Ambry Genetics, 774 F.3d 755, 763, 113 USPQ2d 1241, 1246 (Fed. Cir. 2014) See MPEP 2106.04(a)(2) Abstract Idea Groupings [R-07.2022] Further, claims wherein merely applying a training step using mathematical calculations (see instant specification paragraphs, [0041] to [0049]) to learn relationships among various biological entities do not preclude the interpretation of a mental process as guided by the MPEP 2106.04(a)(2) Abstract Idea Groupings [R-07.2022].
On pages 14-15, Applicant argues at best, these features merely involve an exception (e.g., a mental process) instead of actually reciting or being directed to the exception. This is further supported by Example 39. For example, the claim from Example 39 describes "collecting a set of digital facial images from a database." But, as described by Example 39, the claim of Example 39, including collecting a set of digital facial images, does not recite any judicial exceptions including a mental process because the steps are not practically performed in the human mind. Applicant’s argument is not persuasive because the only similarity between the instant claims and the “Example 39” is the phrase of “training the neural network” in “Example 39” and “training a neural network” of the instant claims. However, unlike the instant claims, the steps of “Example 39” are not practically performed in the human mind.
On page 15, Applicant argues the limitation does not set forth or describe any mathematical relationships, calculations, formulas, or equations using words or mathematical symbols and thus, the claim does not recite a judicial exception." USPTO Memorandum, p. 2. Accordingly, as described above, given that the claims recite features that cannot be performed in the human mind and an almost an identical feature to Example 39 of training a neural network, it is respectfully submitted that the claims are not directed to a mental process and/or a mathematical concept. Applicant’s argument is not persuasive. It is noted that a claim does not have to recite the word "calculating" in order to be considered a mathematical calculation. However, a claim that recites a mathematical calculation, when the claim is given its broadest reasonable interpretation in light of the specification, will be considered as falling within the "mathematical concepts" grouping. The instant claims merely recite applying a training step using mathematical calculations as described by the instant specification (paragraphs [0041] to [0049]) to learn relationships among various biological entities. Therefore, when the claim is given its broadest reasonable interpretation in light of the specification the claim is directed to a mathematical calculation.
On pages 15-17, Applicant argues the claims provide a practical application under Prong Two of Step 2A of the Subject Matter Eligibility Test. As set forth in MPEP § 2106, an element or combination of elements that reflect a technical improvement (e.g., to a computer or any other technology or technical field) provides that a claim is directed to a practical application. See MPEP § 2106.0S(a). In this regard, the claims are directed to improvements to the technological field of training neural network systems to learn consistent representations of multiple data modalities including a KG and additional data modalities such as documents and/or images. For example, MPEP § 2106.0S(a) explicitly states that "[a]n indication that the claimed invention provides an improvement can include a discussion in the specification that identifies a technical problem and explains the details of an unconventional technical solution expressed in the claim, or identifies technical improvements realized by the claim over the prior art." Applicant points to paragraphs [0017], [0019], and [0020] to support the asserted technical improvement(s). Applicant’s argument is not persuasive because the cited disclosure merely uses a training network to characterize the relationships of a natural phenomenal among the molecules, diseases, and other entities involved in the biological processes, however, the claimed invention does not improve computers, other technology or technical field. It is important to keep in mind that an improvement in the abstract idea itself (e.g. a recited fundamental economic concept) is not an improvement in technology.
Applicant asserts the claims recite that the neural network system treats the knowledge graph and the objects of a respective one of the data modalities in a unified manner by jointly learning embeddings of the nodes from the knowledge graph and embeddings of the objects of the respective one of the data modalities. The Office Action fails to consider the additional features of the claims as a whole and in fact does not even provide analysis on the additional features at all. Applicant’s argument is not persuasive the limitation of “training a neural network…” was addressed a step accomplished by mathematical calculations as recited in claim 4. Further, the instant specification (see instant specification paragraphs, [0041] to [0049]) describes the training a neural network by using mathematical calculations to learn relationships among various biological entities.
On pages 17-18, Applicant argues the claims should at least be found to be patent-eligible under Step 2B as the claims recite significantly more than a mental process and/or a mathematical concept. The claims recite an inventive concept sufficient to transform the alleged abstract idea into a patent-eligible invention. Applicant’s argument is not persuasive because the claims does not specifically recite limitation(s) that could reasonably be interpretation as transforming.
On page 18, Applicant argues claims 16-20 are not directed to an abstract idea, have a practical application, and contain a number of particular limitations which ensure that no underlying idea would be tied up. For instance, regarding Step 2A, Prong One, to reject the features of dependent claim 16, the Office states that "[t]hese limitations further narrow the abstract idea or extra-solution activity, but are nonetheless part of the abstract idea identified in claim 1." See Detailed Action, p. 14. Dependent claim 16 recites "generating the embeddings of the nodes of the knowledge graph using the first machine learning model; generating the embeddings of the objects of the respective one of the data modalities using the second machine learning model; jointly training the first machine learning model and the second machine learning model based on the embeddings of the nodes of the knowledge graph and the embeddings of the objects of the respective one of the data modalities." Dependent claims 17- 20 provide further details on how the embeddings are generated and the joint training is performed. Applicant’s argument is not persuasive because the claims considered as a whole recite limitations that under BRI (Broadest Reasonable Interpretation) are directed to a mental process and mathematical calculations for characterizing a natural phenomenal such as relationships among chemicals, proteins, and diseases.
On pages 18-19, Applicant argues Example 39 explicitly contradicts the Office's analysis that generating embeddings using the machine learning models and training or "jointly training" machine learning models are directed to a mathematical calculation and/or a mental process. For example, Example 39 explicitly states that claim features directed to generating training data (e.g., creating a first training set of collected set of digital facial images) and then training a machine learning model using the training data would not recite any judicial exceptions including mental processes or mathematical concepts. See Example 39. Similarly, dependent claim 16 also describes generating training data (e.g., generating embeddings), but moves even a step further than the claim in Example 39 by stating that the training data is generated not merely based on rotating digital facial images, but by using the machine learning models themselves. Further, dependent claim 16 also describes jointly training the machine learning models using the generated embeddings that are generated by the machine learning models (e.g., the embeddings of the nodes of the knowledge graph that are generated using the first machine learning model and the embeddings of the objects of the respective one of the data modalities that are generated using the second machine learning model). Thus, as explicitly stated by Example 39 and further supported by the USPTO Memorandum, the features of dependent claim 16 are not directed to a mental process or a mathematical concept. Applicant’s argument is not persuasive because the only similarity between the instant claims and “Example 39” is the recitation of “training.” However, when the claims considered as a whole recite limitations that under BRI (Broadest Reasonable Interpretation) are directed to a mental process and mathematical calculations for characterizing a natural phenomenal such as relationships among chemicals, proteins, and diseases.
On pages 19-20, Applicant argues “[r]egarding Step 2A, Prong Two, it is respectfully submitted that the new claims, including dependent claim 16, provide a practical application as the features from the new claims provide a technical improvement (e.g., to a computer or any other technology or technical field). See MPEP § 2106.05(a). For example, the Present Published Specification describes that an "obvious approach to combine the [knowledge graph (KG)] structure and documents (which is present in the literature) is to train independent ML models on the KG and documents and then combine them in some post hoc manner." See Present Published Specification, paragraph [0019]. But, "even when trained jointly and end-to-end (as done in previous work), these two embeddings are still likely to be completely different for each entity. That is, the embeddings are not unified, but, rather, completely independent." See Present Published Specification, paragraph [0019]. Applicant further points to Published Specification, paragraphs [0017], [0020], and [0054]-[0083] to support the technical problem being solved by the claimed invention. Applicant’s argument is not persuasive because the claims do not reflect the disclosure and asserted improvement. Currently, the claims are directed to high level steps that under BRI the limitations embody the mental process and/or mathematical calculations. The claims do not include the components or steps of the invention that provide the improvement as asserted by Applicant. Further, the claims considered as a whole recite limitations that under BRI (Broadest Reasonable Interpretation) are directed to a mental process and mathematical calculations for characterizing a natural phenomenal such as relationships among chemicals, proteins, and diseases. The cited disclosure merely uses a training network to characterize the relationships of a natural phenomenal among the molecules, diseases, and other entities involved in the biological processes, however, the claimed invention does not improve computers, other technology or technical field. It is important to keep in mind that an improvement in the abstract idea itself (e.g. a recited fundamental economic concept) is not an improvement in technology.
On page 20-21, Applicant argues dependent claim 16 integrates the alleged judicial exception into a practical application. Furthermore, because dependent claims 17-20 provide additional details on how the embeddings are generated and/or how the joint training of the machine learning models are performed, it is respectfully submitted that dependent claims 17-20 further integrate the alleged judicial exception into a practical application. Applicant’s argument is not persuasive the claims do not recite additional elements beyond high-level of generality such that they improve computers, other technology or technical field. It is important to keep in mind that an improvement in the abstract idea itself (e.g. a recited fundamental economic concept) is not an improvement in technology.
On page 21, Applicant argues “Turning to Step 2B, Applicant acknowledges that the dependent claims 16-20 are free of prior art. See Detailed Action, p. 26. Thus, it is respectfully submitted that dependent claims 16-20 recite an inventive concept sufficient to transform the alleged abstract idea into a patent- eligible invention. Accordingly, because the claims 16-20 are novel and non-obvious and recite a combination of features which go beyond what is well-understood, routine, or conventional, the claims 16-20 recite an inventive concept, which ensures that the claim, as a whole, amounts to significantly more than the alleged abstract idea.” Applicant’s argument is not persuasive because claims 16-20 further narrow the abstract idea or extra-solution activity, however, they are nonetheless part of the abstract idea identified in claim 1. They also do not amount to significantly more than the abstract idea. The claims are similarly rejected under the same rationale as claim 1, supra.
On pages 22-24, Applicant points to paragraphs [0017] and [0020] to support the claimed invention as described by the pointed to disclosure is distinct from the cited prior art. Further, Applicant argues the claims describe the crucial concept of integrating and jointly learning embeddings from a knowledge graph and one or more external data modalities (e.g., biomedical documents, images) to discover novel relationships. Thus, the above-identified feature of the claims describes leveraging information beyond just the symbolic structure of the knowledge graph by incorporating external, heterogeneous data (e.g., the embeddings of the objects of the respective one of the data modalities). See Present Published Specification, paragraphs [0003] and [0017]. Applicant’s argument is not persuasive because the claims are given their broadest reasonable interpretation consistent with the specification. However, the instant claims are not limited to the critical steps that have been cited from the specification by Applicant as limitations that are not disclosed by the prior art. As cited by the MPEP, the court explained that “reading a claim in light of the specification, to thereby interpret limitations explicitly recited in the claim, is a quite different thing from reading limitations of the specification into a claim,’ to thereby narrow the scope of the claim by implicitly adding disclosed limitations which have no express basis in the claim.” The court found that applicant was advocating the latter, i.e., the impermissible importation of subject matter from the specification into the claim.). See also In re Morris, 127 F.3d 1048, 1054-55, 44 USPQ2d 1023, 1027-28 (Fed. Cir. 1997) (MPEP §2111 [R-1]).
As for Applicant’s argument, Costabello merely describes knowledge graph embeddings (i.e., the three knowledge graph embeddings indicating a point representing SARS, a point represent that SARS has a treatment, and a point representing the SARS vaccine). Costabello fails to disclose or suggest the other type of embeddings (e.g., embeddings of the objects of the respective one of the data modalities). Accordingly, because Costabello fails to disclose or suggest both embeddings of the nodes from the knowledge graph and embeddings of the objects of the respective one of the data modalities, then Costabello fails to disclose or suggest at least the above-identified features of the independent claims. Applicant’s argument is not persuasive. As previously cited, Costabello discloses the fit/train engine may convert entities (e.g., nodes) and relations (e.g., links or edges) of the knowledge graph into points in a k-dimensional metric space. For example, as shown in FIG. 1C, the knowledge graph embeddings may include points in a k-dimensional metric space (e.g., shown as a graph in two dimensions for simplicity) ([0021]). The disclosure reasonably describes the argued limitation of “embeddings of the nodes from the knowledge graph.” Further, Costabello discloses the fit/train engine may generate three knowledge graph embeddings (e.g., points) on a two-dimensional metric space. As further shown, the three points may include a point representing SARS (e.g., located at {1.15, 3.45}), a point representing that SARS has a treatment (e.g., located at {3.25, 1.15}), and a point representing the SARS vaccine (e.g., located at {4.82, 5.62}) ([0022]). The disclosure reasonably describes the argued limitation of “embeddings of the objects of the respective one of the data modalities.”
On page 24, Applicant argues Min fails to disclose or suggest the above-identified features of the independent claims, and the Office has not relied on Min for the above-identified features in any event. For instance, Min does address predicting new relationships in a knowledge graph. See Min, Abstract. However, Min's methodology is confined to processing and embedding the structural information of the knowledge graph itself (e.g., triplets of head, relation, tail entities) using convolutional techniques. See Min, Abstract and paragraphs [0025]-[0027]. As such, it is respectfully submitted that Min fails to disclose or suggest incorporating and jointly learning from objects of one or more data modalities alongside the knowledge graph structure to enrich entity representations or discover new relationships. Thus, it is respectfully submitted that Min fails to disclose or suggest at least the above-identified features of the independent claims as well. Applicant’s argument is not persuasive because argued limitations have been addressed by Costabello above.
On page 25, Applicant argues Costabello, Min, O'Keefe, Song, Lee, and Ishizaki fail to disclose or suggest at least the foregoing features of the independent claims. Thus, for at least the foregoing reasons, Costabello, Min, O'Keefe, Song, Lee, and Ishizaki, whether alone or in combination, could not render obvious independent claims 1, 10, and 15 or its dependent claims 2-9 and 11-15. Applicant’s argument is not persuasive because argued limitations have not been rejected under Costabello, Min, O'Keefe, Song, Lee, and Ishizaki.
On pages 25-26 as applied to claim 7, Applicant argues Costabello, Min, and Song do not collectively or individually describe the underlying multi-modal, jointly learned embeddings that are central to this claimed tradeoff. For example, claim 7 describes a hyperparameter to manage the balance between the overall prediction accuracy (of novel relationships) and the inherent characteristics or quality of these unified, multi-modal embeddings. This implies a sophisticated control over the representation learning process that integrates heterogeneous information sources. Applicant’s argument is not persuasive because the cited prior art discloses the limitation of “introducing a hyperparameter for controlling a tradeoff between a prediction accuracy and the unified embeddings” as cited. As for the argument of a hyperparameter to manage the balance between the overall prediction accuracy (of novel relationships) and the inherent characteristics or quality of these unified, multi-modal embeddings...This implies a sophisticated control over the representation learning process that integrates heterogeneous information sources. The argued limitations are not in claim 7, therefore, the disclosure in the cited prior art is not required.
Applicant argues Min cannot teach a hyperparameter to control a tradeoff specific to the quality of multi-modal, unified embeddings. Applicant’s argument is not persuasive because Min has not been cited to describe the argued limitations.
Applicant argues Song does not disclose the fusion of heterogeneous data modalities (e.g., text, images, and knowledge graph structure) into truly "unified embeddings" in the manner of described by the claims. Therefore, Song does not describe a hyperparameter for balancing prediction accuracy with the quality of such multi-modal unified embeddings. Applicant’s argument is not persuasive because the argued limitations are not recited in claim 7.
On pages 26-28 as applied to claims 8 and 9, Applicant argues claims 8 and 9 describe specific applications of the neural network system for predicting relationships in a biological context, namely "gene associated with disease" and "chemical treats disease," followed by ranking and selecting candidates for experiments or personalized drug development. These claims cannot be solely derived from Costabello, Min, and Lee because the claims rely on the fundamental novelty of a unified, multi-modal joint learning of embeddings in a knowledge graph context for novel relationship prediction. For example, the claims detail specific downstream applications of the novel system described in claim 1. The prediction of "gene associated with disease" or "chemical treats disease" relationships (which are often novel or yet undiscovered) directly benefits from the rich, multi-modal embeddings learned by jointly considering the knowledge graph structure and external data modalities (e.g., biomedical documents, images of chemicals/proteins). Further, the ranking and selection of candidates for knockdown experiments or personalized drug development are practical uses of the enhanced predictive power derived from this multi-modal embedding approach. The quality of these predictions is directly tied to the unified embeddings. Applicant describes how Costabello, Min, and Lee individually are distinct from the claimed invention. Applicant’s argument is not persuasive because Costabello, Min, and Lee as a whole render the claimed invention obvious over the cited prior art.
Specifically to Applicant’s argument that “claims 8 and 9 are distinct because these claims apply the unique multi-modal, joint embedding learning capability of claim 1 to specific, highly valuable biological prediction tasks (gene-disease, chemical-disease). It is respectfully submitted that the cited references either focus on explanation (Costabello), predict relationships only from KG structure (Min), or evaluate specific biological features for DTI but without the generalized multi-modal knowledge graph embedding approach (Lee). None of them provide the comprehensive, heterogeneous data integration and joint learning that forms the technical basis for the high-quality predictions outlined in claims 8 and 9”, Applicant’s argument is not persuasive because Costabello, Min, and Lee have been cited to describe the claims limitations. As for limitation not recited in the claim such as “the unique multi-modal, joint embedding learning capability of claim 1 to specific, highly valuable biological prediction tasks (gene-disease, chemical-disease)”, citation of Costabello, Min, and Lee is not required.
On pages 28-29 as applied to claim 14, Applicant’s arguments directed to Costabello and Min are not persuasive as discussed above.
Applicant argues “Ishizaki deals with "vocabulary information" and "token information" for combining structured documents, it is fundamentally different from claim 14 for several reasons. First, "Structured Documents" VS. "Biomedical Documents" is described. For instance, Ishizaki explicitly deals with structured documents and their combination. In contrast, claim 14 is about biomedical documents, which are typically includes unstructured text (e.g., scientific papers, clinical notes) requiring complex natural language processing (NLP) / text mining to extract meaningful information, not merely combining pre-structured data.” Applicant’s argument is not persuasive because the argued limitation of “biomedical documents, which are typically includes unstructured text (e.g., scientific papers, clinical notes) requiring complex natural language processing (NLP) / text mining to extract meaningful information, not merely combining pre-structured data” is not present in the claims.
Ishizaki's purpose is "combining" or merging structured information by resolving common vocabulary. Claim 14's purpose may be to prepare unstructured biomedical text into a format (multisets of tokens) suitable for generating embeddings that can be jointly learned with a knowledge graph for novel relationship prediction. The context and objective are entirely different. Additionally, regarding "multiset of tokens", while Ishizaki uses "token information," Ishizaki's use is within the context of replacing common vocabulary in structured documents. Ishizaki does not explicitly teach preparing a "set of biomedical documents, where each element of the set is a multiset of tokens from the vocabulary" for the purpose of feature extraction in a multi-modal learning system, especially in the nuanced biomedical domain. Biomedical text mining involves domain- specific challenges (e.g., entity recognition, disambiguation, specific vocabularies like MeSH, UMLS) that are not addressed by Ishizaki's general structured document combination.” Applicant’s argument is not persuasive because the claimed invention is not limited to just unstructured text. For example, the instant specification describes the documents 140 may be retrieved by manually or automatically crawling various data sources, including unstructured sources, like academic papers from PubMed, or structured data sources, such as ontologies like the Gene Ontology ([0040]). Therefore, Ishizaki in view of Costabello and Min describe the claimed invention as exemplified by the instant specification.
Applicant asserts the context and objective are entirely different. Additionally, regarding "multiset of tokens", while Ishizaki uses "token information," Ishizaki' s use is within the context of replacing common vocabulary in structured documents. Ishizaki does not explicitly teach preparing a "set of biomedical documents, where each element of the set is a multiset of tokens from the vocabulary" for the purpose of feature extraction in a multi-modal learning system, especially in the nuanced biomedical domain. Biomedical text mining involves domain-specific challenges (e.g., entity recognition, disambiguation, specific vocabularies like MeSH, UMLS) that are not addressed by Ishizaki's general structured document combination. Applicant’s argument is not persuasive because Ishizaki describes claim 14 as recited. As for the argument of “for the purpose of feature extraction in a multi-modal learning system, especially in the nuanced biomedical domain. Biomedical text mining involves domain-specific challenges (e.g., entity recognition, disambiguation, specific vocabularies like MeSH, UMLS) that are not addressed by Ishizaki's general structured document combination”, the limitations are not recited in the claims, therefore, citation in the prior art is not required.
PENDING MATTERS
Claims 1-20, filed April 01, 2026, are examined on the merits.
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 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
Step 1:
The claims recite a method, a computer system, and a computer-readable medium, which are statutory categories of invention.
Step 2A Prong One:
Claim 1 recites “establishing a knowledge graph…”, “annotating entities…”, and “identifying novel relationships…” at a high level of generality such that it could be practically performed in the human mind with the aid of paper and pencil. The limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for generic computer components. These limitations, as drafted, are processes that, under its broadest reasonable interpretation, can be performed as a mental process (that is, “observation, evaluation, judgement, opinion”).
The limitation of “training a neural network…” is directed to a mathematical calculations as exemplified in dependent claim 4.
Claims 10 and 15 are directed to a computer system and computer-readable medium comprising the same steps as in claim 1. These claims are similarly rejected under the same rationale as claim 1, supra.
Step 2A Prong Two
The judicial exception is not integrated into a practical application. In particular, the claims recite additional elements of a “memory” and “one or more processors”, where the claim further recite generic elements of the method, system, or computer-readable medium. The “memory” and “one or more processors” are recited at a high-level of generality such that they amount to no more than mere instructions to apply the exception using a generic component (MPEP 2106.05(f)).
Accordingly, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea.
Step 2B
The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the integration of the abstract idea into a practical application, the additional elements of “memory” and “one or more processors” are recited at a high-level of generality such that they amount to no more than mere instructions to apply the exception using a generic computer component (MPEP 2106.05(f)).
Thus taken alone, the individual elements do not amount to significantly more than the above-identified judicial exception (the abstract idea). Looking at the limitations as an ordered combination adds nothing that is not already present when looking at the elements taken individually. There is no indication that the combination of elements improves the functioning of a computer or improves any other technology.
Claim 2 recites wherein the one or more data modalities include objects in form of biomedical documents describing details of the entities and/or in form of images. These limitations further narrow the abstract idea or extra-solution activity, but are nonetheless part of the abstract idea identified in claim 1. They also do not amount to significantly more than the abstract idea. The claims are similarly rejected under the same rationale as claim 1, supra.
Claim 3 recites wherein a machine learning model maintained by the neural network system is trained with a single sample at a time, wherein a sample is provided as a triple composed of a head entity, a tail entity and a corresponding relation type between the head entity and the tail entity, and wherein, for a particular head entity and a particular relation type, the tail entities are grouped into a positive subset and into a negative subset in such a way that for each tail entity in the positive subset it holds that a respective triple exists in the knowledge graph, while for each tail entity in the negative subset it holds that no respective triple exists in the knowledge graph. These limitations further narrow the abstract idea or extra-solution activity, but are nonetheless part of the abstract idea identified in claim 1. They also do not amount to significantly more than the abstract idea. The claims are similarly rejected under the same rationale as claim 1, supra.
Claim 4 recites wherein the neural network system is configured to minimize for a particular positive sample a loss function that calculates a Euclidean distance between a distance between the embeddings of the nodes from the knowledge graph for the positive sample and a center of the respective negative samples and a distance between the embeddings from the object of a particular of the data modalities for the positive sample and a center of the negative samples. These limitations further narrow the abstract idea, mathematical concept, or extra-solution activity, but are nonetheless part of the abstract idea identified in claim 1. They also do not amount to significantly more than the abstract idea. The claims are similarly rejected under the same rationale as claim 1, supra.
Claim 5 recites wherein the loss function is calculated individually for each pair of utilized data modalities. These limitations further narrow the abstract idea, mathematical concept, or extra-solution activity, but are nonetheless part of the abstract idea identified in claim 1. They also do not amount to significantly more than the abstract idea. The claims are similarly rejected under the same rationale as claim 1, supra.
Claim 6 recites wherein the neural network system measures a prediction accuracy of a respective one of a trained machine learning model based on learned weights of the embeddings. These limitations further narrow the abstract idea, mathematical concept, or extra-solution activity, but are nonetheless part of the abstract idea identified in claim 1. They also do not amount to significantly more than the abstract idea. The claims are similarly rejected under the same rationale as claim 1, supra.
Claim 7 recites introducing a hyperparameter for controlling a tradeoff between a prediction accuracy and the unified embeddings. These limitations further narrow the abstract idea or extra-solution activity, but are nonetheless part of the abstract idea identified in claim 1. They also do not amount to significantly more than the abstract idea. The claims are similarly rejected under the same rationale as claim 1, supra.
Claim 8 recites selecting a particular disease; using the neural network system to predict for the-selected disease relationships of a form ‘gene associated with disease’; ranking predicted genes according to a likelihood of the respective one of the predicted genes to be associated with the selected disease; and selecting a predefined number of the top-ranked genes as candidates for a knockdown experiment. These limitations further narrow the abstract idea, mathematical concept, or extra-solution activity, but are nonetheless part of the abstract idea identified in claim 1. They also do not amount to significantly more than the abstract idea. The claims are similarly rejected under the same rationale as claim 1, supra.
Claim 9 recites selecting a particular disease; using the neural network system to predict for the-selected disease relationships of a form ‘chemical treats disease’: ranking the-predicted chemicals according to a likelihood of the-a respective one of the predicted chemicals to treat the selected disease; and selecting a predefined number of the top-ranked chemicals as candidates for personalized drug development. These limitations further narrow the abstract idea or extra-solution activity, but are nonetheless part of the abstract idea identified in claim 1. They also do not amount to significantly more than the abstract idea. The claims are similarly rejected under the same rationale as claim 1, supra.
Claim 16 recites the neural network system comprises a first machine learning model and a second machine learning model, and wherein training the neural network system to jointly learn the embeddings of the nodes from the knowledge graph and the embeddings of the objects of the respective one of the data modalities comprises: generating the embeddings of the nodes of the knowledge graph using the first machine learning model; generating the embeddings of the objects of the respective one of the data modalities using the second machine learning model; jointly training the first machine learning model and the second machine learning model based on the embeddings of the nodes of the knowledge graph and the embeddings of the objects of the respective one of the data modalities. These limitations further narrow the abstract idea or extra-solution activity, but are nonetheless part of the abstract idea identified in claim 1. They also do not amount to significantly more than the abstract idea. The claims are similarly rejected under the same rationale as claim 1, supra.
Claim 17 reciters the embeddings of the nodes of the knowledge graph comprise a first positive sample and one or more first negative samples associated with the first positive sample, wherein the embeddings of the objects of the respective one of the data modalities comprise a second positive sample and one or more second negative samples associated with the second positive sample, and wherein jointly training the first machine learning model and the second machine learning model comprises: calculating a first Euclidean distance based on the first positive sample and the one or more first negative samples; calculating a second Euclidean distance based on the second positive sample and the one or more second negative samples; calculating an offset loss based on the first Euclidean distance and the second Euclidean distance; and jointly training the first machine learning model and the second machine learning model based on the offset loss. These limitations further narrow the abstract idea or extra-solution activity, but are nonetheless part of the abstract idea identified in claim 1. They also do not amount to significantly more than the abstract idea. The claims are similarly rejected under the same rationale as claim 1, supra.
Claim 18 recites the neural network system further comprises a third machine learning model, and wherein training the neural network system to jointly learn the embeddings of the nodes from the knowledge graph and the embeddings of the objects of the respective one of the data modalities further comprises: generating embeddings of the relationships from the knowledge graph using the third machine learning model, and wherein jointly training the first machine learning model and the second machine learning model further comprises jointly training the first machine learning model, the second machine learning model, and the third machine learning model based on the embeddings of the nodes of the knowledge graph, the embeddings of the objects of the respective one of the data modalities, and the embeddings of the relationships from the knowledge graph. These limitations further narrow the abstract idea or extra-solution activity, but are nonetheless part of the abstract idea identified in claim 1. They also do not amount to significantly more than the abstract idea. The claims are similarly rejected under the same rationale as claim 1, supra.
Claim 19 recites jointly training the first machine learning model, the second machine learning model, and the third machine learning model comprises: calculating an offset loss based on the embeddings of the nodes of the knowledge graph and the embeddings of the objects of the respective one of the data modalities; calculating a prediction loss based on the embeddings of the nodes of the knowledge graph, the embeddings of the objects of the respective one of the data modalities, and the embeddings of the relationships from the knowledge graph; calculating a total loss based on the offset loss and the prediction loss; and jointly updating parameters of the first machine learning model, the second machine learning model, and the third machine learning model based on the total loss. These limitations further narrow the abstract idea or extra-solution activity, but are nonetheless part of the abstract idea identified in claim 1. They also do not amount to significantly more than the abstract idea. The claims are similarly rejected under the same rationale as claim 1, supra.
Claim 20 recites training the neural network system to jointly learn the embeddings of the nodes from the knowledge graph and the embeddings of the objects of the respective one of the data modalities further comprises: generating second embeddings of second objects of a second data modality of the data modalities, wherein the respective one of the data modalities are documents and the second data modality are images, and wherein jointly training the first machine learning model and the second machine learning model comprises: calculating a first offset based on the embeddings of the nodes of the knowledge graph and the embeddings of the objects of the documents; calculating a second offset based on the embeddings of the nodes of the knowledge graph and the second embeddings of the second objects; calculating a third offset based on the embeddings of the objects of the documents and the second embeddings of the second objects; and jointly training the first machine learning model and the second machine learning model based on the first offset, the second offset, and the third offset. These limitations further narrow the abstract idea or extra-solution activity, but are nonetheless part of the abstract idea identified in claim 1. They also do not amount to significantly more than the abstract idea. The claims are similarly rejected under the same rationale as claim 1, supra.
Claims 11-14 are directed to a computer system and computer-readable medium comprising the same steps as in claims 2-9. These claims are similarly rejected under the same rationale as claims 2-9, supra.
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.
Claims 1-3, 10, and 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Costabello et al. (Costabello hereafter, US 2019/0220524 A1) in view of Min et al. (Min hereafter, US 2019/0122111 A1).
Claim 1, Costabello discloses a computer-implemented method of learning novel relationships among various entities, including biological entities such as chemicals, proteins, and diseases, the method comprising:
establishing a knowledge graph wherein each of the entities is represented as a node and each relationship between the entities is represented as an edge between the respective nodes ([0017], e.g. the knowledge graph converter may generate a knowledge graph based on the training data and the ontology (e.g., based on the converted and aggregated training data and the ontology));
annotating entities in the knowledge graph with objects of one or more data modalities ([0019], e.g. the knowledge graph may indicate that SARS is a disease with a vaccine treatment by the SARS vaccine)….
Costabello discloses training with the knowledge graph ([0013], e.g. a user of the user device (e.g., via a user interface provided to the user) may cause the user device to provide, to the prediction platform, training data for training a knowledge graph associated with a particular disease (e.g., severe acute respiratory syndrome (SARS)))…treats the knowledge graph and the objects of a respective one of the data modalities in a unified manner by jointly learning embeddings of the nodes from the knowledge graph ([0022], e.g. the fit/train engine may generate three knowledge graph embeddings (e.g., points) on a two-dimensional metric space) and embeddings of the objects of the respective one of the data modalities ([0022], e.g. the fit/train engine may generate three knowledge graph embeddings (e.g., points) on a two-dimensional metric space. As further shown, the three points may include a point representing SARS (e.g., located at {1.15, 3.45}), a point representing that SARS has a treatment (e.g., located at {3.25, 1.15}), and a point representing the SARS vaccine (e.g., located at {4.82, 5.62})); and using the learned embeddings for identifying novel relationships among the entities ([0011], e.g. knowledge graph embedding models predict existences of labeled links between entities. Such predictions are a result of operations between points (e.g., known as embeddings) in a metric space. The embeddings are learned from the entire knowledge graph during training of the knowledge graph).
However, Costabello does not train a neural network system with the knowledge graph. Min discloses train a neural network system with the knowledge graph ([0027], e.g. a convolutional neural network (CNN) is used to learn the entity and relationship embedding and their connections, and [0069]) where training the CNN model can be treated as a pairwise ranking problem where one positive triplet should have a higher score than the negative triplets).
Min discloses much better performance can be achieved with the ACNN than other competing approaches for exploring unseen relationships and performing knowledge graph completion, which can be used to improve the system performance for many natural language processing applications such as sentence classification, sentiment analysis, question answering, and sentence reasoning ([0030]). One of ordinary skill in the art at the time prior to the effective filing date of the instant invention would have been motivated by Min to improve the system performance Costabello. Therefore, it would have been obvious for one of ordinary skill in the art to use the system of Costabello with the neural network of Min. The benefit would be to achieve much better performance.
Claim 2, Costabello as modified discloses wherein the one or more data modalities include objects in form of biomedical documents describing details of the entities and/or in form of images ([0014], e.g. the training data may include information associated with a subject of the ontology. For example, example implementation 100 relates to an ontology associated with the SARS disease. Thus, the training data may include data associated with the SARS disease that is received from relationship database management systems (RDBMS), comma-separated values (CSV) data stores, and/or the like. As shown in FIG. 1A, the training data may include data indicating a disease (e.g., SARS), a cause of the disease (e.g., virus_XYZ), what organ the disease affects (e.g., lungs), symptoms of the disease (e.g., high fever), a virus identifier (e.g., virus_XYZ), a protein sequence associated with the virus (e.g., ACARBAC), a drug identifier associated with a drug that treats the disease (e.g., SARS vaccine), a drug type (e.g., vaccine), what the drug treats (e.g., SARS), and/or the like). It is noted the limitation of “documents” has been attributed with the definition of “a piece of written, printed, or electronic matter that provides information or evidence or that serves as an official record.” See Oxford Languages dictionary.
Claim 3, Costabello as modified discloses a machine learning model maintained by the neural network system is trained with a single sample at a time, wherein a sample is provided as a triple composed of a head entity, a tail entity and a corresponding relation type between the head entity and the tail entity, and wherein, for a particular head entity and a particular relation type ([0041], e.g. new triplets based on the existing incomplete graph: (1) given a head or tail and one kind of relationship, l, find the associated tail or head, (h, t), in the entity set; (2) given one head, h, and one tail, t, find the relationship, l, between these two entities), the tail entities are grouped into a positive subset and into a negative subset in such a way that for each tail entity in the positive subset it holds that a respective triple exists in the knowledge graph, while for each tail entity in the negative subset it holds that no respective triple exists in the knowledge graph ([0043], e.g. Existing known triplets can be used as positive samples, and negative samples can be created by corrupting positive triplets to train the CNN models. Positive triplets (h, l, t) can have a small distance between h+l and t while negative triplets (h′, l, t′) will have big distance between h′+l and t′. The relationship between two entities corresponds to a translation between the embeddings of entities, that is, h+l+=t when the relation between (h, l, t) is true, and the translation for h+l+t for (h′, l, t′)).
Claims 10 and 15 are directed to a computer system and computer-readable medium comprising the same steps as in claims 1-3. These claims are similarly rejected under the same rationale as claims 1-3, supra.
Claims 6 and 13 is/are rejected under 35 U.S.C. 103 as being unpatentable over Costabello et al. (Costabello hereafter, US 2019/0220524 A1) in view of Min et al. (Min hereafter, US 2019/0122111 A1), as applied to claims 1-3, 10, and 15 above, in further view of O’Keefe et al. (O’Keefe hereafter, US 11621081 B1).
Claims 6 and 13, Costabello as modified discloses the claimed invention except for the limitation of “wherein the neural network system measures a prediction accuracy of a respective one of a trained machine learning model based on learned weights of the embeddings.” O’Keefe discloses wherein the neural network system measures a prediction accuracy of a respective one of a trained machine learning model based on learned weights of the embeddings (column 7, line 57 to column 8, line 13, e.g. encoding function of the neural network can correspond to a word embedding function that serves to further enhance the prediction accuracy predictions about undiagnosed conditions that are generated by the system).
O’Keefe discloses a method for improving overall patient health and reducing health system costs (column 1, lines 15-17). One of ordinary skill in the art at the time prior to the effective filing date of the instant invention would have been motivated by O’Keefe to improve the method of Costabello as modified. Therefore, it would have been obvious for one of ordinary skill in the art to use method of Costabello as modified with the embedding model of O’Keefe. The benefit would be for improving overall patient health and reducing health system costs.
Claim 7 is/are rejected under 35 U.S.C. 103 as being unpatentable over Costabello et al. (Costabello hereafter, US 2019/0220524 A1) in view of Min et al. (Min hereafter, US 2019/0122111 A1), as applied to claims 1-3, 10, and 15 above, in further view of Song et al. (Song hereafter, US 20200090039 A1).
Costabello as modified discloses the claimed invention except for “introducing a hyperparameter for controlling a tradeoff between a prediction accuracy and the unified embeddings.” Song discloses “introducing a hyperparameter for controlling a tradeoff between a prediction accuracy and the unified embeddings” ([0056], e.g. system 200 can generate a unified embedding model that achieves equivalent performance and recognition accuracy when compared to individual specialized models. Further, the unified model can have the same, or even less, model complexity as a single individual specialized model. Hence, this specification describes improved processes and methods for easing or reducing the difficulties in training model embeddings for multiple verticals such that a unified model can be generated).
One of ordinary skill in the art at the time prior to the effective filing date of the instant invention would have been motivated by Song to improve the method of Costabello as modified. Therefore, it would have been obvious for one of ordinary skill in the art to use the method of Costabello as modified with the unified embeddings of Song. The benefit would be for improved processes and methods for easing or reducing the difficulties in training model embeddings for multiple verticals such that a unified model can be generated.
Claims 8 and 9 is/are rejected under 35 U.S.C. 103 as being unpatentable over Costabello et al. (Costabello hereafter, US 2019/0220524 A1) in view of Min et al. (Min hereafter, US 2019/0122111 A1), as applied to claims 1-3, 10, and 15 above, in further view of Lee et al. (Comparison of Target Features for Predicting Drug-Target Interactions by Deep Neural Network Based on Large-Scale Drug-Induced Transcriptome Data, 2019).
Claim 8, Costabello as modified does not explicitly disclose the limitation of “selecting a particular disease; using the neural network system to predict for the-selected disease relationships of a form ‘gene associated with disease’; ranking predicted genes according to a likelihood of the respective one of the predicted genes to be associated with the selected disease; and selecting a predefined number of the top-ranked genes as candidates for a knockdown experiment.” It is noted that “selecting a particular disease…ranking predicted genes according to a likelihood of the respective one of the predicted genes to be associated with the selected disease; and selecting a predefined number of the top-ranked genes as candidates for a knockdown experiment” are well know biomedical drug discovery steps (Lee, page 2, e.g. Although a number of repositioning drug candidates were discovered by CMAP approach based on inverse pattern matching between drug- and disease-expression signatures [18–21], it remains a challenge to identify their underlying physical targets. In LINCS dataset, the expression 955 landmark genes (L1000 genes) were actually measured, and the inferred expression levels are provided for additional ~12,000 genes). Further, Lee discloses the use of deep neural network Deep Neural Network Based on Large-Scale Drug-Induced Transcriptome Data via gene knockdown (page 2, e.g. initial version of CMAP is a collection of ~7000 drug-induced expression profiles (DEPs) in human cancer cell lines treated with 1309 compounds. Recently, CMAP has been extended to the L1000 dataset of Library of Integrated Network-based Cellular Signatures (LINCS) [17], a resource containing 1.3 million gene expression profiles associated with 20,413 chemical perturbagens and ~5000 genetic perturbagens (i.e., single gene knockdown or overexpression)).
Lee discloses Overall, the performances of DNNs improved compared to those in the previous section (Figures 2 and 3), probably due to expanded training dataset (page 6, e.g. Overall, the performances of DNNs improved compared to those in the previous section (Figures 2 and 3), probably due to expanded training dataset). One of ordinary skill in the art at the time prior to the effective filing date of the instant invention would have been motivated by Lee to improve the method of Costabello as modified. Therefore, it would have been obvious for one of ordinary skill in the art to use the method of Costabello as modified with for drug discovery as suggests by Lee. The benefit would be for improved performances of DNNs improved compared to those of the prior art.
Claim 9, Costabello as modified does not explicitly disclose the limitation of “selecting a particular disease…ranking the-predicted chemicals according to a likelihood of the-a respective one of the predicted chemicals to treat the selected disease; and selecting a predefined number of the top-ranked chemicals as candidates for personalized drug development.” It is noted that “selecting a particular disease; using the neural network system to predict for the-selected disease relationships of a form ‘chemical treats disease’: ranking the-predicted chemicals according to a likelihood of the-a respective one of the predicted chemicals to treat the selected disease; and selecting a predefined number of the top-ranked chemicals as candidates for personalized drug development.” are well known biomedical drug discovery steps (Lee, page 1, e.g. In silico prediction of drug–target interaction (DTI) is becoming more data-driven than conventional modeling-based approaches such as docking or molecular dynamic simulation. Deep neural network (DNN) is increasingly being applied to highly complex and challenging problems such as protein folding [1]. The vast majority of drugs and compounds are expected to interact with multiple targets, i.e., polypharmacology [2]. While millions of DTIs have been identified, and increasingly keep being revealed, it is still costly and time-consuming to validate DTIs experimentally even by high throughput screening (HTS) [3]. It is most likely that there still exist unknown DTIs for both approved drugs and clinical candidate compounds. Such hidden DTIs could critically impact the drug development process including unexpected clinical outcome, or may broaden their indications through drug repositioning). Further, Lee discloses the use of deep neural network Deep Neural Network Based on Large-Scale Drug-Induced Transcriptome Data via gene knockdown (page 2, e.g. initial version of CMAP is a collection of ~7000 drug-induced expression profiles (DEPs) in human cancer cell lines treated with 1309 compounds. Recently, CMAP has been extended to the L1000 dataset of Library of Integrated Network-based Cellular Signatures (LINCS) [17], a resource containing 1.3 million gene expression profiles associated with 20,413 chemical perturbagens and ~5000 genetic perturbagens (i.e., single gene knockdown or overexpression)).
Claim 14 is/are rejected under 35 U.S.C. 103 as being unpatentable over Costabello et al. (Costabello hereafter, US 2019/0220524 A1) in view of Min et al. (Min hereafter, US 2019/0122111 A1), as applied to claims 1-3, 10, and 15 above, in further view of Ishizaki (US 2009/0132564 A1).
Claim 14, Costabello as modified discloses the claimed invention except for a biomedical documents mining component that is configured to prepare biomedical documents based on a common vocabulary and to generate a set of biomedical documents, where each element of the set is a multiset of tokens from the vocabulary. Ishizaki discloses a biomedical documents mining component that is configured to prepare biomedical documents based on a common vocabulary and to generate a set of biomedical documents, where each element of the set is a multiset of tokens from the vocabulary ([0009], e.g. an acquiring unit configured to acquire, for each of the plurality of structured documents, correspondence information showing a correspondence between vocabulary information constituting the structured document and token information, and encoded information in which the vocabulary information included in the structured document has been replaced with corresponding token information based on the correspondence information; a replacing unit configured to, if the vocabulary information included in a first correspondence information is common with the vocabulary information included in a second correspondence information replace common vocabulary information included in the first correspondence information with a reference to the common vocabulary information included in the second correspondence information; and a combining unit configured to generate combined information by combining the correspondence information in which the common vocabulary information has been replaced and the encoded information, of each of the plurality of structured documents).
Ishizaki discloses an improvement that provides a technique that enables redundancy of repeated description to be further reduced in the case where a plurality of structured documents are combined ([0008]). One of ordinary skill in the art at the time prior to the effective filing date of the instant invention would have been motivated by Ishizaki to improve the system of Costabello as modified. Therefore, it would have been obvious for one of ordinary skill in the art to use the method of Costabello with the document acquiring unit of Ishizaki. The benefit would be to enable redundancy of repeated description to be further reduced in the case where a plurality of structured documents are combined.
PERTINENT PRIOR ART
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Alshahrani discloses our method combines symbolic methods, in particular knowledge representation using symbolic logic and automated reasoning, with neural networks to generate embeddings of nodes that encode for related information within knowledge graphs. Through the use of symbolic logic, these embed dings contain both explicit and implicit information. We apply these embeddings to the prediction of edges in the knowledge graph representing problems of function prediction, finding candidate genes of diseases, protein-protein interactions, or drug target relations, and demonstrate performance that matches and sometimes outperforms traditional approaches based on manually crafted features (Abstract).
Ali et al. (BioKEEN: a library for learning and evaluating biological knowledge graph embeddings, 2019) discloses developed BioKEEN (Biological KnowlEdge EmbeddiNgs) and PyKEEN (Python KnowlEdge EmbeddiNgs) to facilitate their easy use through an interactive command line interface. Finally, we present a case study in which we used a novel biological pathway mapping resource to predict links that represent pathway crosstalks and hierarchies (Abstract).
STATUS OF PRIOR ART
Claims 16-20 are free of any prior art.
CONCLUSION
THIS ACTION IS MADE FINAL. 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.
Patent applicants with problems or questions regarding electronic images that can be viewed in the Patent Application Information Retrieval system (PAIR) can now contact the USPTO's Patent Electronic Business Center (Patent EBC) for assistance. Representatives are available to answer your questions daily from 6 am to midnight (EST). The toll free number is (866) 217-9197. When calling please have your application serial or patent number, the type of document you are having an image problem with, the number of pages and the specific nature of the problem. The Patent Electronic Business Center will notify applicants of the resolution of the problem within 5-7 business days. Applicants can also check PAIR to confirm that the problem has been corrected. The USPTO's Patent Electronic Business Center is a complete service center supporting all patent business on the Internet. The USPTO's PAIR system provides Internet-based access to patent application status and history information. It also enables applicants to view the scanned images of their own application file folder(s) as well as general patent information available to the public.
For all other customer support, please call the USPTO Call Center (UCC) at 800-786-9199. The USPTO's official fax number is 571-272-8300.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to C. Dune Ly, whose telephone number is (571) 272-0716. The examiner can normally be reached on Monday-Friday from 8 A.M. to 4 PM ET.
If attempts to reach the examiner by telephone are unsuccessful, the examiner's supervisor, Tony Mahmoudi, can be reached on 571-272-4078.
/Cheyne D Ly/
Primary Examiner, Art Unit 2152
6/23/2026
Prosecution Insights