DETAILED ACTION
Applicant’s response, filed 15 Dec. 2025, has been fully considered. The following rejections and/or objections are either reiterated or newly applied. They constitute the complete set presently being applied to the instant application.
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 .
Status of Claims
Claims 18-20 are cancelled.
Claims 1-17 are pending.
Claims 1-17 are rejected.
Priority
The effective filing date of the claimed invention is 31 March 2022.
Drawings
The drawings received 16 March 2022 are objected to as failing to comply with 37 CFR 1.84(p)(5) because they do not include the following reference sign(s) mentioned in the description:
252-1, 252-2,…252-n at para. [0046]; and
and 254-1, 254-2,…254-n at para. [0046].
Corrected drawing sheets in compliance with 37 CFR 1.121(d) are required in reply to the Office action to avoid abandonment of the application. Any amended replacement drawing sheet should include all of the figures appearing on the immediate prior version of the sheet, even if only one figure is being amended. Each drawing sheet submitted after the filing date of an application must be labeled in the top margin as either “Replacement Sheet” or “New Sheet” pursuant to 37 CFR 1.121(d). If the changes are not accepted by the examiner, the applicant will be notified and informed of any required corrective action in the next Office action. The objection to the drawings will not be held in abeyance.
Response to Arguments
Applicant's arguments filed 15 Dec. 2025 regarding the drawing objections have been fully considered but they are not persuasive.
Applicant remarks that FIG. 5C was amended to include the reference signs mentioned in the description (Applicant’s remarks at pg. 9, para. 1).
This argument is not persuasive. While Applicant states replacement drawings include FIG. 5C, no replacement sheet for FIG. 5C was provided. Therefore the figure is objected to for the same reasons discussed in the previous Office action mailed 17 Sept. 2025.
Specification
The disclosure is objected to because of the following informalities. This objection is newly recited.
Para. [0051] recites “…enables medicinal chemists can select lead candidate series explore chemical space similar…, reduces failure rates…, and accelerate the drug discovery process…”, which is grammatically incorrect and nonsensical, and should read “enables medicinal chemists [[can]] to select lead candidate…, explore a chemical space… reduce[[s]] failure rates…, and accelerate…”.
Appropriate correction is required.
Claim Interpretation
Applicant specification at para. [0036] defines “order-independent representation” to refer to a uniquely defined textual or numerical representation of the structure that is independent of arbitrary ordering of the atoms.
Applicant’s specification at para. [0027] defines “order-dependent representation” to refer to a non-unique text representation that defines the structure of the chemical compound.
Claim Rejections - 35 USC § 112(a)
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
Claims 1-17 are rejected under 35 U.S.C. 112(a) as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor at the time the application was filed, had possession of the claimed invention. This rejection is newly recited and necessitated by claim amendment.
Claims 1 and 11, and claims dependent therefrom, recite “synthesizing a new compound…; and testing the new chemical compound to determine a chemical property of the new chemical compound, a pharmacological property of the new chemical compound, or a combination thereof”, which lacks written description in the disclosure.
Applicant remarks that support for the amendments are found in at least para. [0050]-[0052] of the specification (Applicant’s remarks at pg. 9, para. 1). However, Applicant’s specification at para. [0050] merely discloses that the chemical property prediction module may “generate/identify new compounds that are related to the input”, which provides support for computationally identifying a new chemical compound related to the reproduced order-dependent representation of the chemical compound. Applicant’s specification at para. [0051] discusses the advantages of the chemical property prediction module, reducing the time needed for a medicinal chemist to modify a chemical compound and generate a lead compound to achieve a desired level of potency and other chemical/pharmacological properties, and further explains the module enables chemists to select lead candidate series, explore a chemical space similar to the chemical compound, and accelerate and reduce failure rates in drug discovery; however, this does not clearly require or imply synthesizing the new chemical compound and physically testing the new chemical compound for certain properties. Instead, para. [0051] encompasses a medicinal chemist using the chemical property prediction module to explore the chemical space around the input compound and identify the new compound computationally (i.e. selecting a lead compound for drug discovery). See MPEP 2163 II. and In re Robertson, 169 F.3d 743, 745, 49 USPQ2d 1949, 1950-51 (Fed. Cir. 1999) ("To establish inherency, the extrinsic evidence ‘must make clear that the missing descriptive matter is necessarily present in the thing described in the reference, and that it would be so recognized by persons of ordinary skill. Inherency, however, may not be established by probabilities or possibilities. The mere fact that a certain thing may result from a given set of circumstances is not sufficient.’" (citations omitted)). Applicant’s specification at para. [0052] discusses FIG. 6, which overviews the training process of the machine learning model to predict properties of chemical compounds, but does not disclose synthesizing a compound and testing the synthesized compound. After a further review of Applicant’s specification, there appears to be no other discussion of synthesizing and physically testing a new chemical compound identified by the chemical property prediction module.
For the reasons discussed above, the specification does not provide a sufficient disclosure of the limitations of synthesizing and testing recited in claims 1 and 11, and claims dependent therefrom, to demonstrate to one of ordinary skill in the art that the inventor possessed the invention at the time the application was filed. THS IS A NEW MATTER REJECTION. For more information regarding the written description requirement, see MPEP §2161.01- §2163.07(b).
Claim 11 and claims dependent therefrom, recite “A system…, the system comprising: one or more processors and one or more nontransitory computer-readable mediums storing instructions…, when the instructions comprise:…synthesizing a new compound…; and testing the new chemical compound to determine a chemical property of the new chemical compound, a pharmacological property of the new chemical compound, or a combination thereof”, which requires a processor configured to physical synthesizing a compound and testing the physical compound for a chemical or pharmacological property.
First, as discussed above, the disclosure does not provide support for synthesizing a new chemical compound and then testing the new compound for a chemical or pharmacological property. Furthermore, while Applicant’s specification at para. [0057]-[0059] does disclose a system comprising a memory and processor configured to carry out the computational steps of claim 11 up to predicting properties of the chemical compound and identifying a new chemical compound related to the chemical compound, Applicant’s specification does not provide support for any specialized hardware comprising a processor that is configured to carry out a physical synthesis step and/or robotically assay a synthesized compound to determine the recited property.
For the reasons discussed above, the specification does not provide a sufficient disclosure of the limitations of a processor configured to synthesize and test a compound recited in claim 11, and claims dependent therefrom, to demonstrate to one of ordinary skill in the art that the inventor possessed the invention at the time the application was filed. THS IS A NEW MATTER REJECTION. For more information regarding the written description requirement, see MPEP §2161.01- §2163.07(b).
Claim Rejections - 35 USC § 112(b)
The rejection of claims 1-2, 4-14, and 16-20 under 35 U.S.C. 112(b) in the Office action mailed 17 Sept. 2025 has been withdrawn in view of claim amendments and cancellations received 15 Dec. 2025.
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
Claims 3 and 15 are rejected under 35 U.S.C. 112(b) as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor, regards as the invention. This rejection is previously recited.
Claims 3 and 15 are indefinite for recitation of “the graph”. Claims 1 and 11, from which claims 3 and 15 depend, previously recite “…generation a graph of the chemical compound” and “a molecular graph representation of the chemical compound…”. Therefore, it is unclear if “the graph” in each instance is referring to the graph of the chemical compound or the molecular graph. Clarification is requested. For purpose of examination, the graph is interpreted to refer to “the generated graph”.
Response to Arguments
Applicant's arguments filed 15 Dec. 2025 regarding 35 U.S.C. 112(b) have been fully considered but they are not persuasive with respect to claims 3 and 15.
Applicant remarks that claims 1 and 11 were amended in a manner to overcome the rejection of the claims, and thus the rejection of the dependent claims should be withdrawn for the same reasons as claims 1 and 11 (Applicant’s remarks at pg. 10, para. 2-4).
This argument is not persuasive. While claims 1 and 11 were amended to specify “the generated graph”, claims 3 and 15 still refer to “the graph”. As a result, it is unclear which graph is being referenced as discussed above.
Claim Rejections - 35 USC § 101
The rejection of claims 18-20 under 35 U.S.C. 101 in the Office action mailed 17 Sept. 2025 has been withdrawn in view of the cancellation of these claims received 15 Dec. 2025.
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 1-17 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. Any newly recited portion is necessitated by claim amendment.
The Supreme Court has established a two-step framework for this analysis, wherein a claim does not satisfy § 101 if (1) it is “directed to” a patent-ineligible concept, i.e., a law of nature, natural phenomenon, or abstract idea, and (2), if so, the particular elements of the claim, considered “both individually and as an ordered combination,” do not add enough to “transform the nature of the claim into a patent-eligible application.” Elec. Power Grp., LLC v. Alstom S.A., 830 F.3d 1350, 1353 (Fed. Cir. 2016) (quoting Alice, 134 S. Ct. at 2355). Applicant is also directed to MPEP 2106.
Step 1: The instantly claimed invention (claims 1 and 11 being representative) is directed a method and system. Therefore, the instantly claimed invention falls into one of the four statutory categories. [Step 1: YES]
Step 2A: First it is determined in Prong One whether a claim recites a judicial exception, and if so, then it is determined in in Prong Two if the recited judicial exception is integrated into a practical application of that exception.
Step 2A, Prong 1: Under the MPEP § 2106.04, the Step 2A (Prong 1) analysis requires determining whether a claim recites an abstract idea, law of nature, or natural phenomenon.
Claims 1 and 11 recite the following steps which fall under the mathematical concepts and/or mental processes groupings of abstract ideas:
generating a graph of a chemical compound based on at least one of an order-dependent representation of the chemical compound and a molecular graph representation of the chemical compound;
encoding the generated graph based on at least one of an adjacency matrix of a graph convolution neural network (GCN), one or more characteristics of the generated graph, one or more activation functions of the GCN, and one or more weights of the GCN to generate a latent vector representation of the chemical compound;
decoding the latent vector representation based on a plurality of hidden states of a neural network (NN)/recurrent neural network (RNN) to generate a reproduced order-dependent representation of the chemical compound; and
training the machine learning model based on the reproduced order-dependent representation, wherein the machine learning model includes the GCN and the RNN, and wherein the machine learning model is configured to predict one or more properties of the chemical compound (claim 11 only).
The identified claim limitations falls into one of the groups of abstract ideas of mathematical concepts and/or mental processes, for the following reasons. First, the limitations of generating a graph of a chemical compound based on the order-dependent representation and molecular graph representation encompass analyzing the chemical formula or SMILES string (i.e. the order-dependent representation) and a ball and stick model of the chemical compound (i.e. molecular graph representation) and creating, mentally aided with pen and paper, nodes corresponding to identified substructures and edges based on identified fragments, which amounts to a mere analysis of data that can be practically performed in the mind. Furthermore, encoding the generated graph based on characteristics of the generated graph, which involves mentally analyzing the graph and converting the information into a binary format, such as by a one-hot encoding routine as described in Applicant’s specification at para. [0043], which can be practically performed mentally aided with pen and paper. That is, other than reciting these limitations are carried out by a processor, nothing in the claims precludes the steps from being practically performed in the mind.
The steps of encoding the generated graph based on an adjacency matrix of a GCN, activation functions of the GCN, or one or more weights of the GCN to generate a latent vector representation of the chemical compound, decoding the latent vector representation based on a plurality of hidden states of a neural network or recurrent neural network to generate a reproduced order-dependent representation, and training the machine learning model including the GCN and RNN, to predict one or more properties of a chemical compound further recite a mathematical concept. Generating a latent vector representation by encoding the generated graph using an adjacency matrix of the GCN, activation functions, and/or weights of the GCN amounts to a textual equivalent to performing mathematical calculations. As discussed in Applicant’s specification at para. [0036] discloses encoding the graph into a latent vector representation combines features by summing the one or more learned features to generate a fixed-size descriptor vector or a scale-invariant feature vector. Encoding graph data into a latent vector space recites a mathematical process of performing a dimensional reduction using mathematical operations, as described in the specification, and thus recites a mathematical concept. Similarly, decoding the latent vector representation to generate a reproduced order-dependent representation based on hidden states of a neural network recites the mathematical concept of using mathematical calculations to map the low-dimensional data back into the high-dimensional space, as discussed in Applicant’s specification at para. [0040]-0042]. Last, training the machine learning including the GCN and RNN, when given its broadest reasonable interpretation in light of the specification at para. [0047]-[0048], amounts to a textual equivalent of performing mathematical calculations as the only disclosed embodiment, including determining aggregate loss values based on a loss function that derives differences between the input and reproduced order-dependent representation and/or molecular fingerprint. Therefore these limitations recite a mathematical concept. See MPEP 2106.04(a)(2) I.
Dependent claims 2-10 and 12-17 further recite an abstract idea and/or are part of the abstract idea of claims 1 and 11. Dependent claims 2 and 14 further limit the mathematical concept of decoding the latent vector to reproduce a SMILES string for the chemical compound. Dependent claims 3 and 15 further recite the mental process of identifying one or more fragments and substructures of the order-dependent representation and molecular graph representation, generating one or more nodes based on the substructures, generating one or more edges based on the fragments, and generating the graph based on the nodes and edges, which can be performed mentally aided by pen and paper as explained above with respect to claims 1 and 11. Dependent claims 4 and 16 further limit the mathematical concept of the RNN to include grated current units, which represent mathematical relationships that calculate an output vector based on an update grate vector, reset gate vector, and hidden state vector in light of Applicant’s specification at para. [0039]-[0040]. Dependent claim 5 further recites the mathematical concept of training a machine learning model to be based on the reproduced-order dependent representation or the order-dependent representation, which is a mathematical concept as explained above for claim 11. Dependent claims 6 and 17 further recite the mental process and mathematical concept of generating a molecular fingerprint based on analyzing the latent vector representation, which encompasses decoding the latent vector representation (mathematical calculations), and the mathematical concept of training the machine learning model using the molecular fingerprint. Dependent claim 7 further limits the abstract idea of claim 6 to generate a Morgan Fingerprint. Dependent claim 8 further recites the mental process and mathematical concept of determining statistical properties of the latent vector representation and training the machine learning model on the statistical properties. Dependent claims 9 and 12 further recite the mathematical concept of encoding the generated graph based on node aggregation functions of the GCN. Dependent claims 10 and 13 further limit the abstract idea of claim 1 of generating the latent vector representation to be an order independent representation. Therefore, claims 1-17 recite an abstract idea. [Step 2A, Prong 1: YES]
Step 2A: Prong 2: Under the MPEP § 2106.04, the Step 2A, Prong 2 analysis requires identifying whether there are any additional elements recited in the claim beyond the judicial exception(s), and evaluating those additional elements to determine whether they integrate the exception into a practical application of the exception. This judicial exception is not integrated into a practical application for the following reasons.
Claims 2-10 and 12-17 do not recite any elements in addition to the judicial exception.
The additional elements of claims 1 and 11 include:
synthesizing a new chemical compound that is related to the reproduced order-dependent representation of the chemical compound;
and testing the new chemical compound to determine a chemical property of the new chemical compound, a pharmacological property of the new chemical compound, or a combination thereof.
The additional elements of claim 11 further include:
one or more processors and one or more non-transitory computer-readable medium.
First, the additional elements of synthesizing a new chemical compound that is related to the reproduced order-dependent representation of the chemical compound and then testing the new chemical compound to determine a chemical or pharmacological property are not sufficient to integrate the recited judicial exception into a practical application because they amount to mere instructions to apply the exception for the following reasons. MPEP 2106.05(f) states The additional element of a processor and non-transitory computer readable medium are generic computer components. The courts have found the use of a computer or other machinery in its ordinary capacity for economic or other tasks (e.g., to receive, store, or transmit data) or simply adding a general purpose computer or computer components after the fact to an abstract idea (e.g., a fundamental economic practice or mathematical equation) does not integrate a judicial exception into a practical application. See MPEP 2106.05(f). In the instant case, the claims encompass synthesizing any “new chemical compound” that is somehow “related” to the reproduced order-dependent representation of the chemical compound determined by the abstract idea. However, the claim provides no restriction on how the result of identifying and synthesizing a new chemical compound from the order-dependent representation is accomplished, and similarly provides no restriction on what assay is being used to test the new chemical compound to determine any chemical or pharmacological property. Overall, these additional elements only recite the idea of a solution (identifying and testing a new chemical compound), but fail to recite any details to how this solution is accomplished.
Last, the additional element of a processor and non-transitory computer readable medium are generic computer components. The courts have found the use of a computer or other machinery in its ordinary capacity for economic or other tasks (e.g., to receive, store, or transmit data) or simply adding a general purpose computer or computer components after the fact to an abstract idea (e.g., a fundamental economic practice or mathematical equation) does not integrate a judicial exception into a practical application. See MPEP 2106.05(f).
Therefore, the additionally recited elements amount to mere instructions to apply the exception, and, as such, the claims as a whole do no integrate the abstract idea into practical application. Thus, claims 1-17 are directed to an abstract idea. [Step 2A, Prong 2: NO]
Step 2B: In the second step it is determined whether the claimed subject matter includes additional elements that amount to significantly more than the judicial exception. See MPEP § 2106.05.
The claims do not include any additional steps appended to the judicial exception that are sufficient to amount to significantly more than the judicial exception for the following reasons.
Claims 2-10 and 12-17 do not recite any elements in addition to the judicial exception. The additional elements of claims 1 and 11 are outlined above.
First, the additional element of a processor and non-transitory computer readable medium are conventional computer components. The courts have found the use of a computer or other machinery in its ordinary capacity for economic or other tasks (e.g., to receive, store, or transmit data) or simply adding a general purpose computer or computer components after the fact to an abstract idea (e.g., a fundamental economic practice or mathematical equation) does not provide significantly more. See Affinity Labs v. DirecTV, 838 F.3d 1253, 1262, 120 USPQ2d 1201, 1207 (Fed. Cir. 2016) (cellular telephone); TLI Communications LLC v. AV Auto, LLC, 823 F.3d 607, 613, 118 USPQ2d 1744, 1748 (Fed. Cir. 2016) (computer server and telephone unit). Therefore, the additional element is not sufficient to amount to significantly more than the judicial exception.
Furthermore, synthesizing a chemical compound and then testing the chemical compound for a chemical or pharmacological property, even in combination with a computer, are well-understood, routine, and conventional. This position is supported by Lin et al. (A Review on Applicants of Computational Methods in Drug Screening and Design, 2020 March, Molecules, 25(1375), pg. 1-17; newly cited). Lin revies the applications of computational methods in drug screening and design (Abstract), and discloses that with the rapid development of computer hardware, software, and algorithms, drug screening and design have benefited from computational methods which reduce the time and cost of drug development (pg. 1, para. 1). Lin discloses drug design involves designing a chemical compound, synthesizing the chemical compound, and then assessing the biological activities of the chemical compounds for further drug development (pg. 3, para. 3).
Therefore, taken alone, the additional elements do not amount to significantly more than the above-identified judicial exception(s). Even when viewed as a combination, the additional elements fail to transform the exception into a patent-eligible application of that exception. Thus, the claims as a whole do not amount to significantly more than the exception itself. [Step 2B: NO]
Therefore, the instantly rejected claims are not drawn to eligible subject matter as they are directed to an abstract idea without significantly more. For additional guidance, applicant is directed generally to applicant is directed generally to the MPEP § 2106.
Response to Arguments
Applicant's arguments filed 15 Dec. 2025 regarding 35 U.S.C. 101 have been fully considered but they are not persuasive.
Applicant further remarks that a human cannot practically or reasonably replicate the neural network system in the mind and perform the recited operations of synthesizing a new chemical compound and testing the new chemical compound, and a human cannot practically or reasonably decode latent vector representations based on hidden states of neural networks, and thus claims 1-20 do not recite a mental process (Applicant’s remarks at pg. 12, para. 2 to pg. 13, para. 3).
This argument is not persuasive. It is agreed that the steps of “synthesizing” and “testing” do not recite an abstract idea, including a mental process, and therefore these steps have been characterized as additional elements in the above rejection. Regarding Applicant’s argument that the human mind cannot replicate the neural network system or decode a latent vector representation based on hidden states of a neural network, the Office action did not characterize these steps as mental processes, and instead states these recited the abstract idea of a mathematical concept. In the step of “encoding the generated graph” of claims 1 and 11, only the alternative embodiment of “encoding the generated graph based on… one or more characteristics of the generated graph”, which does not require a neural network system or using hidden states of a neural network, was considered to recite a mental process. It is further noted the independent claims recite the encoding being based on a neural network in alternative form, so the independent claims only require encoding the generated graph based on one or more characteristics of the generated graph. Applicant does not present any particular arguments regarding why these steps do not recite a mathematical concept. Therefore, it is not persuasive the claims do not recite an abstract idea under Step 2A, Prong 1.
Applicant remarks that claims 1 and 11 include recitations that integrate the abstract idea into a practical application, including “decoding the latent vector representation…; synthesizing a new chemical compound…; and testing the new chemical compound”, which are directed to an improvement of the technical field of identifying new chemical compounds to be synthesized and tested (Applicant’s remarks at pg. 13, para. 4 to pg. 14, para. 4). Applicant further remarks this improvement is described at para. [0050]-[0052], explaining the chemical property prediction module reduces the amount of time for a medicinal chemist to modify a chemical compound, and enables medicinal chemists to select lead candidate series, explore chemical spaces, reduce failures rates of chemical compounds, and accelerate the drug discovery process (Applicant’s remarks at pg. 14, para. 5 to pg. 15, para. 1).
This argument is not persuasive. First, the step of decoding recite a mathematical concept, and thus is not an additional element. Regarding the steps of synthesizing and testing, the claims encompass synthesizing any “new chemical compound” that is somehow “related” to the reproduced order-dependent representation of the chemical compound determined by the abstract idea. However, the claim provides no restriction on how the result of identifying and synthesizing a new chemical compound from the order-dependent representation is accomplished, and similarly provides no restriction on what assay is being used to test the new chemical compound to determine any chemical or pharmacological property. Overall, these additional elements only recite the idea of a solution (identifying and testing a new chemical compound), but fail to recite any details to how this solution is accomplished. See MPEP 2106.05(f).
Furthermore, while Applicant alleges these additional elements improve drug discovery, computational drug discovery for drug development is already conventional in the field and reduces the time and cost of drug development, as discussed by Lin (newly cited above; Abstract, pg. 1, para. 1). Given the claims do not recite any details regarding how the generated reproduced order-dependent representation of the chemical compound is actually being applied to synthesize a new chemical compound and test the compound, it is not apparent that these additional elements improve drug development, alone or in combination with the judicial exception.
Applicant remarks the claimed technique is similar to the claims in BASCOM, where “the installation of a filtering tool…” provided an inventive concept for reciting a non-generic arrangement of the additional elements, and similarly decoding the latent vector representation based on a plurality of hidden states of a neural network to generate a reproduced order-dependent representation of the chemical compound, synthesizing the compound, and testing the compound are other than what is well-understood, routine, and conventional (Applicant’s remarks at pg. 17, para. 1-3).
This argument is not persuasive. First, the decoding step is part of the abstract idea, and therefore, the conventionality of the decoding step is not considered under step 2B. Furthermore, simply synthesizing a chemical compound and testing the chemical compound for a chemical property is well-understood, routine, and conventional, as discussed by Lin (newly cited) in the above rejection.
Claim Rejections - 35 USC § 102
The rejection of claims 1-2, 4-5, 9-14, 16, and 18-20 under 35 U.S.C. 102(a)(2) as being anticipated by Zavoronkovs (2020), as evidenced by Wu (2019) in the Office action mailed 17 Sept. 2025 has been withdrawn in view of claim amendments and cancellations received 15 Dec. 2025. However, after further consideration a new grounds of rejection is set forth below under 35 U.S.C. 103 in view of the claim amendments.
Claim Rejections - 35 USC § 103
The rejection of claims 3 and 15 under 35 U.S.C. 103 as being unpatentable over Zavoronkovs (2020) in view of Pysmiles (2020) in the Office action mailed 17 Sept. 2025 has been withdrawn in view of claim amendments received 15 Dec. 2025.
The rejection of claims 6-8 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Zavoronkovs (2020) in view of Kearnes (2019) in the Office action mailed 17 Sept. 2025 has been withdrawn in view of claim amendments received 15 Dec. 2025.
However, after further consideration new grounds of rejections are set forth below under 35 U.S.C. 103 in view of the claim amendments.
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.
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 1-2, 4-5, 9-14, and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Zavoronkovs (2020) over Reddy (2007), as evidenced by Wu (2019). This rejection is newly recited and necessitated by claim amendment.
Cited references:
Zavoronkovs et al., US 2023/0075100 A1, effectively filed 19 Feb. 2020 (previously cited);
Reddy et al., Virtual Screening in Drug Discovery- A Computational Perspective, Current Protein and Peptide Science, 2007, 8, pg. 329-351 (newly cited).
Wu et al., A Comprehensive Survey on Graph Neural Networks, 2019, arxiv, pg. 1-22; (previously cited).
Regarding claims 1 and 11, Zavoronkovs discloses a method, and system comprising a processor configured to perform the method (Abstract; [0023]-[0024]), of generating and using a Graph-to-Sequence (G2S) model to create chemical entities ([0037]-[0038]), wherein the method comprises the following steps:
Zavoronkovs discloses representing a chemical molecule as a simplified molecular-input line-entry system (SMILES) (i.e. an order-dependent representation of the chemical compound) and molecular graph of the chemical compound ([0038], e.g. chemical entities represented as sequence and graph data). Zavoronkovs discloses converting the SMILES representation into a graph structure (i.e. a graph of the chemical compound) ([0037]) (generating a graph of a chemical compound based on at least one of an order-dependent representation of the chemical compound and a molecular graph representation of the chemical compound).
Zavoronkovs discloses inputting the generated graph structure into an encoder, which processes the graph data (i.e. characteristics of the graph) to output latent vectors (i.e. a latent vector representation of the chemical compound) corresponding to the input graph data ([0037]; [0043]-[0044]; FIG. 1). (encoding the generated graph based on at least one of an adjacency matrix of a graph convolutional neural network (GCN), one or more characteristics of the generated graph, one or more activation functions of the GCN, and one or more weights of the GCN to generate a latent vector representation of the chemical compound).
Zavoronkovs discloses the latent vectors are input into a decoder, which processes the input data to obtain sequence data of the chemical compound ([0044]; FIG. 1, #104 and 116; [0009], e.g. decoder returns reconstructed input signals) (i.e. a reproduced order-dependent representation of the chemical compound). Zavoronkovs further discloses the decoder is a deep recurrent neural network that utilize hidden layers to predict outputs ([0003]; [0006]; [0091]; claim 19), demonstrating the decoding (i.e. output generation) is based on a plurality of hidden states of a recurrent neural network (decoding the latent vector representation based on a plurality of hidden states of a [recurrent] neural network (NN) to generate a reproduced order-dependent representation of the chemical compound).
Zavoronkovs discloses the autoencoder neural network generates new objects having a desired property using a trained graph to sequence model trained on the reconstructed sequence representation of the generated graph ([0021]-[0022]; [0052]-[0058], e.g. training; [0066], e.g. generating objects using trained model; FIG. 2-3), demonstrating a new compound related to the sequence data of the chemical compound (i.e. the reproduced order-dependent representation of the chemical compound) is identified. Zavoronkovs discloses the new object can be sequence representations of molecules ([0087] ;[0125]).
Zavoronkovs further discloses creating a real version of the new object, which is a physical object with properties ([0021]).
Zavoronkovs further discloses validating the new object to have the desired property after creating the physical object, wherein the desired property includes biochemical properties or structural properties (i.e. testing the new chemical compound to determine a chemical property of the new compound) ([0021]).
Further regarding claim 11¸ Zavoronkovs discloses training the autoenocoder, including the graph convolution neural network (i.e. the encoder) and recurrent neural network (i.e. the decoder) using the reconstructed input signals (claim 19; [0012]; [0043], e.g. encoder is GCN and decoder is recurrent NN; FIG. 2, e.g. training uses computed loss from reconstructed object). Zavoronkovs further discloses the autoencoder predicts properties of the chemical compounds (FIG. 2-3, e.g. object with desired properties based on input desired properties; [0067]-[0068]; [0076]; [0078], e.g. properties of generated objects calculated/predicted and used to train the model in the next iteration) (training the machine learning model based on the reproduced order-dependent representation, wherein the machine learning model includes the GCN and the RNN, and wherein the machine learning model is configured to predict one or more chemical properties of the chemical compound).
Further regarding claims 1 and 11, Zavoronkovs does not disclose the following:
While Zavoronkovs does generally disclose creating a physical version of the new compound that is related to the reproduced order-dependent representation of the chemical compound, Zavoronkovs does not explicitly disclose synthesizing the new chemical compound.
However, Reddy overviews virtual screening in drug discovery (Abstract), as performed in Zavoronkovs ([0004]-[0005]), and discloses an effective interplay between the experimental and computational approaches in drug discovery is important to guide experimentalists in synthesis and screening of compounds in a more rational way (pg. 329, col. 2, para. 1). Reddy discloses that a common solution to the problem of chemical spaces being too large to permit physical synthesis of all compounds is to design virtual libraries of compounds and apply appropriate selection techniques to screen the library for a smaller subset of compounds for physical synthesis and biological testing to identify active molecules (i.e. testing for a pharmacological property) (pg. 340, col. 1, para. 1-2). Reddy discloses efforts combining computational and experimental approaches are expected to yield compounds with improved potency and enhanced efficacy (pg. 329, col. 2, para. 3).
It would have been prima facie obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the method of Zavoronkovs to have created the physical new object by synthesizing the new chemical compound, as shown by Reddy, discussed above. One of ordinary skill in the art would have been motivated to combine the methods of Zavoronkovs and Reddy in order to create and identify lead compounds having improved potency and enhanced efficacy, as shown by Reddy (pg. 329, col. 2, para. 1-3). This modification would have had a reasonable expectation of success given Zavoronkovs discloses a neural network that generates chemical compounds with desired properties to guide drug optimization (Abstract; [0005]), which is a virtual screening technique applicable to the synthesis and testing of Reddy.
Further regarding the dependent claims:
Regarding claims 2 and 14, Zavoronkovs further discloses the decoder outputs reconstructed input signals in a sequence or string representation of a chemical entity, wherein the sequence is a SMILES sequence ([009]; [014]) (the reproduced order-dependent representation is a simplified molecular-input line-entry system (SMILES) string associated with the chemical compound).
Regarding claims 4 and 16, Zavoronkovs further discloses the neural network of the decoder can be a gated recurrent unit network or a long short-term memory (LSTM) unit ([0043]).
Regarding claim 5, Zavoronkovs further discloses training a neural network (i.e. machine learning model) based on a graph structure derived from a SMILES sequence of the chemical compound (i.e. the order-dependent representation) (Fig. 1; [0040], e.g. sequence data used in training).
Regarding claims 9 and 12¸ Zavoronkovs further discloses the encoding uses a graph convolution neural network (GCN)-like encoder ([0043]; [0097]), which multiplies the input values by each neuron’s (i.e. node) learnable weights to which some function is applied at each layer of multiple layers ([0006]).
Regarding claims 10 and 13¸ Zavoronkovs discloses the encoding generates a latent vector from a graph-structure with unordered nodes (i.e. an order-independent representation) using a graph neural network ([0022]; [0037]; [0043]), demonstrating the latent vector is order-independent.
Further regarding claims 9-10 and 12-13, while not explicitly discussed in Zavoronkovs, the functions of the GCN being “aggregation functions” and the latent vectors being unordered are inherent in Zavoronkovs through the use of a graph convolution neural network, as evidenced by Wu. Wu overviews Graph Neural networks, including graph convolution neural network (GCN) s as used in Zavoronkovs, (Abstract; pg. 1, col. 2, para. 2), and demonstrates graph convolution obtains a hidden representation of a node by analyzing features of neighboring nodes which are unordered and variable in size and encode these graphs into a latent space, and further discloses GCNs generate a nodes representation by aggregating its own features (Figure 1(b); pg. 4, col. 1, para. 2; pg. 16, col. .2, para. 7).
Therefore, the invention is prima facie obvious.
Claims 3 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Zavoronkovs in view of Reddy, as applied to claims 1 and 11 above, further in view of Pysmiles (2020). This rejection is newly recited and necessitated by claim amendment.
Cited references: Pysmiles: The lightweight and pure-python SMILES reader and writer, 2020, pg. 1-9 (previously cited).
Regarding claims 3 and 15, Zavoronkovs in view of Reddy disclose the method and system of claims 1 and 11 as applied above.
Further regarding claims 3 and 15, Zavoronkovs does not disclose the following:
Regarding claims 3 and 15, while Zavoronkovs discloses the graph data structure can be obtained through various techniques, Zavoronkovs does not disclose the technique of generating a graph structure by: identifying one or more fragments and one or more substructures of at least one of the order-dependent representation and the molecular graph representation; generating one or more nodes based on the one or more substructures; and generating one or more edges based on the one or more fragments.
However, Pysmiles discloses software for converting a SMILES string to a molecular graph (pg. 1, para. 1-3), which comprises identifying substructures, including aromatic systems, of the molecule and numbers of hydrogen atoms attached to various atoms (i.e. fragments) (pg. 2, para. 1-2). Pysmiles further discloses generating nodes corresponding to atoms of the substructures and edges representing bonds of the fragments (pg. 1, last para to pg. 2, para. 2; pg. 2, Reading Smiles section).
It would have been prima facie obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the molecular graph input of Zavoronkovs to have utilized molecular graphs generated by the Pysmiles software, as shown by Pysmiles (pg. 1-2), thus arriving at the invention of claims 3 and 14.One of ordinary skill in the art would have been motivated to combine the methods of Zavoronkovs and Pysmiles based on the simple substitution of the molecular graph generation in Zavoronkovs with the molecular graph generation in Pysmiles. One of ordinary skill in the art could have substituted the molecular graph generation of Zavoronkovs with the molecular graph generation of Pysmiles given both molecular graph generations result in a molecular graph representing atoms as nodes and bonds as edges (Pysmiles pg. 2; Zavoronkovs ([0037]-[0038]) and Zavoronkovs discloses graph structured data is known and can be obtained through various techniques. Thus the results of the substitution would have predictably resulted in a molecular graph generated according to Pysmiles and used as input into the autoencoder of Zavoronkovs.
Therefore, the invention is prima facie obvious.
Claims 6-8 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Zavoronkovs in view of Reddy, as applied to claims 5 and 11 above, further in view of Kearnes (2019). This rejection is newly recited and necessitated by claim amendment.
Cited references: Kearnes et al., Decoding Molecular Graph Embeddings with Reinforcement Learning, 2019, arXiv, pg. 1-14 (previously cited).
Regarding claims 6-8 and 17, Zavoronkovs in view of Reddy disclose the method and system of claims 5 and 11 as applied above.
Further regarding claims 6 and 17, Zavoronkovs further discloses translating the latent vectors into a 2D map representing biochemical and/or structural properties of chemical compounds corresponding to the latent vectors, using the properties ([0078]; [0118]-[0119]). Zavoronkovs discloses these properties are real object properties of the object (e.g. properties of the chemical compound) ([0052]; FIG. 2 #203), and that for molecule objects, the properties can be any chemical property from structural requirements to physical-chemical properties, including molecular descriptors ([0076]; [0090]).
Zavoronkovs further discloses training the neural network using selected chemical compounds having desired properties corresponding to a region in the 2D map containing chemical compounds with identified properties ([0018]; [0078], e.g. most promising area from 2D map in terms of objects used in training), and that training includes using a loss function and the latent vector representation ([0007]; [0043], e.g. encoder and decoder trained; FIG. 1).
Further regarding claim 8, Zavoronkovs further discloses calculating log-likelihoods (i.e. statistical properties) from the latent space ([0055]; FIG. 2).
Zavoronkovs further discloses training the neural network by performing gradient descent using the calculated log-likelihoods ([0055]; FIG. 2).
Zavoronkovs does not disclose the following:
Regarding claims 6-7 and 17, Zavoronkovs does not disclose the structural properties, including molecular descriptors, translated (i.e. generated) from the latent vectors and used to select chemical compounds with desired properties include a molecular fingerprint, wherein the molecule fingerprint is a Morgan fingerprint of the chemical compound.
However, regarding claims 6-7 and 17, Kearnes discloses a molecular graph based variational autoencoder (Abstract), which comprises a graph encoder that takes as input molecular graphs and encodes the graphs into a latent space and decoder that reconstructs a molecular graph using the latent space (pg. 1, col. 2, para. 5 to pg. 2 col. 2, para. 1; pg. 4, col. 1, para. 1 Fig. 1). Kearnes further discloses training the autoencoder uses a reward function that measures the similarity between morgan fingerprints of the input molecule and the current state of the generator (e.g. the generated molecule) (pg. 2, col. 2, para. 4 to pg. 3, col. 1, para. 1). Kearnes discloses the morgan fingerprints used to calculate the similarity are calculated based on the latent vector embeddings (e.g. morgan fingerprint decoded from perturbed embedding) (pg. 7, col. 1, para. 3 to col. 2, para. 1). Kearnes further discloses the decoder using this similarity metric guarantees a chemically valid output and optimizes for molecules relevant for drug discovery (pg. 1, col. 2, para. 3; pg. 2, col. 2, para. 5; pg. 4, col. 2, para. 3).
It would have been prima facie obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the method of Zavoronkovs in view of Reddy, as applied to claims 5 and 11 above, to have utilized a structural properties of a morgan fingerprint to generate chemical compounds with desired structural properties, as shown by Kearnes, discussed above. It would have been prima facie obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have combined the methods of Zavoronkovs in view of Reddy and Kearnes in order to generate chemically valid outputs relevant for drug discovery, as shown by Kearnes (pg. 4, col. 2, para. 3), rather than using structural properties like logP or QED that are only useful as baseline estimates, as shown by Kearnes (pg. 4, col. 2, para. 3). This modification would have had a reasonable expectation of success given both Zavoronkovs and Kearnes train an autoencoder that takes a molecular graph as input to generate a molecule and train the autoencoders on structural properties, such that the structural properties of a morgan fingerprint used in Kearnes is applicable to the method of Zavoronkovs.
Therefore, the invention is prima facie obvious.
Response to Arguments
Applicant's arguments filed 15 Dec. 2025 regarding 35 U.S.C. 102, as applicable to the 103 rejection above, have been fully considered but they are not persuasive.
Applicant remarks that Zavoronkovs does not teach synthesizing a new chemical compound related to the reproduced order-dependent representation of the chemical compound, or testing the new compound, and that the real version of the new object of Zavoronkovs is not synthesizing a new chemical compound that is related to the reproduced order-dependent representation, and furthermore, validating the new object to have the desired property of Zavoronkovs is not the same as testing the new chemical compound to determine a chemical property of the new compound (Applicant’s remarks at pg. 18, para. 2 to pg. 19, para. 1).
This argument is not persuasive. First, Zavoronkovs discloses creating a physical version of the new object (e.g. the new molecule) and then validating the biochemical or structural property of the physical molecule ([0021]); this clearly shows testing the new chemical compound to determine a chemical property of the new compound. Validating a predicted biochemical property reads on “determining a chemical property” given the actual chemical property of the physical molecule has yet to be determined and is only predicted in-silico.
With regard to the synthesis step, Zavoronkovs does not explicitly disclose synthesizing the new chemical compound, and instead discloses the new chemical compound is physical created (e.g. encompassing expressing a molecule in a cell). However, this limitation is taught by Reddy (newly cited) as applied in the new grounds of rejection above.
Zavoronkovs also discloses the “new object” is related to the reproduced order-dependent representation of the chemical compound. Zavoronkovs discloses the autoencoder neural network generates new objects having a desired property using a trained graph to sequence model trained on the reconstructed sequence representation of the generated graph ([0021]-[0022]; [0052]-[0058], e.g. training; [0066], e.g. generating objects using trained model; FIG. 2-3), demonstrating a new compound related to the sequence data of the chemical compound (i.e. the reproduced order-dependent representation of the chemical compound) is identified. Zavoronkovs discloses the new objects are determined based on a certain distribution of latent codes of physical objects ([0092]). In other words, the autoencoder of Zavoronkovs generates new synthetic data (i.e. new objects) resembling the training set; and thus the new objects of Zavoronkovs are related to the reproduced order-dependent representation of the chemical compound used to train the model.
Applicant remarks that dependent claims 2, 4-5, and 9-10 are patentable over the cited are based on their dependency to claim 1, and claim 11 and dependent claims 12-14, and 16 are patentable over the art for the reasons stated above for claim 1 (Applicant’s remarks at pg. 10, para. 2-4).
This argument is not persuasive for the same reasons discussed above for claim 1.
Double Patenting
The previous provisional rejections on the ground of nonstatutory double patenting as being unpatentable over claims 1-20 of copending Application No. App. 18/312,620 have been withdrawn in view of claim amendments and cancellations received 15 Dec. 2025. However, after further consideration a new grounds of rejection is set forth below in view of the amendments.
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).
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 1-3, 5, and 9-15 are provisionally rejected on the ground of nonstatutory double patenting as being unpatentable over claims 1-20 of copending Application No. App. 18/312,620 (reference application) in view of Reddy (2007). This rejection is newly recited and necessitated by claim amendment.
Cited Reference: Reddy et al., Virtual Screening in Drug Discovery- A Computational Perspective, Current Protein and Peptide Science, 2007, 8, pg. 329-351 (newly cited).
Regarding instant claims 1 and 11 reference claims 1, 14-15, and 17 disclose the limitations of instant claims 1 and 11. Reference claims 14-15 disclose the generation of a graph as input into a graph convolution neural network that uses an encoder to generate a latent vector space as instantly claimed. Reference claims 1 and 17 use the neural network to decode the latent space to produce a chemical compound.
Regarding instant claims 2 and 14¸ instant claim 1 only requires the input is one of an order-dependent representation and a molecular graph. Given reference claim 1 discloses the input can be a molecular graph representation, the embodiment in which the reproduced order dependent representation is used is not required. Thus instant claims 2 and 14, are disclosed by the reference claims as applied above.
Regarding instant claims 3 and 15, reference claim 16 discloses these steps for generating the molecular graph.
Regarding instant claim 5, reference claim 1 discloses training the neural network based on the order-dependent representation, and reference claims 6-7 disclose the neural network includes a graph convolution neural network.
Regarding instant claims 9 and 12, reference claim 7 discloses these limitations.
Regarding claims 10 and 13, reference claim 9 discloses this limitation.
The reference claims do not disclose the following limitations:
Further regarding instant claims 1 and 11, the reference claims do not disclose synthesizing the new chemical compound or testing the new chemical compound as claimed.
However, Reddy overviews virtual screening in drug discovery (Abstract), as performed in the reference claims, and discloses an effective interplay between the experimental and computational approaches in drug discovery is important to guide experimentalists in synthesis and screening of compounds in a more rational way (pg. 329, col. 2, para. 1). Reddy discloses that a common solution to the problem of chemical spaces being too large to permit physical synthesis of all compounds is to design virtual libraries of compounds and apply appropriate selection techniques to screen the library for a smaller subset of compounds for physical synthesis and biological testing to identify active molecules (i.e. testing for a pharmacological property) (pg. 340, col. 1, para. 1-2). Reddy discloses efforts combining computational and experimental approaches are expected to yield compounds with improved potency and enhanced efficacy (pg. 329, col. 2, para. 3).
It would have been prima facie obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the method of the reference claims to have created the physical new object by synthesizing the new chemical compound and tested the new chemical compound, as shown by Reddy, discussed above. One of ordinary skill in the art would have been motivated to combine the methods of the reference claims and Reddy in order to create and identify lead compounds having improved potency and enhanced efficacy, as shown by Reddy (pg. 329, col. 2, para. 1-3). This modification would have had a reasonable expectation of success given the reference claims use a neural network to identify a candidate chemical compound, which is a virtual screening technique applicable to the synthesis and testing of Reddy.
This is a provisional nonstatutory double patenting rejection because the patentably indistinct claims have not in fact been patented.
Claims 4, 6-8, and 16-17 are provisionally rejected on the ground of nonstatutory double patenting as being unpatentable over claims 1-20 of copending Application No. App. 18/312,620 in view of Reddy, as applied to claims 1, 5, and 11 above, further in view of Kearnes (2019). This rejection is newly recited and necessitated by claim amendment.
Cited reference: Kearnes et al., Decoding Molecular Graph Embeddings with Reinforcement Learning, 2019, arXiv, pg. 1-14 (previously cited).
Regarding instant claim 8, reference claim 11 discloses calculating gradient values (i.e. statistical properties) and convergence conditions based on the latent vectors in training the neural network.
Regarding claims 4, 6-7, and 16-17, the reference claims do not disclose these limitations.
However, regarding instant claims 4, 6-7, and 16-17, Kearnes discloses a molecular graph based variational autoencoder (Abstract), which includes a gated recurrent unit (pg. 1, col. 1, para. 2) and comprises a graph encoder that takes as input molecular graphs and encodes the graphs into a latent space and decoder that reconstructs a molecular graph using the latent space (pg. 1, col. 2, para. 5 to pg. 2 col. 2, para. 1; pg. 4, col. 1, para. 1 Fig. 1). Kearnes further discloses training the autoencoder uses a reward function that measures the similarity between morgan fingerprints of the input molecule and the current state of the generator (e.g. the generated molecule) (pg. 2, col. 2, para. 4 to pg. 3, col. 1, para. 1). Kearnes discloses the morgan fingerprints used to calculate the similarity are calculated based on the latent vector embeddings (e.g. morgan fingerprint decoded from perturbed embedding) (pg. 7, col. 1, para. 3 to col. 2, para. 1). Kearnes further discloses the decoder using this similarity metric guarantees a chemically valid output and optimizes for molecules relevant for drug discovery (pg. 1, col. 2, para. 3; pg. 2, col. 2, para. 5; pg. 4, col. 2, para. 3).
It would have been prima facie obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified the method of the reference claims to have utilized the neural network with structural properties of a morgan fingerprint to generate chemical compounds with desired structural properties, as shown by Kearnes, discussed above. It would have been prima facie obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have combined the methods of the reference claims and Kearnes in order to generate chemically valid outputs relevant for drug discovery, as shown by Kearnes (pg. 4, col. 2, para. 3), rather than using structural properties like logP or QED that are only useful as baseline estimates, as shown by Kearnes (pg. 4, col. 2, para. 3). This modification would have had a reasonable expectation of success given both the reference claims and Kearnes train an autoencoder that takes a molecular graph as input to generate a molecule and train the autoencoders on properties, such that the structural properties of a morgan fingerprint used in Kearnes is applicable to the method of the reference claims
This is a provisional nonstatutory double patenting rejection.
Response to Arguments
Applicant's arguments filed 15 Dec. 2025 regarding double patenting have been fully considered but they are not persuasive.
Applicant remarks claims 1-18 as presented render the double patenting rejections moot (Applicant’s remarks at pg. 21, para. 1).
This argument is not persuasive because they do not take into account the newly cited reference, Reddy.
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.
Inquiries
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KAITLYN L MINCHELLA whose telephone number is (571)272-6485. The examiner can normally be reached 7:00 - 4:00 M-Th.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Olivia Wise can be reached at (571) 272-2249. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/KAITLYN L MINCHELLA/Primary Examiner, Art Unit 1685