DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Applicant’s amendments and arguments, filed 5/19/2026 have been entered and carefully considered, but are not completely persuasive.
Claims 1, 5-13, 16-27 are pending in this application. Claims 5-13 and 16-24 stand withdrawn from consideration as being drawn to non-elected inventions or species. Claims 25-27 are newly added, and are generic to the elected invention. Claims 2-4, 14, 15 have been canceled.
Claims 1, 25-27 are under examination. Applicant elected Invention I (a DNA product), that is the product of species A (a DNA synthesis process), using evaluation by species B vii (using a biological product language), in the reply filed on 1/20/2026.
Currently, claims 1 and 25-27 are each generic to that election.
This application is a continuation of PCT/US2025/0318891, filed 6/2/2025, which claims
priority to two US provisional applications. This Application has TrackOne Status.
The claims have been heavily amended, and the earliest filed provisional providing support for the amended claims now appears to be 63/803471, filed 5/9/2025.
The rejections under 35 USC 102 of claims 1-4, 14-15 over a multiplicity of separate references have been withdrawn. New prior art rejections are made below in view of Applicant’s amendments.
Claim Interpretation
The claims in this application are given their broadest reasonable interpretation (BRI) using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art.
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, 25-26 is/are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea of mental steps, mathematic concepts, organizing human activity, or a natural law without significantly more.
The claims have been heavily amended; however, the basis of the analysis remains the same.
Applicant is directed to MPEP 2106 for the most current and complete guidelines in the analysis of patent- eligible subject matter. The current MPEP is the primary source for the USPTO’s patent eligibility guidance.
With respect to step (1): YES, the claims are drawn to statutory categories: Computer-implemented processes.
With respect to step (2A) (1): YES, the claims recite an abstract idea, law of nature and/or natural phenomenon. The claims explicitly recite elements that, individually and in combination, constitute one or more judicial exceptions (JE).
Mathematic concepts, Mental Processes or Elements in Addition (EIA) in the claim(s) include:
1. (Currently Amended) A method performed by one or more computers for computationally designing a modified molecule that maintains at least a first feature of a first molecule while also achieving a second feature of a second molecule, the method comprising:
(Preamble and EIA: setting forth a method, and the goal of the method, using a general-purpose computer, an additional element MPEP 2106.05(a)).
selecting the first molecule having the first feature;
selecting the second molecule having the second feature; and
(EIA- Data gathering steps, of selecting representations of unrestricted molecules based on the presence of unrestricted features. MPEP 2106.05(g)).
selecting the modified molecule by performing operations comprising:
generating a set of candidate molecules based at least in part on the first molecule and the second molecule;
(Mental Process, computationally combining the selected molecules in various, unrestricted ways, by observing each molecule, the function, and creating combinations. MPEP 2106.04(a)(2) Section III).
processing each of the first molecule, the second molecule, and each candidate molecule in the set of candidate molecules using an embedding neural network to generate a respective embedding of each of the first molecule, the second molecule, and each candidate molecule in the set of candidate molecules in an embedding space,
wherein the embedding neural network comprises at least one embedding layer defined by a set of neural network parameters and has been configured, through machine learning training, to generate feature-aware embeddings that capture semantic and functional attributes of input molecules;
(Mathematic concept, mental process and EIA: embedding is a mathematic transformation of data. The trained “embedding neural network” is an additional element, which acts “to generate feature-aware” embeddings, which capture the semantic and functional attributes, which are identified by a mental process of observation. MPEP 2106.04(a)(2) 1, III)
generating a ranking of the set of candidate molecules using Pareto front optimization based on, for each candidate molecule:
(i) a classification of whether the embedding of the candidate molecule is within a threshold conservation distance of the embedding of the first molecule in the embedding space to maintain the first feature of the first molecule, and
(ii) a distance between the embedding of the candidate molecule and the embedding of the second molecule to design towards the second feature of the second molecule; and
(Mathematic concept of ranking, using a type of calculating, Pareto Front optimization. MPEP 2106.04(a)(2) section I)
selecting one or more of the candidate molecules based on the ranking of the candidate molecules.
(Mental Process- of observation and selection of a result, based on the calculated rank. MPEP 2106.04(a)(2) section III.)
25. (New) The method of claim 1, wherein calculation of embeddings and embedding distances is parallelized across a plurality of AI processing cores, wherein the AI processing cores comprise at least one of a graphics processing unit (GPU), a neural processing unit (NPU) or a field-programmable gate array (FPGA) configured with custom instruction sets for molecular analysis.
(EIA- additional element related to the structure of the general-purpose computer MPEP 2106.05(a, b))
26. (New) The method of claim 1, further comprising: selecting one or more of the candidate molecules for physical synthesis based on the ranking of the set of candidate molecules.
(Mental process of observing the results of claim 1, and making a judgement of what should be selected to be synthesized. The physical synthesis is an intended use of the selected molecule data. MPEP 2106.04(a)(2)(III).)
With respect to step 2A (2): NO, the claims do not integrate the JE into a practical application (MPEP 2106.04(d)):
“Examiners evaluate integration into a practical application by: (1) identifying whether there are any additional elements recited in the claim beyond the judicial exception(s); and (2) evaluating those additional elements individually and in combination to determine whether they integrate the exception into a practical application, using one or more of the considerations introduced in subsection I supra, and discussed in more detail in MPEP §§ 2106.04(d)(1), 2106.04(d)(2), 2106.05(a) through (c) and 2106.05(e) through (h).”
Claim(s) 1 recite(s) the additional non-abstract element(s) of data gathering, or a description of the data gathered.
Data gathering steps are not an abstract idea, they are extra-solution activity, as they collect the data necessary to carry out the JE. MPEP 2106.05(g).
The data gathering does not impose any meaningful limitation on the JE, or how the JE is performed. MPEP 2106.05(g).
The data gathering steps constitute a general link to a technological environment. (MPEP 2106.05(h), citing Mayo, Bilski, electric Power Group, Genetic Techs Ltd v Merial LLC.)
The additional limitation (data gathering) must have more than a nominal or insignificant relationship to the identified judicial exception to provide integration into a practical application. (MPEP 2106.05(g) citing Mayo, PerkinElmer, Inc. v. Interna Ltd, Intellectual Ventures LLC v. Erie Indem. Co., Electric Power Group LLC v. Alstom S.A.).
Claim(s) 1, 25 recite the additional non-abstract element (EIA) of a general-purpose computer system or parts thereof.
The claims do not provide any details of how specific structures of the computer elements are used to implement the JE. MPEP 2106.05(a), contrasting decisions identifying how the computer implements an abstract idea, such as in McRo to decisions which found no specific interaction with the computer, such as in Affinity Labs of Tex v. DirecTV, LLC.
The computer elements of the claims do not provide improvements to the functioning of the computer itself. MPEP 2106.05(a) I, contrasting decisions indicating an improvement to the computer, such as DDR Holdings, LLC v. Hotels.com LP, with decisions that did not identify an improvement to the computer, such as FairWarning IP, LLC v. Iatrix Sys.
The computer elements of the claims do not provide improvements to any other technology or technical field. MPEP 2106.05(a) II: contrasting decisions indicating an improvement to the technology, such as Diamond v. Diehr, Trading Techs. Int’l v. CQG Inc, or Intellectual Ventures I v. Symantec Corp, with decisions that did not identify an improvement to the technology, such as Alice Corp, Versata Dev. Group, Inc. v. SAP AM. Inc, or TLI Communications.
The computer elements of the claims do not utilize a particular machine. MPEP 2106.05(b): contrasting decisions wherein a particular machine was identified, such as MacKay Radio & Tel. Co. v. Radio Corp. of America, Eibel Process Co. v. Minn. & Ont. Paper Co., with decisions where a general-purpose computer does not qualify as a particular machine, such as Ultramercial, Inc. v. Hulu, LLC, TLI communications, or Eon Corp. IP holdings LLC v. AT&T Mobility LLC.
Hence, these are mere instructions to apply the JE using a computer, and therefore the claim does not recite integrate that JE into a practical application.
Dependent claim(s) 26 recite(s) an abstract limitation to the JE reciting additional mathematic concepts, or mental processes. Additional abstract limitations cannot provide a practical application of the JE as they are a part of that JE.
In combination, the limitations of data gathering, for the purpose of carrying out the JE, using a general-purpose computer merely provide extra-solution activity, and fail to integrate the JE into a practical application.
With respect to step 2B: NO, the claims do not recite a specific inventive concept. The judicial exception alone cannot provide that inventive concept or practical application (MPEP 2106.05).
“… an "inventive concept" is furnished by an element or combination of elements that is recited in the claim in addition to (beyond) the judicial exception, and is sufficient to ensure that the claim, as a whole, amounts to significantly more than the judicial exception itself. Alice Corp…”
With respect to claim(s) 1: The limitation(s) identified above as non-abstract elements (EIA) related to data gathering do not rise to the level of significantly more than the judicial exception.
The data gathered in claim 1 includes first/second molecules, having a first/second feature:
Dahiya (2023) obtains DNA molecules representations, each having differing features, in a method to generate a sequence containing both features. “The platform transforms the conventional paradigms in synthetic biology by enabling the context-sensitive and host-specific engineering of 5′ regulatory elements—promoters and 5′untranslated regions (UTRs) along with an array of codon-optimised coding sequence (CDS) variants. This allows us to generate context-sensitive 5′regulatory sequences and CDSs, achieving an unparalleled level of specificity and adaptability in different target hosts.” (abstract); Dahiya et al (2023) From Context to Code: Rational De Novo DNA Design
and Predicting Cross-Species DNA Functionality Using Deep Learning Transformer Models. BioRxiv, 10/15/2023. 19 pages.
Brereton (2020) obtains drug molecule representations each having a first or second feature, in a method of embedding, and predicting drug properties of generated drug molecules. “POEM’s predictive strength is obtained by combining multiple different representations of molecular structures in a context-specific manner, while maintaining low dimensionality.” (Abstract) Brereton et al. (2020) Predicting drug properties with parameter-free machine learning: pareto-optimal embedded modeling (POEM). Machine Learning Science and Technology, vol 1: 025008, 30 pages.
Wang (2024) obtains molecular representations for differing chemicals, each having differing features, in a method of generating new structures having predictable properties. “The algorithm first extracts an update set from the sampled molecules through the designed aggregation-based molecular clustering. Then, the final reward is computed by constructing the Pareto frontier ranking of the molecules from the updated set.” (abstract) Wang et al. (2024) Multi-objective molecular generation vial clustered Pareto-based reinforcement learning. Neural Networks, V 179: 106596, 15 pages.
Makowski (2022) obtains biological molecule representations, each having differing features, in methods of optimizing antibody affinity and specificity. “Here we evaluate the use of machine learning to simplify antibody co-optimization for a clinical-stage antibody (emibetuzumab) that displays high levels of both on-target (antigen) and off-target (non-specific) binding.” (Abstract) Makowski, E. et al. (2022) Co-optimization of therapeutic antibody affinity and specificity using machine learning models that generalize to novel mutational space. Nature Communications, vol 13: 3788, 14 pages and some supplemental material.
Liu (2025) obtains molecules, each of which having differing features or functions, to generate a new molecule having aspects of both features. Liu exemplifies their method on polypeptides having anti-fungal features. P19012. Liu et al. (2025) Multi-objective molecular design through learning latent pareto set. The Thirty-ninth AAAI conference on artificial intelligence (AAAI-25), p19006-19014.
Hong (2024) obtains molecules, each having differing features, in a method of protein sequence design, through multiobjective optimization. At least three protein sequences are obtained, each having differing higher-dimensional design problems. P3. Hong, L. et al. (2024) An integrative approach to protein sequence design through multiobjective optimization. PLOS Computational Biology, Vol 20 no 7: e1011953, 37 pages.
These elements meet the BRI of the identified data gathering limitations. As such, the prior art recognizes that this data gathering element is routine, well understood and conventional in the art. MPEP 2106.05(d): “If, however, the additional element (or combination of elements) is no more than well-understood, routine, conventional activities previously known to the industry, which is recited at a high level of generality, then this consideration does not favor eligibility.”
Data gathering steps are not an abstract idea, they are extra-solution activity, as they collect the data necessary to carry out the JE. MPEP 2106.05(g).
The data gathering does not impose any meaningful limitation on the JE, or how the JE is performed. MPEP 2106.05(g).
The additional limitation (data gathering) must have more than a nominal or insignificant relationship to the identified judicial exception to provide an inventive concept. (MPEP 2106.05(g) citing Mayo, PerkinElmer, Inc. v. Interna Ltd, Intellectual Ventures LLC v. Erie Indem. Co., Electric Power Group LLC v. Alstom S.A.)
The data gathering steps constitute a general link to a technological environment: the trait prediction methods are intended to be applied to plant populations. (MPEP 2106.05(h), citing Mayo, Bilski, electric Power Group, Genetic Techs Ltd v Merial LLC.)
Therefore, simply appending well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception are insufficient to provide significantly more (as discussed in Alice Corp.,).
With respect to claim(s) 1, 25: the limitations identified above as non-abstract elements (EIA) related to general-purpose computer systems do not rise to the level of significantly more than the judicial exception.
Claim 1 requires a general-purpose computer system. Claim 25 adds GPU, FPGA and NPU processors to the computer system.
Feala (US 2022/0122692) employs computer systems, which can comprise GPU, in methods of embedding polypeptide sequences each of which having differing features or functions. [0078]
Ragnathan (US 2022/0348903) employs computer systems, which can comprise FPGA, in methods of evolutionary data-driven design of proteins and other sequence designed molecules. [0266]
Abeliuk (US 2021/0257049) employs computer systems, which can comprise GPU, in methods of optimizing a phenotype, with a combination of generative and predictive models.
Guturu (US 2023/0253113) employs computer systems, which can comprise GPU, in methods of creating biomolecule embeddings. Polypeptides are a preferred embodiment.
Each of the above of disclose computer systems or computing elements which meet the BRI of the claimed computer system or computer system elements, comprising input, output/ display, a processor, and memory.
As such, the prior art recognizes that these computing elements are routine, well understood and conventional in the art.
The claims do not provide any details of how specific structures of the computer elements are used to implement the JE. MPEP 2106.05(a), contrasting decisions identifying how the computer implements an abstract idea, such as in McRo to decisions which found no specific interaction with the computer, such as in Affinity Labs of Tex v. DirecTV, LLC.
The computer elements of the claims do not provide improvements to the functioning of the computer itself. MPEP 2106.05(a) I, contrasting decisions indicating an improvement to the computer, such as DDR Holdings, LLC v. Hotels.com LP, with decisions that did not identify an improvement to the computer, such as FairWarning IP, LLC v. Iatrix Sys.
The computer elements of the claims do not provide improvements to any other technology or technical field. MPEP 2106.05(a) II: contrasting decisions indicating an improvement to the technology, such as Diamond v. Diehr, Trading Techs. Int’l v. CQG Inc, or Intellectual Ventures I v. Symantec Corp, with decisions that did not identify an improvement to the technology, such as Alice Corp, Versata Dev. Group, Inc. v. SAP AM. Inc, or TLI Communications.
The computer elements of the claims do not utilize a particular machine. MPEP 2106.05(b): contrasting decisions wherein a particular machine was identified, such as MacKay Radio & Tel. Co. v. Radio Corp. of America, Eibel Process Co. v. Minn. & Ont. Paper Co., with decisions where a general-purpose computer does not qualify as a particular machine, such as Ultramercial, Inc. v. Hulu, LLC, TLI communications, or Eon Corp. IP holdings LLC v. AT&T Mobility LLC.
Hence, these are mere instructions to apply the JE using a computer, and therefore the claim does not provide significantly more.
Dependent claim(s) 26 recite(s) a limitation requiring additional mathematic concepts or mental processes. Additional abstract limitations cannot provide significantly more than the JE as they are a part of that JE (MPEP 2106.05).
In combination, the data gathering steps providing the information required to be acted upon by the JE, performed in a generic computer or generic computing environment fail to rise to the level of significantly more than that JE. The data gathering steps provide the data for the JE, which is carried out by the general-purpose computers. No non-routine step or element has clearly been identified.
The claims have all been examined to identify the presence of one or more judicial exceptions. Each additional limitation in the claims has been addressed, alone and in combination, to determine whether the additional limitations integrate the judicial exception into a practical application. Each additional limitation in the claims has been addressed, alone and in combination, to determine whether those additional limitations provide an inventive concept which provides significantly more than those exceptions. For these reasons, the claims, when the limitations are considered individually and as a whole, are rejected under 35 USC § 101 as being directed to non-statutory subject matter.
Applicant’s Arguments:
The Examiner notes, that claim 27 was not rejected under this statute. Should claim 1 be amended to positively recite the physical synthesis of the selected compound predicted to have both desired features, this rejection would be overcome.
Applicant’s amendments and arguments with respect to this rejection have been carefully considered, but are not persuasive.
The newly added “embedding neural network” of claim 1 generally links the abstract idea to the technological environment of embedding neural networks, which necessarily comprise an embedding layer. The recitation “configured, through machine learning training, to generate feature-aware embeddings that capture semantic and functional attributes of input molecules” fails to set forth how the neural network works with the data at hand to achieve this goal, nor does it provide additional structural data (layers, etc) to the embedding neural network itself. (MPEP 2106.05(f), (h)).
With respect to claim 25, adding GPU, NPU or FPGA parallel processing to the method, this generally links the abstract idea to the technology of parallel processing. The recitation in claim 25 that the calculation of embeddings and embedding distances are parallelized, does not clearly set forth how this is achieved, nor how this provides an improvement to the technology of processing embeddings. (MPEP 2106.05(a), (h))
Ex parte Desjardins had claims drawn to the use of a machine learning model, trained on one task with a first set of data, and set parameter weights, then trained again using differing data on a different task, adjusting parameters and weights, while protecting performance of the first task. Further, in Desjardins, the retraining of the particular ML changed the structure of that ML in a way that provided "'[a]n improvement in the functioning of a computer, or an improvement to other technology or technical field,' as discussed in MPEP §§ 2106.04(d)(l) and 2106.05(a)":
"By training the same machine learning model on multiple tasks as described in this specification, once the model has been trained, the model can be used for each of the multiple tasks with an acceptable level of performance. As a result, systems that need to be able to achieve acceptable performance on multiple tasks can do so while using less of their storage capacity and having reduced system complexity. For example, by maintaining a single instance of a model rather than multiple different instances of a model each having different parameter values, only one set of parameters needs to be stored rather than multiple different parameter sets, reducing the amount of storage space required while maintaining acceptable performance on each task. In addition, by training the model on a new task by adjusting values of parameters of the model to optimize an objective function that depends in part on how important the parameters are to previously learned task(s), the model can effectively learn new tasks in succession whilst protecting knowledge about previous tasks." Ex parte Desjardins, p3, quoting the specification.
The independent claim in Ex parte Desjardins contained specific limitations as to how at least some aspects of the asserted improvements are achieved:
"When evaluating the claim as a whole, we discern at least the following limitation of independent claim 1 that reflects the improvement: "adjust the first values of the plurality of parameters to optimize performance of the machine learning model on the second machine learning task while protecting performance of the machine learning model on the first machine learning task." We are persuaded that constitutes an improvement to how the machine learning model itself operates, and not, for example, the identified mathematical calculation." Ex parte Desjardins, p9
In contrast, the claims do not clearly set forth the link between the data gathered, the initial training of the ML, the structure of the ML, and how training or retraining affects the structure to obtain the desired results or asserted improvement.
Arguments that the claims provide an improvement in the computer itself are not persuasive, as no changes to the computer itself are provided, nor is there any specific interaction between specific elements of the computer or processor with the gathered data.
MPEP 2106: “To show that the involvement of a computer assists in improving the technology, the claims must recite the details regarding how a computer aids the method, the extent to which the computer aids the method, or the significance of a computer to the performance of the method. Merely adding generic computer components to perform the method is not sufficient. Thus, the claim must include more than mere instructions to perform the method on a generic component or machinery to qualify as an improvement to an existing technology. See MPEP § 2106.05(f)”
MPEP: 2106.05: “The recitation of claim limitations that attempt to cover any solution to an identified problem with no restriction on how the result is accomplished and no description of the mechanism for accomplishing the result, does not integrate a judicial exception into a practical application or provide significantly more because this type of recitation is equivalent to the words "apply it". See Electric Power Group, LLC v. Alstom, S.A., 830 F.3d 1350, 1356, 119 USPQ2d 1739, 1743-44 (Fed. Cir. 2016); Intellectual Ventures I v. Symantec, 838 F.3d 1307, 1327, 120 USPQ2d 1353, 1366 (Fed. Cir. 2016); Internet Patents Corp. v. Active Network, Inc., 790 F.3d 1343, 1348, 115 USPQ2d 1414, 1417 (Fed. Cir. 2015). In contrast, claiming a particular solution to a problem or a particular way to achieve a desired outcome may integrate the judicial exception into a practical application or provide significantly more. See Electric Power, 830 F.3d at 1356, 119 USPQ2d at 1743.”
Further, with respect to the arguments regarding the alleged improvement, it is unclear that the independent claims recite all the necessary and sufficient steps required to achieve that improvement. MPEP 2106.05(a): “An important consideration in determining whether a claim improves technology is the extent to which the claim covers a particular solution to a problem or a particular way to achieve a desired outcome, as opposed to merely claiming the idea of a solution or outcome. McRO, 837 F.3d at 1314-15, 120 USPQ2d at 1102- 03; DDR Holdings, 773F.3d at 1259, 113 USPQ2d at 1107
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1, 25-27 are rejected on the basis that it contains an improper Markush grouping of alternatives. See In re Harnisch, 631 F.2d 716, 721-22 (CCPA 1980) and Ex parte Hozumi, 3 USPQ2d 1059, 1060 (Bd. Pat. App. & Int. 1984). A Markush grouping is proper if the alternatives defined by the Markush group (i.e., alternatives from which a selection is to be made in the context of a combination or process, or alternative chemical compounds as a whole) share a “single structural similarity” and a common use. A Markush grouping meets these requirements in two situations. First, a Markush grouping is proper if the alternatives are all members of the same recognized physical or chemical class or the same art-recognized class, and are disclosed in the specification or known in the art to be functionally equivalent and have a common use. Second, where a Markush grouping describes alternative chemical compounds, whether by words or chemical formulas, and the alternatives do not belong to a recognized class as set forth above, the members of the Markush grouping may be considered to share a “single structural similarity” and common use where the alternatives share both a substantial structural feature and a common use that flows from the substantial structural feature. See MPEP § 2117.
The Markush grouping of “candidate molecules” as the product of claim 1, is improper because the alternatives defined by the Markush grouping do not share both a single structural similarity and a common use for the following reasons: Claim 1 is completely unlimited as to the nature of the molecule. The molecules which fall within the scope of claim 1 do not share any single structural similarity or common use. The molecules are not specific chemical structures, they are generic encompassing ANY type of molecule. The listed products are not alternative chemical compounds, that share any single structural similarity and common use. The listed products are not functionally equivalent. Specification [1487] enumerates possible molecules to be chosen, or generated: protein, enzyme, DNA, RNA, plasmid, metabolite, and biologic strain.
Each encompassed protein (within the generic types of “enzyme” or “non-enzyme”) has a differing primary structure, differing secondary structure, and differing biological and biochemical characteristics, not shared by any other product.
Each encompassed DNA sequence product has a differing primary structure, differing secondary structure, and differing biochemical/ biological / functional characteristics not shared by any other product.
Each encompassed RNA sequence product has a differing primary and secondary structures, and differing biochemical/ biological/ functional characteristics not shared by any other product.
Each encompassed plasmid has a differing primary structure, and differing biochemical/ biological/ functional characteristics and features not shared by any other product.
Each encompassed metabolite has a differing chemical structure and differing set of biologic or chemical or functional characteristics not shared by any other product.
Each encompassed “biologic strain” has differing physical and chemical structures, differing underlying genomic sequences, as well as differing biochemical/ biological/ functional features or characteristics not shared by any other product.
The list of types of products do not list any specific product having a specific structure, or substructure or specific characteristic that could be shared. None of the types of products are functional equivalents. No specific feature or function is claimed or identified. Each identified type of selected product molecule has a multitude of differing uses, and no specific common use has been claimed.
The Markush grouping of “first/ second molecule” (replacing “parent”) is improper because the alternatives defined by the Markush grouping do not share both a single structural similarity and a common use for the following reasons: The first and second molecules of claim 1, substitute the term: “molecule” for the previously pending “parent” and are set forth in the specification as encompassing “a parent DNA or RNA sequence” [1487], “a parent protein” [1487], “a parent cell line” [1487], “a parent strain of a microbe” [1487], “Other examples of the first (or second) biologic parent include an enzyme protein, a non-enzyme protein, a DNA sequence, an RNA sequence, a plasmid, a metabolite, a biologic strain… a metabolite…” [multiple places], “The biologic parent may include… a metabolic precursor, a cell or cell line, a strain of a biological species, or the like. The biologic parent may be or may include a material that is similar to the biologic product, and that is to be modified to generate the biologic product” [1510]. Each of these possible biologic parent molecules encompassed by the claim has separate and distinct chemical structures, biological sequences, biochemical properties, biological functions, biological features, and a plethora of differing uses. The biologic parent molecules encompassed by the claim share no common structure or substructure. The biologic parent molecules encompassed by the claim share no common specific use that depends on a shared or common structure. Claim 1 is completely unlimited as to the nature of the first/ second molecule. The possible parent molecules encompassed by the claim are not functional equivalents. The biologic parent molecules encompassed by the claim do not share any single structural similarity or common use. The biologic parent molecules encompassed by the claim are not specific chemical structures, they are generic. The biologic parent molecules encompassed by the claim are not alternative chemical compounds, that share any single structural similarity and common use.
To overcome this rejection, Applicant may set forth each alternative (or grouping of patentably indistinct alternatives) within an improper Markush grouping in a series of independent or dependent claims and/or present convincing arguments that the group members recited in the alternative within a single claim in fact share a single structural similarity as well as a common use.
Applicant’s Arguments
Applicant’s arguments are not persuasive.
As set forth by MPEP 2117:
“The improper Markush grouping rejection of the claim should be maintained until (1) the claim is amended such that the Markush grouping includes only members that share a single structural similarity and a common use; or (2) the applicant presents convincing arguments why the members of the Markush grouping share a single structural similarity and common use (i.e., are members of a physical, chemical, or art-recognized class that share a common use, or are chemical compounds that share a substantial structural feature that is essential to the common use).”
The claims have been broadened by amendment, from “biologic parent” and “biologic product” to “molecule” and “candidate molecule.” These terms encompass nearly any type of small molecule, metabolite, chemical, drug, DNA, RNA, protein, macromolecule, compound or composition. These do not chare a single structural similarity and a common use, nor have convincing arguments been provided.
MPEP 2117: “Note that no Markush claim can be allowed until any improper Markush grouping rejection has been overcome or withdrawn, and all other conditions of patentability have been satisfied.”
Claims 1, 25-27 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
The metes and bounds of claim 1 remain unclear. The claim fails to particularly point out and distinctly claim the particular molecules to be selected, the nature of any feature to be selected and fail to particularly point out and distinctly claim how the actual selections, and combinations are to be performed. Claim 1 sets forth the selection of two “parent” molecules, each of which has an unnamed feature, and then performs an evaluation of combinations of the parents. There is no particular basis for selection of the molecule, or the related features. Any molecule can have an astounding number of possible associated features, from molecular weight, atomic structure, chemical properties, biological properties, biological sequence information, names, abbreviations, pH, solubility, stability… Without specificity to either the overall goal features desired, or the molecules to be modified or used in the combination, it is entirely unclear how to select two molecules and generate combinations which retain the unexplained, unidentified features. The two molecules do not have to be the same type of molecule, or be compatible for combination. One of skill would not be able to determine the metes and bounds of the desired “candidate molecule” or the metes and bounds of the “molecule, having a… feature” as it is entirely unclear how to determine what a selected molecule could be, which would lead to the selection of the desired product, what features are to be observed or selected from that indefinite molecule, or how to evaluate any product for a combination of features. Even within the election of a DNA sequence product and the disclosures quoted above, the concept of a “selecting a DNA molecule having a … feature” is indefinite: DNA sequences exist in a multitude of tangible, or data-related forms, with a variety of parameters, associated information, physio-chemical attributes, and biochemical activities. If the elected desired outcome is a “Candidate DNA sequence product”, what is the “molecule having a … feature”? Is it a) sequence data, b) a cell line, c) a plasmid, d) a cell, e) an organism, f) a chromosome, g) isolated polynucleotides, primers, probes? These do not appear to be functional equivalents, nor do they clearly have compatible features for combination. It is entirely unclear how to select a “molecule having a first/ second feature” when no particular desired features are identified.
The claim is not limited to selecting a first/ second “parent DNA sequence”. The claim is not limited to evaluating specific combinations of DNA sequences and features from each parent using a biological language model. The claim does not direct the synthesis of a selected candidate DNA sequence product.
The nature of the evaluation is not set forth in claims 1 and 25-27 such that the point of the selection could be carried out. The aspect of the combination to be evaluated in claims 1 and 25-27 is not identified. It is entirely unclear how to “generate” the candidate molecules, even if limited to the elected DNA sequence product, when the features of the molecules, the overall desired features, and the aspect of the evaluation is completely generic. The preamble states that the goal is “designing a modified molecule”, however it is not clear that any specific molecule is modified in the “generating” process. The claim does not clearly take molecule A, and modify it with some aspect of Molecule B, to “generate” the set of candidates. It is entirely unclear how to create the desired combinations of the (undefined) molecules, in order to then embed and analyze those (undefined) combinations by the neural network and optimization steps. Even within the elected DNA products, it is entirely unclear how Applicant intends to “generate” the set of candidate DNA molecules. It is unclear whether this is a genetic algorithm, randomly breaking and combining two DNA sequence datasets, a set of in vitro plasmid manipulation steps that generate a physical combined product, a directed set of instructions to manipulate sequence data, et al. As the molecules do not have to be of the same type or class, it is entirely unclear how to combine elements from differing classes of molecules, to achieve the goal of the preamble. One of skill would not know how to create or select one combination over another, without a guiding principle, underlying scientific goal, or particular desired function.
Claims 1, and 25-27 do not achieve “designing a modified molecule that maintains at least a first feature of a first molecule while also achieving a second feature of a second molecule” as the selected candidate model is not clearly tested or analyzed for the presence or activity of each feature. The combinations are merely “based at least in part on the first and second molecule” without regard to the features., or what parts of the molecule are related to the features desired. The selected candidate molecule is based on data transformations, and rankings by the neural network and the optimization steps, and do not assess the presence or activity of any features in the selected candidate.
While claims are read in light of the specification, limitations from the specification cannot be read into the claims. MPEP 2111.
"Though understanding the claim language may be aided by explanations contained in the written description, it is important not to import into a claim limitations that are not part of the claim. For example, a particular embodiment appearing in the written description may not be read into a claim when the claim language is broader than the embodiment." Superguide Corp. v. DirecTV Enterprises, Inc., 358 F.3d 870, 875, 69 USPQ2d 1865, 1868 (Fed. Cir. 2004).”
The following is a quotation of 35 U.S.C. 112(d):
(d) REFERENCE IN DEPENDENT FORMS.—Subject to subsection (e), a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
The following is a quotation of pre-AIA 35 U.S.C. 112, fourth paragraph:
Subject to the following paragraph [i.e., the fifth paragraph of pre-AIA 35 U.S.C. 112], a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
Claim 26 is rejected under 35 U.S.C. 112(d) or pre-AIA 35 U.S.C. 112, 4th paragraph, as being of improper dependent form for failing to further limit the subject matter of the claim upon which it depends, or for failing to include all the limitations of the claim upon which it depends.
Claim 26 recites the same limitation that ends claim 1 “selecting one or more of the candidate molecules … based on the ranking…” the difference being in the intended use of the selected product “for physical synthesis”. However, claim 26 does not carry out the intended use, and this claim fails to further limit claim 1.
Applicant may cancel the claim(s), amend the claim(s) to place the claim(s) in proper dependent form, rewrite the claim(s) in independent form, or present a sufficient showing that the dependent claim(s) complies with the statutory requirements.
Applicant’s Arguments:
Applicant’s arguments have been considered, but are not completely persuasive. The Examiner recognizes some of the amendments were made to address this rejection, however, indefiniteness remains or was introduced by those amendments.
MPEP 2173: “The essential inquiry pertaining to this requirement is whether the claims set out and circumscribe a particular subject matter with a reasonable degree of clarity and particularity. "As the statutory language of ‘particular[ity]' and 'distinct[ness]' indicates, claims are required to be cast in clear—as opposed to ambiguous, vague, indefinite—terms. It is the claims that notify the public of what is within the protections of the patent, and what is not." Packard, 751 F.3d at 1313, 110 USPQ2d at 1788.”
The Examiner has identified multiple incompatible reasonable interpretations of the claims, and set forth why the claims fail to particularly point out and distinctly claim what Applicant considers their invention.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claim(s) 1, 26-27 is/are rejected under 35 U.S.C. 102a1 as being anticipated by Makowski (2022).
Makowski, E. et al. (2022) Co-optimization of therapeutic antibody affinity and specificity using machine learning models that generalize to novel mutational space. Nature Communications, vol 13: 3788, 14 pages and some supplemental material.
Applicant’s elected species are: Invention I (a DNA product), that is the product of species A (a DNA synthesis process), using evaluation by species B vii (using a biological product language).
Makowski is directed to generating and optimizing antibody sequences. “Here we evaluate the use of machine learning to simplify antibody co-optimization for a clinical-stage antibody (emibetuzumab) that displays high levels of both on-target (antigen) and off-target (non-specific) binding. We mutate sites in the antibody complementarity determining regions, sort the antibody libraries for high and low levels of affinity and non-specific binding, and deep sequence the enriched libraries.” (Abstract).
With respect to claim 1 and “A method performed by one or more computers for computationally designing a modified molecule that maintains at least a first feature of a first molecule while also achieving a second feature of a second molecule, the method comprising:
selecting the first molecule having the first feature;
selecting the second molecule having the second feature; and”
Makowski obtains DNA sequences encoding emibetuzumab, including Vh and Vl CDR encoding sequences, as set forth at p10, Methods. “Sites in the heavy chain CDRs of emibetuzumab were selected for mutagenesis using chemical rules reported previously for predicting antibodies with drug-like specificity38. Briefly, CDR sites in 10 the VH domain were selected for mutagenesis if they were (i) flagged by one or more of the six maximum chemical rules, (ii) hydrophobic or positively charged, (iii) solvent exposed (>10%), and (iv) relatively uncommon in human antibodies…”
“The first set of features are Unified Representation (UniRep) features, which are deep learning features obtained from a neural network trained on over twenty million unlabeled protein sequences to perform next amino acid prediction35. A compelling aspect of these features is that they are not biased by assumptions regarding which molecular features are most important for antibody affinity and specificity. We used the previously reported 64-unit neural network to generate 64 UniRep features per antibody35. The second set of features, which we refer to as PhysChem features, are 26 physicochemical features that are based on the VH domain sequence, including the isoelectric point, average residue hydrophobicity, and number of specific amino acids…” p6
With respect to claim 1 and “selecting the modified molecule by performing operations comprising:
generating a set of candidate molecules based at least in part on the first molecule and the second molecule;”
Makowski generates a set of candidate molecules, in scFAb libraries, as set forth at p10.
“For each of eight sites that were identified in the heavy chain CDRs for mutagenesis (Y33, R50, R54, R55, G56, A95, W97, and Y102), degenerate codons were selected to sample the wild-type residue in addition to five additional residues which sample a range of physico-chemical properties and were predicted to reduce non-specific binding (Fig. S1). The final library of yeast-displayed single-chain Fabs (scFabs; theoretical diversity of 1.7 ×106) was constructed via homologous recombination following electroporation of EBY100 Saccharomyces cerevisiae.”
With respect to claim 1 and “processing each of the first molecule, the second molecule, and each candidate molecule in the set of candidate molecules using an embedding neural network to generate a respective embedding of each of the first molecule, the second molecule, and each candidate molecule in the set of candidate molecules in an embedding space,
wherein the embedding neural network comprises at least one embedding layer defined by a set of neural network parameters and has been configured, through machine learning training, to generate feature-aware embeddings that capture semantic and functional attributes of input molecules;”
Makowski analyzes the candidate sequences as set forth in Fig 1, and the results section beginning at p2. The high affinity sequences, and the high/low non-specific binding sequences were embedded and featurized, including sequence information, physiochemical information, in a deep learning neural network.
“The sorted libraries were deep sequenced, and the resulting antibody sequences were used to train models for predicting metrics correlated with antibody affinity and specificity (non-specific binding) using different types of molecular features. These features included antibody sequences encoded as binary vectors, physicochemical features, and deep learning features. The resulting models were used not only to predict the classification of antibody affinity and specificity (e.g., high or low affinity), but also continuous metrics correlated with each property to predict intraclass variability (e.g., high vs. very high affinity).” Legend to Fig 1.
With respect to claim 1 and “generating a ranking of the set of candidate molecules using Pareto front optimization based on, for each candidate molecule:
(i) a classification of whether the embedding of the candidate molecule is within a threshold conservation distance of the embedding of the first molecule in the embedding space to maintain the first feature of the first molecule, and
(ii) a distance between the embedding of the candidate molecule and the embedding of the second molecule to design towards the second feature of the second molecule; and
selecting one or more of the candidate molecules based on the ranking of the candidate molecules.”
Makowski generates a ranked set of candidates, using Pareto Front optimization, for each generated candidate. Fig 1 illustrates the Pareto Front.
“The model predictions were also used to identify antibody mutants in the library at the Pareto frontier that maximize antibody affinity to different extents while minimizing tradeoffs due to reduced specificity (i.e., increased non-specific binding). Some of the models, which generalized to novel mutational space, were used to identify antibodies with even greater improvements in affinity and specificity than was possible in the experimentally sorted libraries.” Legend to Fig 1.
“To evaluate the predictions of Pareto optimal antibody variants, we next identified and produced 41 antibody mutants (as soluble IgGs) that were predicted to be at or near the Pareto frontier (Fig. 4A) and experimentally evaluated their levels of antigen (Fig. 4B) and non-specific (Fig. 4C) binding.” P5
“We identified a lead candidate (EM1) for further optimization that displayed an attractive combination of increased antigen binding (1.20x of wild type) and reduced non-specific binding (0.51x of wildtype; Fig.4D). We also selected additional clones for further mutagenesis, although to a more limited extent, to investigate the potential for optimizing antibody mutants with a diverse range of properties.” P5
As such, claims 1, 26 and 27 are anticipated.
Claim(s) 1, 25-27 is/are rejected under 35 U.S.C. 102a1 as being anticipated by Ahmed (2023).
Ahmed, M. M. (2023) Attention-based generative model in deep evolutionary learning: a many-objective approach to multi-target SMILES fragment-based drug design for cancer. Brock University, St. Catherines, Ontario, Canada. 159 pages.
This reference is directed to drug-like molecules as the candidate molecules. These fall within the scope of claim 1, but were not elected.
With respect to claim 1 and “A method performed by one or more computers for computationally designing a modified molecule that maintains at least a first feature of a first molecule while also achieving a second feature of a second molecule, the method comprising:
selecting the first molecule having the first feature;
selecting the second molecule having the second feature; and”
Ahmed selects representations of chemical molecules, each of which has an associated feature. The overall source of the molecules, are the publicly available ZINC and DrugBank databases (Section 5.1), or Protein Data Bank (PDB). These databases provide various features (attributes) associated with the molecules listed therein.
“The ZINC dataset [131] is a comprehensive collection of commercially available chemical
compounds, widely utilized in molecular generative tasks for drug discovery, providing a rich source of molecular structures to train deep learning models in predicting and generating novel compounds with desired properties. The DrugBank dataset [132], encompassing detailed drug data and extensive drug target information, plays a crucial role in molecular generative tasks by offering insights into drug mechanisms and interactions, thereby aiding in the development of novel therapeutics through informed molecular design and deep learning applications. Molecular attributes, such as SAS (Synthetic Accessibility Score), logP or solubility, and BAS (Binding Affinity Scores) with the protein targets CA9, GPX4, LPAl, LPA2, and LPA3, are generated and seamlessly integrated into the dataset to enrich the information available for analysis. SAS and logP have been computed using RDKit [105), and BAS are generated by the docking module utilizing QuickVina 2 [10]. For the BRICS fragmentation based DEL models, the molecules are cleaved into SMILES fragments following the BRICS algorithm [22] and are stored in a column, named fragments, in the training dataset, whereas for the HierVAE fragmentation based DEL models, this column is substituted by the fragments obtained using the HierVAE fragmentation strategy.” P75-76.
Section 2.5 of Ahmed provides representations of molecules in the domain of drug design, from string notations to topological descriptions or atomic coordinates. (p39) Whole representations, or fragments of representations can be chosen, depending on the design parameters. Fragment based design is discussed beginning at p44.
Ahmed selects representations of ligands which bind to 4 separate proteins. (Section 4). Each ligand has a feature, including Synthetic Accessibility Score (SAS), octanol-water partition coefficient (logP) and Binding Affinity Score (BAS). “Design Task 1” seeks to generate or design a single molecule having high affinity to 3 separate proteins. “Design Task 2” takes the representations of ligands which bind LPA1, LPA2, or LPA3, to generate a drug molecule with a) high affinity to LPA1 (feature 1) and b) low affinity to LPA2 and LPA3 (feature 2). (Section 4.7.2). Section 5 sets out the initially selected molecules for each task.
With respect to claim 1 and “selecting the modified molecule by performing operations comprising:
generating a set of candidate molecules based at least in part on the first molecule and the second molecule;”
Ahmed generates candidate molecules, based on the initially selected molecules using Deep Evolutionary Learning, which combines Deep Generative Modeling and Multi-objective evolutionary computation. (p4-5).
“DEL achieves the co-evolution of both the molecular sample data as well as the molecular generative model simultaneously across multiple generations guided by multiple properties concerned with drug design. The crux of DEL lies in this iterative process in which each evolutionary generation improves both the molecular sample set and the generative model's learning, thereby implementing a new learning paradigm called data-model co-evolution.” P5
Generative modeling is discussed in Section 2.1.
With respect to claim 1 and “processing each of the first molecule, the second molecule, and each candidate molecule in the set of candidate molecules using an embedding neural network to generate a respective embedding of each of the first molecule, the second molecule, and each candidate molecule in the set of candidate molecules in an embedding space,
wherein the embedding neural network comprises at least one embedding layer defined by a set of neural network parameters and has been configured, through machine learning training, to generate feature-aware embeddings that capture semantic and functional attributes of input molecules;”
Ahmed provides multiple options for transforming the representations of the molecules into an embedding space. One option are Autoencoders, which when supervised (labeled), can encode “context aware” embedding. (Section 2.1.2). Variational Autoencoders are discussed beginning at 2.1.3 and 2.1.4. Variational Transformers are disclosed at 2.1.5. “The model consists of an encoder and a decoder, both of which leverage the Transformer architecture. The encoder processes the input sequence and produces a latent representation…” p28. Transformers are disclosed at 2.1.6.
Input embedding is discussed beginning at p29, including an embedding layer, which embeds the representation of the molecule into an embedding space. (p30). The embeddings capture “semantic and functional attributes”. “It should also be noted that we give more weights to the word embeddings than the positional encodings because we need to make sure that we don't lose the semantic information of the embedding when we add the positional encoding.” P31. The Transformer layers each comprise a neural network (a feed forward network) which has been trained on the training set.
Input embedding is discussed with respect to training samples at p 54.
“A subset of training samples (if first generation) or population samples (otherwise) are projected to the latent space using the encoder of the VAE. The latent vectors are used as individuals for subsequent evolutionary operations”
Figures 4.1 and 4.2, and their description in the text illustrate contextual embedding of input molecules. P61.
With respect to claim 1 and “generating a ranking of the set of candidate molecules using Pareto front optimization based on, for each candidate molecule:
(i) a classification of whether the embedding of the candidate molecule is within a threshold conservation distance of the embedding of the first molecule in the embedding space to maintain the first feature of the first molecule, and
(ii) a distance between the embedding of the candidate molecule and the embedding of the second molecule to design towards the second feature of the second molecule; and
selecting one or more of the candidate molecules based on the ranking of the candidate molecules.”
Ahmed generates ranked lists of candidate molecules meeting either Design Task 1 or Design task 2. Ahmed provides several methods of optimization and ranking, including Pareto front optimization. Section 4, beginning at p52. P54-55 sets forth non-dominated ranking of feasible solutions including Pareto fronts/ ranks and crowding distance computations as steps of the multi-objective sorting operations, before and after combined molecule generation. P55-56 define the multi-objective sorting algorithms in the latent space.
“Using this concept of domination, all feasible solutions in a collection can be sorted to form Pareto frontiers (or fronts, ranks) (F = {F1, F2, F3, ... } ) . For instance, the first front F1 dominates F2 but solutions in F1 do not dominate each other. A function F(z) is defined to retrieve the rank of any feasible solution z in the population.” P56-57.
Section 5, Fig 5.1, p75, shows ranking occurring after generative modeling and filtering, and the sum or ranks is used to ultimately select the top candidates.
Pareto front optimization is defined at p36-37, as one type of multi-objective prioritization.
“While in a SingleOOP, the goal is to optimise only one objective function, in a MultiOOP, more than one objective must be simultaneously optimised. If the objective functions are not conflicting, a solution can be found where each objective reaches its optimum value. However, in MultiOOPs, those objectives are frequently conflicting (the improvement of one objective leads to the degradation of another objective), and also non-commensurable (when dealing with objectives that have different units or scales of measurement). In this case, there is usually no optimal solution but a set of trade-off solutions representing a compromise between the conflicting objectives. Such solutions are called non-dominated solutions which form the Pareto (optimal) set, which is the set of solutions that are equally optimal concerning the considered objectives. A decision vector (x* E S) is a Pareto-optimal solution if no other (x E S) dominates x*. The set of all Pareto optimal solutions in the decision space is called Pareto-optimal set, or simply Pareto set (PS), and its image in the objective space is called Pareto-optimal front or Pareto front (PF). Therefore, considering that the ideal approach for solving MultiOOPs would be to find many different trade-off solutions as close as possible to the PF and as diverse as possible along that front, it becomes clear that the classical methods used in SingleOOPs need a great effort to meet these goals…”
Ahmed addresses multi-objective evolutionary algorithms beginning at p38.
“MOEAs use a population of candidate solutions instead of a single solution, as in classical methods, to find many Pareto-optimal solutions in a single run. Although they do not guarantee to find the optimal trade-off solutions, they can provide a satisfactory approximation set, which is not too far away from the true Pareto Front.” P38
Ahmed uses the Hypervolume indicator as a quality indicator, that evaluate the quality of approximation sets produced by algorithms. P66
“These metrics assign a real value to approximation sets based on specific quality attributes, such as (i) convergence to the Pareto optimal region, and (ii) diversity across the Pareto Front. The Hypervolume (HV) indicator is the predominant metric for gauging the quality of non-dominated solution sets and is an effective quantitative scalar metric, which measures the volume dominated by the derived solution [7] and is strictly monotonic in terms of Pareto dominance. The larger the value of HV, the better the set of solutions covers the objective space and trade-offs between objectives. Especially, the calculation of the HV index does not require the ideal Pareto front of the test problem, which greatly facilitates the use of HV in practical applications.” P66-67.
“The Pareto-fronts obtained by DEL using the two different base models can be compared in terms of hypervolume (HV) to reflect the overall quality of the solutions generated. It is a good indicator of diversity and quality of the solutions obtained. We selected [7.2893,12.6058, 0, 0, 0] and [-7.2893, -12.6058, 0, -12.6, -13.2] as the reference points for task 1 ({SAS, logP, BCA9, BGPX4, BLPA1}) and task 2 ( {SAS, logP, BLPAl, BLPA2, BLPA3} ), respectively, which represents the property values of the worst possible samples in the training dataset. The obtained HVs in the generations 1, 5 and 10 (final population) are listed in Table 5.2.” p81-82.
“The generated molecular samples located on the first Pareto front (rank= 0) of the final (10th) population is considered as the high-rank molecules. In our experiments, the population size M was set to 20,000. In this section, we apply the virtual screening criteria and protein-ligand complex visualization on the molecular samples present only on the first Pareto front.” P88.
As such, claims 1 and 26 are anticipated by Ahmed.
Ahmed discloses the use of GPU based parallelism of docking procedures in section 6, meeting claim 25. LSTM and GRU are disclosed in Section 2.
Ahmed considers physical synthesis of selected candidate molecules in multiple places, including p2:
“There's a nuanced art of molecular optimization: designing new molecules with specific, desired attributes, rather than naively enumerating the training data [3]. This transforms drug design into an intricate task of simultaneously optimizing multiple conflicting objectives, such as molecular weight, octanol-water partition coefficient (logP), polar surface area, toxicity and protein-ligand binding affinity, while ensuring synthetic feasibility, which means that the designed drug candidate must be realistically synthesizable, and safety.”
Further consideration of synthesis, meeting claim 27, occurs at p7, “The third contribution of this study is to design molecules that not only possess realistic synthesizability and high solubility, but also concurrently target multiple protein targets associated with cancer…” p35, (minimizing synthesis costs), the Synthesis Accessibility Score is a critical objective, (p36) using BRICS to emphasize real-world applicability of computationally designed molecules (p62), p105, et al. Section 4.7, 5, Appendix A, each set forth the selected molecules that should be synthesized and tested in vitro.
Claim(s) 1, 25-26 is/are rejected under 35 U.S.C. 102a1 as being anticipated by Liu (2021).
Liu et al. (2021) DrugEx v2: de novo design of drug molecules by Pareto-based multi-objective reinforcement learning in polypharmacology. J Cheminformatics, vol 13:85, 15 pages and some supplemental material.
This reference is directed to drug-like molecules as the candidate molecules. These fall within the scope of claim 1, but were not elected.
With respect to claim 1 and “A method performed by one or more computers for computationally designing a modified molecule that maintains at least a first feature of a first molecule while also achieving a second feature of a second molecule, the method comprising:
selecting the first molecule having the first feature;
selecting the second molecule having the second feature; and”
Liu et al select molecules, each having at least one feature, as set forth in the Materials and Methods section, p2, including SMILES representations of drug-like molecules from ChEMBL database. Additional molecules (ligands) were extracted from ChEMBL to create a Ligand Set, with bioactivity measurements towards 3 receptors.
With respect to claim 1 and “selecting the modified molecule by performing operations comprising:
generating a set of candidate molecules based at least in part on the first molecule and the second molecule;
processing each of the first molecule, the second molecule, and each candidate molecule in the set of candidate molecules using an embedding neural network to generate a respective embedding of each of the first molecule, the second molecule, and each candidate molecule in the set of candidate molecules in an embedding space,
wherein the embedding neural network comprises at least one embedding layer defined by a set of neural network parameters and has been configured, through machine learning training, to generate feature-aware embeddings that capture semantic and functional attributes of input molecules;”
Liu et al. perform crossover and mutation operations on the selected molecules, to analyze how DrugEx2 performs in designing molecules with competing or differing objectives. (p2).
“Since the first version of DrugEx (vl) demonstrated effectiveness for designing novel A2AAR ligands, we began to extend this method for drug design toward multiple targets. In this study, we updated DrugEx to the second version (v2) through adding crossover and mutation operations, which were derived from evolutionary algorithms, to the reinforcement learning (RL) framework… For the multi-target case, desired molecules should have a high affinity towards both the A1AR and A2AAR. In the target-specific case, on the other hand, we required molecules to have only high affinity towards the A2AAR but a low affinity to the A1AR. In order to decrease toxicity and risk of adverse events, molecules were additionally obliged to have a low affinity for hERG in both cases.” P2
The evolutionary algorithm strategies are disclosed at p5-6.
“we used a DL method to define model-based mutation and crossover operations. Moreover, we employed an RL method to replace the sample selection step for the update of model or population in EDA or EA, respectively.” P6
Figure 2 illustrates differing evolutionary algorithm flowcharts. Fig 3 illustrates how generating candidates was carried out, including sequential generation and embedding as the candidate is generated.
The first and second molecules are embedded into a vector in an embedding space, by multiple models, including neural networks. P3
“Descriptors used as input were ECFP6 fingerprints [22] with 2048 bits (2048 dimensions, or 2048D) calculated by the RDKit Morgan Fingerprint algorithm (using a three-bond radius). Moreover, the following 19D physico-chemical descriptors were used:… . Hence, each molecule in the dataset was transformed into a 2067D vector. Before being input into the model, the value of input vectors were normalized to the range of [O, 1] by the MinMax method. Model output value is the probability whether a given chemical compound was active based on this vector. Four algorithms were benchmarked for QSAR model construction, Random Forest (RF), Support Vector Machine (SVM), Partial Least Squares regression (PLS), and Multi-task Deep Neural Network (MT-DNN)… In the MT-DNN, the architecture contained three hidden layers activated by a rectified linear unit (ReLU) between input and output layers, and the number of neurons were 2048, 4000, 2000, 1000 and 3 in these subsequent layers. The training process consisted of 100 epochs with 20% of hidden neurons randomly dropped out between each layer. The mean squared error was used to construct the loss function and was optimized by the Adam algorithm [25] with a learning rate of 10-3.” P3
Generation of candidate molecules is disclosed at p3-4.
“Each SMILES-format molecule in the ChEMBL and LIGAND sets was split into a series of tokens. Then all tokens existing in this dataset were collected to construct the SMILES vocabulary. The final vocabulary contained 84 tokens (Additional file 1: Table Sl) which were selected and arranged sequentially into valid SMILES sequences through correct grammar… SMILES sequence construction under the RL framework can be viewed as a series of decision-making steps (Fig. 1). The generator (G) and the predictors (Q) are regarded as the policy and reward function, respectively. In this study we used multi-objective optimization (MOO) and the aim is to maximize each objective for each scenario, albeit with differences in desirability.”
The candidate molecules are also embedded into an embedding layer, as set forth at p3-4.
With respect to claim 1 and “generating a ranking of the set of candidate molecules using Pareto front optimization based on, for each candidate molecule:
(i) a classification of whether the embedding of the candidate molecule is within a threshold conservation distance of the embedding of the first molecule in the embedding space to maintain the first feature of the first molecule, and
(ii) a distance between the embedding of the candidate molecule and the embedding of the second molecule to design towards the second feature of the second molecule; and
selecting one or more of the candidate molecules based on the ranking of the candidate molecules.”
Liu et al. analyzes each candidate molecule using a Pareto Front Scheme, as set forth at page 5. This scheme calculates a desirability score, which characterizes how well the candidate meets the different objectives. Distance calculations and ranking are provided.
“After the dominance between all pair of solutions being determined, the non-dominated scoring algorithm [29] is exploited to obtain different layers of Pareto frontiers which consist of a set of solutions. The solutions in the top layer are dominated by the other solutions in the lower layer [30]. In order to speed up the non-dominated sorting algorithm, we employed PyTorch to implement this procedure with GPU acceleration. After obtaining the frontiers ranking from dominated solutions to dominant solutions, the molecules were ranked based on the average of Tanimoto-distance instead of crowding distance with other molecules in the same frontier, and molecules with larger distances were ranked on the top.” P5
The PF scheme was analyzed for performance, as set forth at p9-11. PF optimization was indicated as generating molecules with a larger diversity, but the desirability scores were not as high in the weighted scoring scheme (WS). Table 2 indicates that PF “assists models to achieve better diversity and distribution of substructures.” P10.
“When investigating the SA and QED scores, we observed that the PF scheme helped to make all generated molecules more drug-like because of higher QED scores than the molecules generated
by the WS scheme in both multi-target (Fig. 5A-D) and target-specific cases (Fig. 5E-H). Comparing these methods, the molecules generated by REINVENT were supposedly easier to synthesize and more drug-like than others, but the molecules of DrugEx v 1 had more similar distributions with the molecules in the LIGAND set.” P10
PF was analyzed with respect to the chemical space explored by the models, and “the PF scheme could guide all of the generators to better cover chemical space than the WS scheme.” p10
From the ranked set of candidate molecules, Liu selects 16 possible antagonists as set forth at p11 and Fig 7.
“As an example, 16 possible antagonists (without ribose moiety and with a molecular weight< 500) generated by DrugEx v2 with the PF scheme were selected as candidates for both multi-target cases and target specific case, respectively. These molecules were ordered by the selectivity which was calculated as the difference of pXs between two different protein targets. In the multi-target case (Fig. 7 A) rows and columns are sorted by selectivity for the A2AAR and A1AR over hERG respectively, because the desired ligands prefer A1AR and A2AAR to hERG. Conversely in the target-specific case the generated molecules are required to bind only A2AAR rather than A1AR and hERG (Fig. 7B). Hence, here selectivity for the A2AAR over A1AR and hERG were represented by the rows and columns respectively.”
As such, claims 1 and 26 are anticipated.
With respect to claim 25, Liu uses GPU acceleration as set forth above.
Claim(s) 1, 26 is/are rejected under 35 U.S.C. 102a1 as being anticipated by Hong (2024).
Hong, L. et al. (2024) An integrative approach to protein sequence design through multiobjective optimization. PLOS Computational Biology, Vol 20 no 7: e1011953, 37 pages.
This reference is directed to polypeptides as the candidate molecules. These fall within the scope of claim 1, but were not elected.
With respect to claim 1 and “A method performed by one or more computers for computationally designing a modified molecule that maintains at least a first feature of a first molecule while also achieving a second feature of a second molecule, the method comprising:
selecting the first molecule having the first feature;
selecting the second molecule having the second feature; and”
Hong provides selecting polypeptide molecules, each with a differing feature. One molecule (polypeptide) is E.coli RfaH. “RfaH is a difficult model system for multistate design because it is a foldswitching protein that undergoes extensive conformational changes between an N-terminal domain bound all-α (RfaHα) and a dissociated all-β (RfaHβ) C-terminal domain], which involves considerable differences in the secondary and tertiary structure and solvent accessibility of its amino acid residues.” P3. Two other molecules (polypeptides) selected include the E. coli P pilus chaperone protein PapD and the vertebrate calcium-binding protein CaM. (Fig 1).
See also Methods section, Structure Preparation, p19-21.
With respect to claim 1 and “selecting the modified molecule by performing operations comprising:
generating a set of candidate molecules based at least in part on the first molecule and the second molecule;”
Hong generates candidate molecules, based at least in part on the molecules. Hong utilizes “biophysically-informed mutation operators” which mutate the polypeptide sequences. “first, new design candidates are proposed through a mutation operator; here, this operator is composed of ESM-1 v, which is used to rank residue positions, and ProteinMPNN (pMPNN), which is used to redesign the least nativelike positions.” (Legend to Fig 1.) pMPNN, and the genetic algorithm NSGA-II are described further in the Methods section, p21-23.
With respect to claim 1 and “processing each of the first molecule, the second molecule, and each candidate molecule in the set of candidate molecules using an embedding neural network to generate a respective embedding of each of the first molecule, the second molecule, and each candidate molecule in the set of candidate molecules in an embedding space,
wherein the embedding neural network comprises at least one embedding layer defined by a set of neural network parameters and has been configured, through machine learning training, to generate feature-aware embeddings that capture semantic and functional attributes of input molecules;”
Hong provides embedding of the initial molecules, and the candidate molecules into an objective space in a context-aware manner, which meets the BRI of an embedding layer. Figure 2 illustrates different combinations of a genetic algorithm, a mutation operator, and objective optimizer, in improving RfaH design outcomes. The pMPNN embed features into an embedding space, or objective space, using a neural network.
“For PapD, we redesign its multi-specific interface between PapD-PapE, PapD-PapK, and PapD-PapD, which results in a three-dimensional objective space; for CaM, we model a wide variety of functional states that involve changes in its homo-oligomerization states, central linker conformation, binding partners, and calcium-binding state, which results in a 14-dimensional objective space, and redesign almost all but the Ca2+ binding residues (see Methods for more details).” P14.
The molecular language model ESM-1v is a pre-trained protein language model used as a mutational effect predictor, and is discussed at p24.
With respect to claim 1 and “generating a ranking of the set of candidate molecules using Pareto front optimization based on, for each candidate molecule:
(i) a classification of whether the embedding of the candidate molecule is within a threshold conservation distance of the embedding of the first molecule in the embedding space to maintain the first feature of the first molecule, and
(ii) a distance between the embedding of the candidate molecule and the embedding of the second molecule to design towards the second feature of the second molecule; and
selecting one or more of the candidate molecules based on the ranking of the candidate molecules.
Hong provides Pareto front optimization to rank candidate molecules, for the desired objective or feature. “A key feature of an evolutionary multiobjective optimization algorithm is that the solutions approximate the Pareto front in the objective space. To the extent that objective functions such as the pMPNN-SD log likelihood score and the AF2Rank composite score correlate with desirable biophysical properties such as stability, this Pareto optimality condition necessarily needs
to manifest in the sequence space as mutations that improve such properties for either or both states of RfaH.” P8.
“Nevertheless, the fact that all three embeddings generate similar clustering patterns suggest
that the multistate design methods examined so far have all succeeded, to various extent,
in recapitulating the WT RfaH sequence profile, which is distinct from hypothetical non-foldswitching
single-state RfaH sequence profiles.” P9.
“In this work, we examined the potential of evolutionary multiobjective optimization as an integrative framework for protein sequence design. This framework was chosen because of its ability
to explicitly approximate the Pareto front in a user-specified objective space, and the flexibility it affords to construct informative mutation operators to guide sampling in the sequence space. Using the multistate design problem of the two-state foldswitching protein RfaH as an in-depth case study, as well PapD and CaM as examples of higher-dimensional optimization problems, we showed that this approach led to design candidates with reduced bias and variance in native sequence recovery, without the need for post hoc filtering or pMPNN hyperparameter tuning.
Moving forward, as more models become available that capture additional aspects of protein sequence-structure-function relationships, we anticipate such an evolutionary multiobjective optimization framework to be broadly relevant for and readily adaptable to even more complex design tasks. For example, the framework may be adapted for the design of biologics that simultaneously optimizes stability, affinity, specificity, immunogenicity, and pharmacokinetics, or for the construction of entire signaling or metabolic pathways, where multiple biomolecules are simultaneously designed to optimize functional properties” P18.
As such, claims 1, and 26 are anticipated.
Double Patenting
The nonstatutory double patenting rejection is based on a judicially created doctrine grounded in public policy (a policy reflected in the statute) so as to prevent the unjustified or improper timewise extension of the “right to exclude” granted by a patent and to prevent possible harassment by multiple assignees. A nonstatutory double patenting rejection is appropriate where the conflicting claims are not identical, but at least one examined application claim is not patentably distinct from the reference claim(s) because the examined application claim is either anticipated by, or would have been obvious over, the reference claim(s). See, e.g., In re Berg, 140 F.3d 1428, 46 USPQ2d 1226 (Fed. Cir. 1998); In re Goodman, 11 F.3d 1046, 29 USPQ2d 2010 (Fed. Cir. 1993); In re Longi, 759 F.2d 887, 225 USPQ 645 (Fed. Cir. 1985); In re Van Ornum, 686 F.2d 937, 214 USPQ 761 (CCPA 1982); In re Vogel, 422 F.2d 438, 164 USPQ 619 (CCPA 1970); In re Thorington, 418 F.2d 528, 163 USPQ 644 (CCPA 1969).
A timely filed terminal disclaimer in compliance with 37 CFR 1.321(c) or 1.321(d) may be used to overcome an actual or provisional rejection based on nonstatutory double patenting provided the reference application or patent either is shown to be commonly owned with the examined application, or claims an invention made as a result of activities undertaken within the scope of a joint research agreement. See MPEP § 717.02 for applications subject to examination under the first inventor to file provisions of the AIA as explained in MPEP § 2159. See MPEP § 2146 et seq. for applications not subject to examination under the first inventor to file provisions of the AIA . A terminal disclaimer must be signed in compliance with 37 CFR 1.321(b).
The filing of a terminal disclaimer by itself is not a complete reply to a nonstatutory double patenting (NSDP) rejection. A complete reply requires that the terminal disclaimer be accompanied by a reply requesting reconsideration of the prior Office action. Even where the NSDP rejection is provisional the reply must be complete. See MPEP § 804, subsection I.B.1. For a reply to a non-final Office action, see 37 CFR 1.111(a). For a reply to final Office action, see 37 CFR 1.113(c). A request for reconsideration while not provided for in 37 CFR 1.113(c) may be filed after final for consideration. See MPEP §§ 706.07(e) and 714.13.
The USPTO Internet website contains terminal disclaimer forms which may be used. Please visit www.uspto.gov/patent/patents-forms. The actual filing date of the application in which the form is filed determines what form (e.g., PTO/SB/25, PTO/SB/26, PTO/AIA /25, or PTO/AIA /26) should be used. A web-based eTerminal Disclaimer may be filled out completely online using web-screens. An eTerminal Disclaimer that meets all requirements is auto-processed and approved immediately upon submission. For more information about eTerminal Disclaimers, refer to www.uspto.gov/patents/apply/applying-online/eterminal-disclaimer.
Claim s 1, 25-27 are provisionally rejected on the ground of nonstatutory double patenting as being unpatentable over claim s 1-18 of copending Application No. 19/415,089 (reference application). Although the claims at issue are not identical, they are not patentably distinct from each other because both are directed to selecting molecules, that have features, generating new molecules, embedding the original and new molecules using embedding neural networks, that embed features in a context-aware manner, and evaluating the generated molecules using optimization processes.
This is a provisional nonstatutory double patenting rejection because the patentably indistinct claims have not in fact been patented.
Claim s 1, 25-27 are provisionally rejected on the ground of nonstatutory double patenting as being unpatentable over claim s 1-20 of copending Application No. 19/415,171 (reference application). Although the claims at issue are not identical, they are not patentably distinct from each other because both are directed to selecting molecules, that have features, generating new molecules, embedding the original and new molecules using embedding neural networks, that embed features in a context-aware manner, and evaluating the generated molecules using optimization processes.
This is a provisional nonstatutory double patenting rejection because the patentably indistinct claims have not in fact been patented.
Claim s 1, 25-27 are provisionally rejected on the ground of nonstatutory double patenting as being unpatentable over claim s 1-20 of copending Application No. 19/431,513 (reference application). Although the claims at issue are not identical, they are not patentably distinct from each other because both are directed to selecting molecules, that have features, generating new molecules, embedding the original and new molecules using embedding neural networks, that embed features in a context-aware manner, and evaluating the generated molecules using optimization processes.
This is a provisional nonstatutory double patenting rejection because the patentably indistinct claims have not in fact been patented.
Claim s 1, 25-27 are provisionally rejected on the ground of nonstatutory double patenting as being unpatentable over claim s 3-7 and 18-32 of copending Application No. 19/340736 (reference application). Although the claims at issue are not identical, they are not patentably distinct from each other because both are directed to selecting molecules, that have features, generating new molecules, embedding the original and new molecules using embedding neural networks, that embed features in a context-aware manner, and evaluating the generated molecules using optimization processes.
This is a provisional nonstatutory double patenting rejection because the patentably indistinct claims have not in fact been patented.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MARY K ZEMAN whose telephone number is 5712720723. The examiner can normally be reached on 8am-2pm M-F. Email may be sent to mary.zeman@uspto.gov if the appropriate permissions have been filed.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Larry Riggs can be reached on 571 270-3062. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/MARY K ZEMAN/ Primary Examiner, Art Unit 1686
Prosecution Insights