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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 14 July 2025 has been entered.
This Application is a Track One Request Application. The request filed 05 June 2024 was granted.
Status of the Claims
The claim set received 14 July 2025 has been entered into the application.
Claims 1, 11, and 20 are amended.
Claims 1, 11, and 20 are objected to.
Claim(s) 1-20 are pending
Election/Restrictions
Applicant elected Group I encompassing claims 1-10.
Claims 11-20 were withdrawn from further consideration pursuant to 37 CFR 1.142(b) as being drawn to a nonelected, there being no allowable generic or linking claim. Election was made without traverse in the reply filed on 29 Oct 2024. However, upon initial search and review, the claimed limitations in view of the specification would not provide an undue burden with respect to examining all three groups together.
Accordingly, the restriction requirement is withdrawn.
Therefore, claims 1-20 are pending examination.
Priority
No priority is claimed.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on filed 10 November 2025, 18 September 2025, and 18 June 2025 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement has been considered by the examiner.
Claim Objections
Claims 1, 11, and 20 objected to because of the following informalities: “causing physical synthesis of a physical protein having assembled physical amino acids and having the protein property predicted using the biological language reasoning machine learning model”. The claimed step should be amended to recite “
Claim Rejections - 35 USC § 112
35 USC § 112(b)
It is noted the amendments received 14 July 2025 necessitated new ground(s) of rejection.
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 1-20 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.
Claims 1, 11, and 20 were amended to recite “causing physical synthesis of a physical protein having assembled physical amino acids and having the protein property predicted using the biological language reasoning machine learning model”. The claimed step renders the claim indefinite because it is not clear how the method is to cause synthesis of a physical protein because neither the previous or subsequent steps recite physical steps or laboratory methods for synthesizing a physical protein. It is noted that claim 11 is drawn to a computer system while claim 20 is drawn to a computer program product. The steps encompassed within computer system of claim 11 and the software of claim 20 do not contain any physical or laboratory steps for synthesizing the physical protein. Thus, it is unclear how a computer system and computer program product are to synthesize a physical protein. It is recommended to amend the claimed step to clarify what is synthesizing the physical protein and/or add physical steps or laboratory steps of synthesizing the physical protein. It is noted that any claim amendments should contain language and limitations consistent with and supported by specification.
Claims 2-10 and 12-19 are rejected because they fail to provide limitations to overcome the deficiencies of their base claim(s) 1 and 11.
Claim Rejections - 35 USC § 101
The rejection of claims 1-10 under 35 U.S.C § 101 in the Office Action mailed 20 March 2025 is withdrawn in view of the amendments received 14 July 2025.
Claim Rejections - 35 USC § 103
It is noted the amendments received 14 July 2025 necessitated new ground(s) of rejection.
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.
Claim(s) 1-6, 11-15, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Yang et al. (Protein engineering, design and selection, 2023-01, Vol.36) in view of Xue et al. (Proceedings of the 33rd ACM International Conference on Multimedia, 2025, p.729-738: Cited in the IDS filed 10 November 2025 NPL Cite No. 2) in view of Pritzel et al. (US Patent Pub: US 2024/0087686; Patent Pub Date: 14 March 2024) in view of Lee et al. (US Patent Pub: US 2022/0384058; Patent Pub Date: 01 December 2022) in view of Mi et al. (Protein science, 2025-02, Vol.34 (2), p.e70041-n/a).
Claims 1, 11, and 20 recite for a specific amino acid in a protein, determining physically neighboring amino acids of the specific amino acid based on physical distances with respect to the specific amino acid in a local physical protein structure.
Claims 1, 11, and 20 recite including representations of the determined physically neighboring amino acids in a structure encoder input for the specific amino acid.
Claims 1, 11, and 20 recite providing the structure encoder input to an autoencoder trained using geometric loss to determine a local structure token representing the local physical protein structure for the specific amino acid separate from an individual amino acid token for the specific amino acid.
Claims 1, 11, and 20 recite combining the local structure token in a second input track with the individual amino acid token in a first input track into a combined sequence data and using the combined sequence data to train a biological language reasoning machine learning model.
Claims 1, 11, and 20 recite predicting a protein property including by unmasking a masked token in an input token sequence of the biological language reasoning machine learning model.
Claims 1, 11, and 20 recite causing physical synthesis of a physical protein having assembled physical amino acids and having the protein property predicted using the biological language reasoning machine learning model.
Yang et al. (Yang) teaches the input edge features for the ith residues consist of the dihedral and planar angles and the Euclidean distance between the Cβ atom of the residue and the Cβ atoms of its k-nearest neighbor [page 15 Methods paragraph under equation 3], as in claims 1, 11, and 20 for a specific amino acid in a protein, determining physically neighboring amino acids of the specific amino acid based on physical distances with respect to the specific amino acid in a local physical protein structure.
Yang teaches using a masked amino acid sequence and using associated protein structures (i.e., local structures such as alpha helices and beta sheets). Yang teaches the using bi-directional encoder representations from transformers (BERT)) [page 1 introduction]. Yang teaches a structured conditioned MLM pretraining. Yang teaches “The backbone atoms of amino-acid residues i and j with their dihedral and planar angles highlighted.” [page 3 figure 1 a)], as in claims 1, 11, and 20 including representations of the determined physically neighboring amino acids in a structure encoder input for the specific amino acid.
Dependent claims 2-3, 6, and 12-13.
With respect to claim 2 and 12, the claims are rendered obvious because Yang teaches “The input edge features for the ith residue consist of the dihedral and planar angles and the Euclidean distance between the Cβ atom of residue i and the Cβ atoms of its k-nearest neighbors [page 15 equation (3)]. It is obvious that based on the distance values, the candidate amino acid would be included with the candidate polypeptide.
With respect to claim 3 and 13, the claims are rendered obvious because Yang teaches “The input edge features for the ith residue consist of the dihedral and planar angles and the Euclidean distance between the Cβ atom of residue i and the Cβ atoms of its k-nearest neighbors [page 15 equation (3)]. Yang teaches “The backbone atoms of amino-acid residues i and j with their dihedral and planar angles highlighted.” [page 3 figure 1 d]. Here, using the Euclidean distance between the Cβ atom of residue i and the Cβ atoms of its k-nearest neighbors of Yang makes obvious the steps of claims 3 and 13 because it is obvious that the residue “i” can be used as a reference location for any amino acid and their associated subsequent reference or candidate amino acid and can be used to base a physical distance between atoms.
Yang teaches using Euclidean distance between the Cβ carbon and nearest neighbors [page 15 middle para under equation 3], as in claim 6.
Yang does not teach claims 1, 11, and 20 providing the structure encoder input to an autoencoder trained using geometric loss to determine a local structure token representing the local physical protein structure for the specific amino acid separate from an individual amino acid token for the specific amino acid. Yang does not teach claims 1, 11, and 20 combining the local structure token in a second input track with the individual amino acid token in a first input track into a combined sequence data and using the combined sequence data to train a biological language reasoning machine learning model. Yang does not teach claims 1, 11, and 20 predicting a protein property including by unmasking a masked token in an input token sequence of the biological language reasoning machine learning model. Yang does not teach claims 1, 11, and 20 causing physical synthesis of a physical protein having assembled physical amino acids and having the protein property predicted using the biological language reasoning machine learning model. Yang does not teach claims 4-5 and 14-15.
Xue et al. (Xue) teaches using three different types of input tokens are fed into a single transformer and distinguished by input embeddings. Xue teaches the raw amino acid sequence is first tokenized by a protein tokenizer then converted into amino acids embeddings. Xue teaches topological structure token [page 3 section 3]. Xue teaches “To alleviate the above limitations, we propose a multimodal protein model, which tries to extract a better protein representation by jointly learning the sequence-structure-function (S2F) features of proteins, inspired by the recent multimodal pre-training models in natural language processing and computer vision [page 2 middle para.]. Xue teaches “the model learns to figure out whether the input tokens of different modalities are from the same example (Figure 2(c)).” Xue teaches the “(b) The pipeline of the topology encoder that extracts the topological structure token from the domain structure. (c) An illustration of the positive data augmentation and negative sampling processes for pre-training tasks.” [page 4 section 3.2 figure 2]. Xue teaches the model is used for cross-species interactions, antibody-antigen interaction, and mutation-driven affinity change predictions [pages 5-6, figure 3]. Xue teaches using 494 example from SabDab and using positive and negative samples selected from antibodies that can bind to the receptor of the S protein in SARS-Cov-2 [page 7 section 4.3] which inherently contains local secondary structures (i.e., alpha helices, beta sheets ), as in claims 1, 11, and 20 combining the local structure token in a second input track with the individual amino acid token in a first input track into a combined sequence data and using the combined sequence data to train a biological language reasoning machine learning model.
Pritzel et al. (Pritzel) discloses a method for unmasking a masked representation of a protein using a protein reconstruction neural network [disclosure abstract]. Pritzel discloses “wherein a predicted embedding corresponding to a masked embedding in a representation of the amino acid sequence of the protein defines a prediction for an identity of an amino acid at a corresponding position in the amino acid sequence, wherein a predicted embedding corresponding to a masked embedding in a representation of the structure of the protein defines a prediction for a corresponding structural feature of the protein.” [Pritzel, disclosure abstract], as in instant claims 1, 11, and 20 predicting a protein property including by unmasking a masked token in an input token sequence of the biological language reasoning machine learning model.
Lee et al. (Lee) disclose train, using the data set, one or more first layers of a machine learning model to determine relationships of one or more components of a portion of a string described by a first dialect, wherein the one or more components pertain to amino acids associated with first activity level information of the one or more sequences, and the one or more first layers comprise one or more nodes executing one or more objective functions that optimize at least one secondary objective related to the first activity level information [Lee, claim 1]. Lee discloses generate, using the one or more first layers and the final layer, the string comprising the portion and the remainder by arranging, according to the logical rules, a sequence of amino acids included in the string, wherein the string represents a first candidate drug compound comprising the sequence of amino acids associated with the first activity level information and the second activity level information. Lee discloses synthesize, via at least a reaction chamber of an automated flow synthesis platform, the first candidate drug compound in order to create a drug compound. [Lee, claim 1], as in instant claims 1, 11, and 20 causing physical synthesis of a physical protein having assembled physical amino acids and having the protein property predicted using the biological language reasoning machine learning model.
Mi et al. (Mi) teaches a schematic representation of the GDFold2 pipeline of combining PDB data and geometric information obtained from different models which is further processed using loss functions [page 5 figured 2]. Mi teaches “Firstly, it is capable of integrating various sources of predicted geometric information with statistical constraints to rapidly construct satisfactory all atom protein structures. Secondly, it supports highly flexible and fully customizable constraining loss functions, which can be rewritten by users according to their own definitions and requirements.” [page 4 first para].
Mi teaches “2D geometric information predicted from various methods and statistics summarized from high-resolution protein structures are converted into geometric constraints (Figure 2A) and statistical constraints (Figure 2B), respectively, which are then taken as optimization aims for the user-defined loss functions in the neural network (named atomic coordinate network) for atomic coordinate generation.” [page 4 overview of the GDFold2 pipeline figure 1],
Mi teaches “A) Evolution of the loss function and the protein structure in the folding process of GDFold2. The horizontal axis represents the number of optimization steps and the vertical axis represents the sum value of all applied losses. At steps 400 and 800, the statistical loss and van der Waals loss are added to the optimization process, respectively, where the insets present the gradually refined local backbone structures by these two loss functions.”
Mi teaches “The folding process of GDFold2 can be divided into three main stages (Figure 4A), separated by the step points at which new loss terms are applied. The first stage optimizes the backbone structure of the target protein using the loss functions designed for the predicted geometric constraints” [page 11 first para, figure 4], as in Claims 1, 11, and 20 geometric loss function. Here, the limitations of Mi construct a system that uses loss function based on geometric information and constraints. As such and with respect to claims 1, 11, and 20 providing the structure encoder input to an autoencoder trained using geometric loss to determine a local structure token representing the local physical protein structure for the specific amino acid separate from an individual amino acid token for the specific amino acid, the claimed steps are rendered obvious because Yang teaches using autoencoders [page 3 figure 1]. Xue teaches using sequence-structure-function model that uses tokenized data [page 3 section 3.1]. Xue teaches the model is used for cross-species interactions, antibody-antigen interaction, and mutation-driven affinity change predictions [pages 5-6, figure 3]. Xue teaches using 494 examples from SabDab and using positive and negative samples selected from antibodies that can bind to the receptor of the S protein in SARS-Cov-2 [page 7 section 4.3] which inherently contains local secondary structures (i.e., alpha helices, beta sheets), inherently encompasses amino acids that are separate, and provides encoder input into an autoencoder. Mi teaches “A) Evolution of the loss function and the protein structure in the folding process of GDFold2. The horizontal axis represents the number of optimization steps and the vertical axis represents the sum value of all applied losses. At steps 400 and 800, the statistical loss and van der Waals loss are added to the optimization process, respectively, where the insets present the gradually refined local backbone structures by these two loss functions.” [page 11 figure 4].
Dependent claims 4-5 and 14-15.
Mi teaches “The local coordinate system of each amino acid residue is defined in the same way as AlphaFold27, by placing the Cα atom at the origin, C atom on the x-axis and N atom in the xy-plane. Thus, given the optimized Cα coordinates and relative rotation matrices of all residues, the positional vectors of all atoms could be inferred in the global frame, based on prior knowledge about the intra-residue and peptide-bond internal coordinate information.” [page 7 3D protein structure modeling], as in claims 4-5 and 14-15.
It would have been obvious to one of ordinary skill in the art by the effective filing date of the claimed invention to modify Yang in view of Xue because Xue teaches a multimodal pretraining system, Sequence-Structure-Function (S2F) that uses transformers/autoencoder pipeline [page 4 figure 2] for processing amino acid sequence data and primary-tertiary amino acid structure data for predicting protein-protein interactions. One of ordinary skill in the art would be motivated to combine Yang with Xue because Xue teaches using transformers, autoencoders, masking methods, and using tokens to investigate cross-species protein-protein interactions, antibody-antigen interactions, and mutation-driven affinity change prediction [pages 4-6]. Here, one of ordinary skill in the art would recognize that the modules, transformers (i.e., BERT) and tokenization methods of Xue could be combined with and/or substituted into the system of Yang because Yang and Xue teach using similar bidirectional transformers for processing amino acid data to construct proteins but Xue expands on using masked tokenization processes for determining protein-protein interactions. As such, there is a reasonable expectation that combining Yang in view of Xue would yield a predictable claimed step that can combine structure tokens of amino acids into tracks for and using the data to train a biological language machine learning model for subsequently protein structure synthesization.
It would have been obvious to one of ordinary skill in the art by the effective filing date of the claimed invention to modify Yang in view of Xue in view of Pritzel because Pritzel discloses predicting protein representations from masked protein representations [title] and teaches unmasking methods for protein reconstruction [Pritzel, claim 1]. One ordinary skill in the art would recognize that although Pritzel does not teach using encoders, decoders, or autoencoders, Pritzel utilizes embedding methods that can be combined with and/or substituted into the method of Yang in view of Xue to provide steps for unmasking proteins representations to predict protein properties to subsequently synthesize a protein using said protein properties. Therefore, there is a reasonable expectation of success that combining Yang in view of Xue in view of Pritzel would yield a predictable method that can utilize unmasking techniques for predicting properties to proteins that can be subsequently synthesized.
It would have been obvious to one of ordinary skill in the art by the effective filing date of the claimed invention to modify Yang in view of Xue in view of Pritzel in view of Lee because Lee discloses using autoencoders (VAE) framework [page 18 right col para 0177] and neural networks for analyzing amino acid data to generate a candidate drug and synthesizing the candidate drug [Lee, claim 1]. One of ordinary skill in the art would recognize that Lee utilizes autoencoders and neural networks to process protein dialects (i.e., tokens, symbols, words) and lexical elements for predicting a protein structure (i.e., candidate drug compound) that is subsequently synthesized [Lee claim 1]. Here, Lee teaches that proteins structures predicted by computerized processes can be physically synthesized from the predicted amino acid property data. As such, there is a reasonable expectation of success that combining Yang in view of Xue in view of Pritzel in view of Lee would yield a predictable method than can process amino acid data using a biological language machine learning model to predict protein properties and structures that can be subsequently physically synthesized.
It would have been obvious to one of ordinary skill in the art by the effective filing date of the claimed invention to modify Yang in view of Xue in view of Pritzel in view of Lee in view of Mi because Mi teaches using the computer algorithm GDFolf2 for analyzing geometric information of amino acids/amino structure (i.e., primary, secondary, tertiary, quintenary) using loss functions [page 5 figure 2] and generating decoy structures [page 21 figure S4]. One of ordinary skill in the art would recognize that loss functions of Mi can be rewritten by users to their own definitions and requirements for generating all-atom protein structures [page 4 second para]. Here, one of ordinary skill in the art would be motivated to combine Yang in view of Xue in view of Pritzel in view of Lee in view of Mi because Yang, Xue, Pritzel, and Lee utilize loss functions but Mi teaches that loss function can be rewritten upon user parameters. Thus, there is a reasonable expectation of success that rewriting and/or creating a loss function using geometric information as taught by Mi and incorporating the “rewritten” geometric loss function as a loss function into the method of Yang, Xue, Pritzel, and Lee would construct a predictable method using a geometric loss function for predicting protein properties that can be utilized to synthesize a physical protein.
Claim(s) 7-8 and 16-17 are rejected under 35 U.S.C. 103 as being unpatentable over Yang in view of Xue in view of Pritzel in view of Lee in view of Mi, as applied to claims 1-6, 11-15, and 20 above, and in further view of Wu et al. (Vector-quantized Masked Auto-encoders on Molecular Surfaces: Fang Wu: International Conference on Machine Learning,2024, https://openreview.net/forum?id=szxtVHOh0C; Cited in the Office Action mailed 21 November 2024).
Yang in view of Xue in view of Pritzel in view of Lee in view of Mi teach claims 1-6, 11-15, and 20.
Yang, Xue, Pritzel, Lee, and Mi teach a method using language and transformer models, autoencoders, and masking and unmasking techniques for predicting protein properties to generate a protein structure based on the predicted protein properties for subsequent physical synthesis.
Yang in view of Xue in view of Pritzel in view of Lee in view of Mi, and do not teach claims 7-8 and 16-17.
Wu teaches using VQMAE autoencoder [abstract]. Wu teaches a framework with discrete latent representations by enforcing and parameterizing the posterior distribution of latent variables to be categorical, which facilitates the feature extractor to acquire more condensed semantics [page 2, left col]. Wu teaches establishes a surface pattern codebook to enforce a discrete posterior distribution of latent variables and achieve more condensed semantics [abstract] and recites we randomly mask a portion of patches and replace them with relaxed codebook vectors. Both visible token embeddings and sampled codebook vectors are forwarded to SurfFormer to gain a global point cloud understanding [page 3 figure 1]. Wu teaches using sampled latent vectors when tokenizing [page 5 left tokenization], as in claims 7 and 16.
Wu et al. (Wu) teach establishes a surface pattern codebook to enforce a discrete posterior distribution of latent variables and achieve more condensed semantics [abstract] and recites we randomly mask a portion of patches and replace them with relaxed codebook vectors. Both visible token embeddings and sampled codebook vectors are forwarded to SurfFormer to gain a global point cloud understanding [page 3 figure 1], as in claim 8 and 17.
It would have been obvious to one of ordinary skill in the art by the effective filing date of the claimed invention to modify Yang in view of Xue in view of Pritzel in view of Lee in view of Mi, and in further view of Wu because Wu teaches vector-quantized masked auto-encoders on molecular surfaces [title]. One of ordinary skill in the art would be motivated to combine Yang in view of Xue in view of Pritzel in view of Lee in view of Mi, and in further view of Wu because Wu teaches using codebooks in the Surface-VQMAE pipeline. One of ordinary skill in the art would expect a reasonable success combining Yang in view of Xue in view of Pritzel in view of Lee in view of Mi, and in further view of Wu because Wu teaches enhancing the prevalent masked auto-encoder (MAE) with the vector quantization (VQ) technique, which establishes a surface pattern codebook to enforce a discrete posterior distribution of latent variables and achieve more condensed semantics [abstract]. Therefore, combining Yang in view of Xue in view of Pritzel in view of Lee in view of Mi, and in further view of Wu would yield a method using encoded latent structures and learned codebooks for predicting protein properties and synthesizing a physical protein based on the said protein properties.
Claim(s) 9-10 and 18-19 are rejected under 35 U.S.C. 103 as being unpatentable over Yang in view of Xue in view of Pritzel in view of Lee in view of Mi in view of Wu, as applied to claims 1-8 11-17, and 20 above, and in further view of Pritzel et al. (2023) (US Patent Pub: US 2023/0410938; Patent Pub Date: 21 December 2025).
Yang in view of Xue in view of Pritzel in view of Lee in view of Mi in view of Wu teach claims1-8 11-17, and 20.
Yang, Xue, Pritzel, Lee, and Mi teach a method using language and transformer models, autoencoders, and masking and unmasking techniques for predicting protein properties to generate a protein structure based on the predicted protein properties for subsequent physical synthesis.
Yang in view of Xue in view of Pritzel in view of Lee in view of Mi in view of Wu do not teach claims 7-8 and 16-17.
Pritzel et al. (Pritzel (2023)) discloses geometric attention block [figure 8]. Pritzel discloses “An example architecture of the folding neural network that implements a geometric attention mechanism is described with reference to FIG. 8.” [Pritzel, disclosure page 9 left col para 0119], as in instant claims 9 and 18.
With respect to claims 10 and 19, the claims are render obvious because Pritzel (2023) discloses “if the predicted structure parameters define predicted spatial locations of each atom in each amino acid of the protein, then the structure loss may be an average error (e.g., squared-error) between: (i) the predicted spatial locations of the atoms, and (ii) the target (e.g., ground truth) spatial locations of the atoms.” [disclosure page 9 right para 0124]. Mi teaches “extracted the following statistical parameters: peptide bond length, distance between neighboring residues, angles between the adjacent covalent bonds within the peptide plane, van der Waals radii of backbone atoms, vectors of carbonyl oxygen (O) and hydrogen (H) atoms in the peptide plane coordinate system and residue coordinate systems of common amino acids [page 6 statistical constraints]. Mi teaches “A) Evolution of the loss function and the protein structure in the folding process of GDFold2. The horizontal axis represents the number of optimization steps and the vertical axis represents the sum value of all applied losses. At steps 400 and 800, the statistical loss and van der Waals loss are added to the optimization process, respectively, where the insets present the gradually refined local backbone structures by these two loss functions.” Mi teaches “The folding process of GDFold2 can be divided into three main stages (Figure 4A), separated by the step points at which new loss terms are applied. The first stage optimizes the backbone structure of the target protein using the loss functions designed for the predicted geometric constraints” [page 11 first para, figure 4]. Thus, Pritzel (2023) and Mi would construct a geometric loss function to which can be used to model a determined error loss based on bond vectors in a protein structure and a ground truth structure.
It would have been obvious to one of ordinary skill in the art by the effective filing date of the claimed invention to modify Yang in view of Xue in view of Pritzel in view of Lee in view of Mi in view of Wu, and in further view of Pritzel (2023) because Pritzel (2023) discloses using geometric attention blocks and mechanism for predicting parameters and final protein structure [figure 8]. One of ordinary skill in the art would recognize that although Pritzel (2023) does not disclose using autoencoders the geometric attention block and mechanism of Pritzel (2023) can be incorporated into the system of Yang in view of Xue in view of Pritzel in view of Lee in view of Mi to process embeddings and other input data/parameters that can be used for predicting protein properties. Thus, there is a reasonable expectation of success that combining the method of Yang in view of Xue in view of Pritzel in view of Lee in view of Mi with the geometric attention mechanisms and blocks of Pritzel (2023) would yield a method that can use geometric attention mechanisms and blocks, language encoders, decoders, transformers, and autoencoders (i.e., biological language reasoning machine learning model) for processing amino acid sequence and structure embeddings for predicting protein properties that can be subsequently utilized for synthesizing a physical protein based on said predicted protein properties.
Pertinent Prior Art
Mi, Tianyu, Nan Xiao, and Haipeng Gong. “GDFold2: A Fast and Parallelizable Protein Folding Environment with Freely Defined Objective Functions.” Protein science 34.2 (November 11, 2024): e70041-n/a. Web.
It is noted this article is a revision of the article the Mi posted-on 14 March 2025.
Conclusion
Claims 1-20 are rejected.
No claims are allowed.
Finality
This Office action is a Non-Final action. A shortened statutory period for reply to this action is set to expire THREE MONTHS from the mailing date of this action.
Inquiries
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JOSEPH C PULLIAM whose telephone number is (571)272-8696. The examiner can normally be reached 0730-1700 M-F.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Karlheinz Skowronek can be reached at (571) 272-9047. 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.
/J.C.P./Examiner, Art Unit 1687
/Karlheinz R. Skowronek/Supervisory Patent Examiner, Art Unit 1687