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 .
Claim Status
Claims 1–19 are pending.
Claims 1–19 are rejected.
Priority
Receipt is acknowledged of certified copies of papers required by 37 CFR 1.55.
The instant application claims the benefit of foreign priority to application CN.202111423752.A filed 26 November 2021. As such, the effective filing date of claims 1-19 is 26 November 2021.
Information Disclosure Statement
The IDS was considered by the examiner.
Drawings
The drawings are allowable.
Specification
The disclosure is objected to because of the following informalities:
Page 2, line 2; page 12, line 30; and page 16, line 5: Each listed reference contains a sentence that include the phrase “…the method for predicting protein-protein interaction fusion-represents the amino acid sequence, the function information and the structure information corresponding to individual proteins by the pre-trained protein representation model …”. It is unclear what is meant by “fusion-represents.”
Appropriate correction is required.
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 1-19 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, 10, and 19 recite “acquiring a plurality of proteins to be treated, and an amino acid sequence, function information and structure information corresponding to individual proteins.” It is unclear how one would acquire “an amino acid sequence” from a “plurality of proteins.” This rejection can be overcome by correcting the claim language to clarify the relationship between amino acid sequences and proteins to be treated. The examiner suggests language such as “…acquiring a plurality of proteins to be treated, and the amino acid sequence, function information and structure information corresponding to each of the individual proteins.”
Dependent claims 2-9 and dependent claims 11-18 are rejected for the same reason because they depend from claims 1 and 10 and do not resolve the indefiniteness issues.
Claims 1, 10, and 19 further recite “obtaining a fusion representation vector corresponding to the individual proteins based on the amino acid sequence, the function information and the structure information corresponding to the individual proteins by a pre-trained protein representation model; and inputting the fusion representation vector corresponding to the individual proteins to a protein-protein interaction prediction model, to predict the protein-protein interaction.” It is unclear how one would obtain “a fusion representation vector corresponding to the individual proteins” and a how one would input “the fusion representation vector corresponding to the individual proteins.” This objection can be overcome by correcting he claim language to clarify the relationship between the vector and individual proteins. The examiner suggests language such as “corresponding to each of the individual proteins” in place of each use of the phrase “corresponding to the individual proteins.”
Dependent claims 2-9 and dependent claims 11-18 are rejected for the same reason because they depend from claims 1 and 10 and do not resolve the indefiniteness issues.
Claims 2 and 11 recite “…wherein the pre-trained protein representation model is obtained by: acquiring an amino acid sequence, function information and structure information of a protein; and pre-training the protein representation model based on the amino acid sequence, the function information and the structure information.” It is unclear whether the wherein clause is intended to require acquiring the information and pre-training the protein representation model within the metes and bounds of the claimed invention, or if it is only further limiting the type of information used in the invention such that acquiring and pretraining are not required within the metes and bounds of the invention. As set forth in MPEP 2111.04.I, “wherein” clauses raise the question as to the limiting effect of the language in a claim. As the claims do not recite an active performance of the steps, the metes and bounds of the claims are unclear. The rejection may be overcome by clarifying what steps are required to be performed.
Dependent claims 3-9 and dependent claims 12-18 are rejected for the same reason because they depend from claims 2 and 11 and do not resolve the indefiniteness issues.
Claims 3 and 12 recite “pre-training the protein representation model based on the amino acid sequence, the function information and the structure information comprises one or more of: replacing the function information with a mask character, and pre-training the protein representation model based on the amino acid sequence, the structure information and the protein; replacing the function information and the structure information with a mask character respectively, and pre-training the protein representation model based on the amino acid sequence and the protein; and replacing the structure information with a mask character, and pre-training the protein representation model based on the amino acid sequence, the function information and the protein.” It is unclear how one would pre-train the protein representation model using “the protein.”
Dependent claims 4-8 and dependent claims 13-17 are rejected for the same reason because they depend from claims 3 and 12 and do not resolve the indefiniteness issues.
Claim 4 recites “…wherein the pre-trained protein representation model is obtained further by masking an amino acid to be masked in the amino acid sequence, to obtain a masked amino acid sequence; and pre-training the protein representation model based on the amino acid to be masked, the masked amino acid sequence, the function information and the structure information.” It is unclear whether the wherein clause is intended to require masking an amino acid and pre-training the protein representation model within the metes and bounds of the claimed invention, or if it is only further limiting the type of information used in the invention such that masking and pretraining are not required within the metes and bounds of the invention. As set forth in MPEP 2111.04.I, “wherein” clauses raise the question as to the limiting effect of the language in a claim. As the claims do not recite an active performance of the steps, the metes and bounds of the claims are unclear. The rejection may be overcome by clarifying what steps are required to be performed.
Dependent claims 5 and 6 are rejected for the same reason because they depend from claim 4 and do not resolve the indefiniteness issues.
Claims 4, 5, 13, and 14 recite “…pre-training the protein representation model based on the amino acid to be masked, the masked amino acid sequence, the function information and the structure information.” It is unclear how one would pre-train the model based on the amino acid to be masked.
Dependent claim 6 and dependent claim 15 are rejected for the same reason because they depend from claims 5 and 14 and do not resolve the indefiniteness issues.
Claim 7 recites “…wherein the pre-trained protein representation model is obtained further by: masking a character to be masked in the function information, to obtain masked function information; and pre-training the protein representation model....” It is unclear whether the wherein clause is intended to require masking a character and pre-training the protein representation model within the metes and bounds of the claimed invention, or if it is only further limiting the type of information used in the invention such that masking and pretraining are not required within the metes and bounds of the invention. As set forth in MPEP 2111.04.I, “wherein” clauses raise the question as to the limiting effect of the language in a claim. As the claims do not recite an active performance of the steps, the metes and bounds of the claims are unclear. The rejection may be overcome by clarifying what steps are required to be performed.
Dependent claim 8 is rejected for the same reason because it depends from claim 7 and does not resolve the indefiniteness issues.
Claims 7, 8, 16, and 17 recite “…pre-training the protein representation model based on the character to be masked, the masked function information, the function information and the structure information.” It is unclear how one would pre-train the model based on the character to be masked. Additionally, it is unclear how one would pre-train on both the “masked function information” and the “function information” for a given protein.
Claim 8 further recites “… inputting the masked function information, the function information and the structure information to the protein representation model, to obtain a second fusion representation vector…” It is unclear how one would pre-train the protein representation model by inputting both the “masked function information” and the “function information” for a given protein.
Claims 9 and 18 recite “…wherein the structure information is obtained by: acquiring a structure file for the protein; extracting point cloud composed of heavy atoms of the protein from the structure file; determining barcode information of a topological complex of the protein according to the point cloud; and discretizing the barcode information, to obtain the structure information of the protein.” It is unclear whether the wherein clause is intended to require acquiring a structure file, extracting point cloud, determining barcode information, and discretizing the barcode information within the metes and bounds of the claimed invention, or if it is only further limiting the type of information used in the invention such that the above listed steps regarding structure information are not required within the metes and bounds of the invention. As set forth in MPEP 2111.04.I, “wherein” clauses raise the question as to the limiting effect of the language in a claim. As the claims do not recite an active performance of the steps, the metes and bounds of the claims are unclear. The rejection may be overcome by clarifying what steps are required to be performed.
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 1-19 are rejected under 35 U.S.C. 101 because the claimed invention is directed to abstract ideas without significantly more.
Step 2A, Prong 1
In accordance with MPEP § 2106, claims found to recite statutory subject matter (Step 1: YES) are then analyzed to determine if the claims recite any concepts that equate to an abstract idea, law of nature or natural phenomenon (Step 2A, Prong 1). In the instant application, the claims recite the following limitations that equate to an abstract idea:
Claim 1 recites a method for predicting protein-protein interactions, comprising acquiring the amino acid sequences, function information, and structure information for a plurality of proteins; obtaining a fusion representation vector based on the protein sequence, structure, and function; and inputting the vectors to predict protein-protein interactions.
Claim 2 recites a protein representation model comprising acquiring protein sequence, structure, and function information and using that to pre-train a protein representation model.
Claim 3 recites pre-training the protein representation model by masking the structure and/or function information and then pre-training based on the sequence and unmasked structure and/or function information.
Claim 4 recites pre-training the protein representation model by masking an amino acid sequence and pre-training with the masked sequence and the unmasked structure and/or function information.
Claim 5 recites obtaining a first fusion vector from the masked amino acid sequence, the structure information, and the function information, determining an amino acid result, and pre-training the protein representation model.
Claim 6 recites that obtaining a first fusion vector comprises determining character and position vectors corresponding to characters in the masked amino acid sequence, structure information and masked function information to obtain a combined vector, and inputting the combined vector.
Claim 7 recites the protein representation model comprises masking a character in the masked function information and pre-training the protein representation model based on the masked character, masked function information, function information, and structure information.
Claim 8 recites inputting the masked function information, the function information, and the structure information to obtain a second fusion representation vector, and pre-training the protein representation model.
Claim 9 recites acquiring a structure file for the protein, extracting point cloud of heavy atoms from the structure file, determining barcode information according to the point cloud, and discretizing the barcode to obtain protein structure information.
Claims 10-18 recite an electronic device for the methods of claims 1-9, respectively.
Claim 19 recites a non-transitory computer readable medium for the method of claim 1.
The limitations for the described methods of pre-training a protein representation model and predicting protein-protein interactions are evaluations or judgements that can be made through mental observations or mathematical calculations which fall under the “mental processes” and “mathematical concepts” groupings of abstract ideas.
While claims 10-19 recite performing some aspects of the methods using an electronic
device or using non-transitory computer-readable media, there are no additional limitations that indicate that the processor and code require anything other than carrying out the recited mental process or mathematical concept in a generic computer environment. Merely reciting that a mental process is being performed in a generic computer environment does not preclude the steps from being performed practically in the human mind or with pen and paper as claimed. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation on generic computer components, then it falls into the “mental processes” grouping of abstract ideas. As such, claims 1-19 recite abstract ideas (Step 2A, Prong 1: YES).
Step 2A, Prong 2
Claims found to recite a judicial exception under Step 2A, Prong 1 are then further
analyzed to determine if the claims as a whole integrate the recited judicial exception into a
practical application or not (Step 2A, Prong 2). This judicial exception is not integrated into a
practical application because the claims do not recite an additional element that reflects an
improvement to technology or applies or uses the recited judicial exception in some other
meaningful way. Rather, the instant claims recite additional elements that amount to mere
instructions to implement the abstract idea or insignificant extra-solution activity. Specifically,
the claims recite the following additional elements:
Claim 10 and dependent claims 11-18 recite performing the abstract steps with at least one processor.
Claim 19 recites performing the abstract steps with a non-transitory computer-readable medium.
There are no limitations that indicate that the processors or non-transitory computer-readable
medium require anything other than a generic computing system. As such, these limitations
equate to mere instructions to implement the abstract ideas on a generic computer that the courts
have stated do not render an abstract idea eligible in Alice Corp., 573 U.S. at 223, 110 USPQ2d
at 1983. See also 573 U.S. at 224, 110 USPQ2d at 1984.
The above recited additional elements do not provide a practical application of the recited
judicial exception. As such, claims 1-19 are directed to an abstract idea (Step 2A, Prong 2: NO).
Step 2B
Claims found to be directed to a judicial exception are then further evaluated to determine if the claims recite an inventive concept that provides significantly more than the judicial exception itself (Step 2B). The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the claims recite additional elements that equate to mere instructions to apply the recited exception in a generic computing environment.
As discussed above, there are no additional limitations to indicate that the claimed
processor or non-transitory computer readable medium require anything other than generic
computer components in order to carry out the recited abstract ideas in the claims. Claims that
amount to nothing more than an instruction to apply the abstract idea using a generic computer
do not render an abstract idea eligible. Alice Corp., 573 U.S. at 223, 110 USPQ2d at 1983. See
also 573 U.S. at 224, 110 USPQ2d at 1984.
The additional elements do not comprise an inventive concept when considered individually or as an ordered combination that transforms the claimed judicial exception into a
patent-eligible application of the judicial exception. Therefore, the claims do not amount to significantly more than the judicial exception itself (Step 2B: No). As such, claims 1-19 are not patent eligible.
Claim Interpretation
For the purposes of applying prior art:
Claims 1, 10, and 19 are interpreted as “…acquiring a plurality of proteins to be treated, and the amino acid sequence, function information and structure information corresponding to each of the individual proteins; obtaining a fusion representation vector corresponding to each of the individual proteins based on the amino acid sequence, the function information and the structure information corresponding to each of the individual proteins by a pre-trained protein representation model; and inputting the fusion representation vector corresponding to each of the individual proteins to a protein-protein interaction prediction model, to predict the protein-protein interaction.”
Claims 2 and 11 are interpreted as “…wherein the pre-trained protein representation model comprises: acquiring an amino acid sequence, function information and structure information of a protein; and pre-training the protein representation model based on the amino acid sequence, the function information and the structure information.”
Claims 3 and 12 are interpreted as: “…pre-training the protein representation model based on the amino acid sequence, the function information and the structure information comprises one or more of: replacing the function information with a mask character, and pre-training the protein representation model based on the amino acid sequence and the structure information; replacing the function information and the structure information with a mask character respectively, and pre-training the protein representation model based on the amino acid sequence; and replacing the structure information with a mask character, and pre-training the protein representation model based on the amino acid sequence and the function information.”
Claim 4 is interpreted as: “…wherein the pre-trained protein representation model comprises: masking an amino acid to be masked in the amino acid sequence, to obtain a masked amino acid sequence...”
Claims 4, 5, 13, and 14 are interpreted as “…pre-training the protein representation model based on the masked amino acid sequence, the function information and the structure information.”
Claim 7 is interpreted as: “… wherein the pre-trained protein representation model comprises: masking a character to be masked in the function information, to obtain masked function information…”
Claims 7, 8, 16, and 17 are interpreted as “…pre-training the protein representation model based on the masked function information.”
Claim 8 is interpreted as “… inputting the masked function information and the structure information to the protein representation model, to obtain a second fusion representation vector…”
Claims 9 and 18 are interpreted as “…wherein the structure information comprises: acquiring a structure file for the protein…”
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.
Claims 1, 10, and 19 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Jha et al. (IJCNN, 18 July 2021, pages 1-8) (Herein referred to as Jha.)
Regarding claims 1, 10, and 19, Jha teaches “…a deep multi-modal architecture that exploits multiple sources of information about proteins, which helps in predicting PPI [protein-protein interaction].” (Page 1, right column, lines 33-35) Jha further teaches “...framework that incorporates structural information and GO[gene ontology]-based information…[and]… have also added sequence-based information. (Page 1, right column, lines 35-37 and page 2, left column, lines 3-4) Gene ontology (GO) is known to be a method of determining and classifying by protein function. Jha also teaches that “[t]he proposed multi-modal architecture integrates several deep learning techniques such as a pre-trained residual network…” (Page 5, left column, lines 16-18) Additionally, Jha teaches “[t]he feature vectors (Fstruc and FGO) from different modalities (structural and GO-based) are merged and the merged vector is fed to a batch normalization layer (fBN). The output of fBN is input to the fully connected layer (fFC), followed by a sigmoid layer to classify PPI.” (Page 3, right column, lines 25-29) Jha further teaches that they “…add sequence (QSO) modality to our bi-modal approach…” (Page 6, right column, lines 1-2) See also Tables III and IV on page 4 for references to the “Trimodal: Structural+ GO + QSO” vectors.
Regarding claims 10, the instant application claims using an electronic device comprising of at least one processor and memory to execute the methods of claim 1. The court held that broadly providing an automatic or mechanical means to replace a manual activity which accomplished the same result is not sufficient to distinguish over the prior art. See MPEP 2144.04 III.
Regarding claim 19, the instant application claims using a non-transitory computer- readable medium to implement the methods of claim 1. The court held that broadly providing an automatic or mechanical means to replace a manual activity which accomplished the same result is not sufficient to distinguish over the prior art. See MPEP 2144.04 III.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 2 and 11 are rejected under 35 U.S.C. 103 as being unpatentable over Jha in view of Lanchantin et al. (BCB 21, 1 August 2021, pages 1-10) (Herein referred to as Lanchantin.)
Jha teaches the methods of claim 1 and 10 as rejected above under 35 U.S.C. 102(a)(1). (Page 1, right column, lines 33-35; Page 1, right column, lines 35-37 and page 2, left column, lines 3-4; Page 5, left column, lines 16-18; Page 3, right column, lines 25-29; Page 6, right column, lines 1-2; and Page 4, Tables III and IV.) Jha further teaches pre-training based on structure by stating “[t]he model is pre-trained on PubMed abstracts to add semantics to those words based on their use in biomedical literature…[and]… pre-trained Word2Vec model is updated to generate the feature vectors…for biological entities (proteins).” (Page 2, right column, lines 46-51) Jha additionally taches pre-training based on structure using “a pre-trained model, i.e., ResNet50 (fRN) (trained on ImageNet dataset), to get an initial feature vector (OpRN) from volumetric descriptions of a protein p.” (Page 3, left column, lines 14-17) Jha also teaches pre-training based on function and structure in Figure 2. (Page 4)
Jha does not teach pre-training based on the amino acid sequence.
Lanchantin teaches to “…pretrain the network on the Masked Language Model (MLM) task from a large repository of unlabeled protein sequences…[and]…further pretrain the network on a set of Structure Prediction (SP) tasks including secondary structure (SS), contact (CT), and remote homology (RH). (Page 3, Figure 3) Lanchantin further teaches “to pretrain the DeepVHPPI using Masked Language Model (MLM) pretraining …[and]…to further pretrain the network using Structure Prediction (SP) to learn 3D structural representations…” (Page 4, right column, lines 25-29)
It would have been prima facie obvious to one of ordinary skill in the art at the effective
filing date of the invention to have applied the method of Lanchantin to pre-train using amino acid sequence in the method of Jha. Lanchantin teaches “[p]retraining the network allows it to learn representations that transfer well to the PPI task for novel (i.e., unseen) virus sequence.” (Page 4, right column, lines 31-33) Lanchantin further teaches “[r]ecent literature on learning self-supervised representations of natural language have shown that pretraining using self-supervised and supervised methods encourages the model to learn semantics about the input domain that can help prediction accuracy on new tasks.” (Pages 4, right column, lines 38-41 and page 5, left column, line 1) Therefore, one of ordinary skill in the art would have been motivated to incorporate pre-training based on amino acid sequence to a method that also pre-trains based on protein structure and protein function. The invention is therefore prima facie obvious.
Claims 3, 7, 8, 16, and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Jha and Lanchantin as applied to claims 2 and 11 above, and further in view of Rives et al. (PNAS, 5 April 2021, pages 1-12) (Herein referred to as Rives.)
Jha and Lanchantin do not teach pre-training the protein representation model after replacing the function information with a mask character, replacing the function information and the structure information with a mask character respectively, and replacing the structure information with a mask character.
Rives teaches “[t]he idea that biological function and structure are recorded in the statistics of protein sequences selected through evolution has a long history” and “[u]nlocking the information encoded in protein sequence variation is a longstanding problem in biology.” (Page 1, left column, lines 35-37 and 43-44) Rives further teaches “…inferring biological structure and function from evolutionary statistics has motivated development of machine
learning on individual sequence families.” (page 2, left column, lines 10-12) Rives discloses “…models using the masked language modeling objective…Each input sequence is corrupted by replacing a fraction of the amino acids with a special mask token.” (Page 2, right column, lines 3-5) Rives additionally teaches “[t]he interchangeability of amino acids within a given structural or functional context in a protein depends on their biochemical properties. Self-supervision can
be expected to capture these patterns to build a representation space that reflects biochemical knowledge.” (Page 3, right column, lines 26-30) Rives attempts to indirectly mask protein structure and protein function through the masking of protein sequence.
It would have been prima facie obvious to one of ordinary skill in the art at the effective filing date of the invention to have applied the method of Rives to indirectly mask protein structure and protein function to the methods of Jha and Lanchantin. Rives teaches “[l]earning biological properties from sequence data is a logical step toward generative and predictive artificial intelligence for biology…[and] information emerges in the learned representations on fundamental properties of proteins such as secondary structure, contacts, and biological activity.” (Page 1, “Significance”, lines 1-3 and 6-8) Rives further teaches the method “…observes patterns in the sequences of its training data that are determined by structure.” (page 5, right column, lines 6-8) and “[t]he detail and coverage of these experiments provides a view into the mutational fitness landscape of individual proteins, giving quantitative relationships between sequence and protein function.” (Page 9, left column, lines 24-27) Rives additionally discloses, “[i]f neural networks can transfer knowledge learned from protein sequences to design functional proteins, this could be coupled with predictive models to jointly generate and optimize sequences for desired functions.” (Page 11, left column, lines 6-9) Rives’ method allows for the analysis of impacts to protein structure and function through the masking of amino acids. Therefore, one of ordinary skill in the art would have been motivated to directly mask protein structure and protein function when modeling protein-protein interactions. The invention is therefore prima facie obvious.
Claims 4 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Jha, Lanchantin, and Rives as applied to claim 3 and 12 above, and further in view of additional teachings of Lanchantin and Rives.
Jha does not teach masking amino acids.
Lanchantin teaches “…using Masked Language Model (MLM) pretraining in order to learn generic representations from unlabeled protein sequences.” (Page 4, right column, lines 25-27) Lanchantin discloses this MLM “…is a self-supervised technique to allow a model to build rich representations of sequences… replacing the true character with a dummy mask character.” (Page 5, left column, lines 4 and 9-10) Rives teaches “Each input sequence is corrupted by replacing a fraction of the amino acids with a special mask token.” (Page 2, right column, lines 3-5)
It would have been prima facie obvious to one of ordinary skill in the art at the effective filing date of the invention to have applied the methods of Lanchantin and Rives to mask amino acids in a protein sequence to the method of Jha and additional methods of Lanchantin and Rives. Lanchantin teaches “…deep learning methods are well-positioned to aid and augment biological experiments, hoping to help identify more accurate virus-host protein interaction maps…[and] computational methods can quickly adapt to predict how virus mutations change protein interactions with the host proteins.” (Page 1, left column, lines 7-11) As mentioned above in Rives, masking amino acids enhances protein modeling techniques. Therefore, one of ordinary skill in the art would have been motivated to mask amino acids when modeling protein-protein interactions. The invention is therefore prima facie obvious.
Claims 5 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Jha, Lanchantin, and Rives as applied to claim 4 and 13 above, and further in view of Bepler and Berger. (Cell Systems, 16 June 2021, pages 654-699, e1-e3) (Herein referred to as Bepler.)
Jha, Lanchantin, and Rives do not teach determining an amino acid predicting result.
Bepler teaches “…randomly replace 10% of the amino acids in a sequence with either an auxiliary mask token or a uniformly random draw from the amino acids and train our model to predict the original amino acids at those positions.” (Page e1, Method Details, “Masked language modeling module” paragraph, lines 1-3)
It would have been prima facie obvious to one of ordinary skill in the art at the effective filing date of the invention to have applied the method of Bepler to predict masked amino acids to the methods of Jha, Lanchantin, and Rives. Bepler teaches “[t]he drawback to this formulation [autoregressive language model] is that the representations learned for each position depend only on preceding positions, potentially making them less useful as contextual representations. The masked position prediction formulation (also known as masked language modeling) addresses this problem by considering the probability distribution over each token at each position conditioned on all other tokens in the sequence.” (Page 658, left column, lines 4-10) Therefore, one of ordinary skill in the art would have been motivated to determine an amino acid predicting result when modeling protein-protein interactions. The invention is therefore prima facie obvious.
Claims 6 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Jha, Lanchantin, Rives, and Bepler as applied to claim 5 and 14 above, and further in view of Rives.
Jha, Lanchantin, and Bepler do not teach determining character and position vectors or combining said character and position vectors to obtain a combined vector.
Rives additionally teaches “[t]he Transformer model…[c]onsists of an embedding step with token E(x) and positional H(x) embeddings, followed by K layers of Transformer blocks, before a projection W to log probabilities.” (Supplemental Information, Page 3, Section 3: “The Transformer”, “Architecture”, lines 20 and 25-26) This approach of creating an input vector using the sum of character vectors and position vectors in known to be an improvement in the art. Devlin et al (arXiv, 11 October 2018, pages 1-16) teaches “[f]or a given token, its input representation is constructed by summing the corresponding token, segment and position embeddings.” (Page 3, right column, lines 22-25) Devlin et al further teaches “[o]ur input representation is able to unambiguously represent both a single text sentence or a pair of text sentences…in one token sequence.” (Page 3, right column, lines 19-21)
It would have been prima facie obvious to one of ordinary skill in the art at the effective filing date of the invention to have applied the additional teachings of Rives to create the fusion representation vector with the methods of Jha, Lanchantin, Rives, and Bepler. As disclosed by Devlin et al, their method “…addresses the previously mentioned unidirectional constraints by proposing a new pre-training objective…” (Page 1, right column, lines 38-40) Therefore, one of ordinary skill in the art would have been motivated to determine a character vector and a position vector for use in pre-training a protein representation model and when modeling protein-protein interactions. The invention is therefore prima facie obvious.
Claims 9 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Jha and Lanchantin as applied to claims 2 and 11 above, and further in view of Wang et al. (Nature Machine Intelligence, 14 February 2020, pages 116-123) (Herein referred to as Wang.)
Jha and Lanchantin do not teach extracting a point cloud of heavy atoms, determining barcode information of a topological complex of the protein based on the point cloud, or discretizing the barcode information to obtain the structure information of the protein.
Wang teaches “[a] topological representation should be able to extract patterns of different biological or chemical aspects (for example, hydrogen bonds between oxygen and nitrogen atoms, …) from a PPI system that is represented by a set of atomic coordinates (that is, a point cloud). We construct simplicial complexes using selected subsets of atomic coordinates and modified distance matrices to achieve this goal.” (Page 117, right column, lines 7-13) Wang further teaches “[u]sing persistent homology, the original 3D point-cloud data are characterized by topological barcodes that are represented as collections of intervals that capture geometric patterns, topological patterns and PPIs while dramatically simplifying complicated structural representations of a PPI complex.” (Page 117, right column, lines 35-40) Additionally, Wang teaches to “…construct feature vectors from these sets of intervals [from the barcodes] for machine learning models. One method of vectorization is to discretize the range of the filtration parameter into bins and record the behaviour of the bar codes in each bin.” (Page 117, right column, lines 46-50)
It would have been prima facie obvious to one of ordinary skill in the art at the effective filing date of the invention to have applied the teachings of Wang to pre-train the protein representation model with the methods of Jha and Lanchantin. As stated above, Wang discloses using persistent homology is known for “…dramatically simplifying complicated structural representations of a PPI complex.” (Page 117, right column, lines 38-40) Wang further discloses “…topological barcodes for different samples are in the same range of filtration values, which improves the scalability in comparison with the direct use of the original 3D data.” (Page 117, right column, lines 44-46) Wang teaches that “[o]ne advantage of binned barcode vectorization is that it keeps the distance information that reflects the strength of hydrogen bonds, van der Waals interactions and so on. The bin representation of barcode features can be easily incorporated into a CNN [convolutional neural network], which captures and discriminates local patterns; that is, the impact of mutations.” (Page 117, right column, lines 59-63) Therefore, one of ordinary skill in the art would have been motivated to use persistent homology to store protein structure information in barcodes based on point clouds and to discretize the barcode information. These methods are known to improve understanding of protein structure and simplify representations of protein-protein interactions. The invention is therefore prima facie obvious.
Regarding claims 11-18, the instant application claims using an electronic device comprising of at least one processor and memory to execute the methods of claims 2-18. The court held that broadly providing an automatic or mechanical means to replace a manual activity which accomplished the same result is not sufficient to distinguish over the prior art. See MPEP 2144.04 III.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Lei et al teaches a deep learning framework for multi-level peptide-protein interaction prediction. (Nature Communications, September 2021) Hu et al teaches a survey of efforts made towards the development of effective computational models for protein-protein interaction prediction. (Briefings in Bioinformatics, March 2021) Das and Chakrabarti teach a machine learning algorithm to classify and predict protein-protein interaction interfaces. (Nature: Scientific Reports, January 2021) Bramer and Wei teach atom-specific persistent homology to provide a local atomic level representation of a molecule via a global topological tool. (Computational and mathematical biophysics, February 2020) Dutta and Saha teach multimodal approaches for PPI identification using protein sequence, structure, and textual information. (Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, July 2020) Filipavicius et al teach using pre-training masked language models to study protein-protein binding interactions. (arXiv, December 2020) Sledzieski et al teach a deep learning method for predicting a physical interaction between two proteins based on the protein sequences. (bioRxiv, January 2021) Xiao et al teach using large-scale language models to model evolutionary-scale protein sequences, encoding protein biology information in representation. (arXiv, August 2021) Vig et al teach analyzing Transformer protein models through the lens of attention, and present a set of interpretability methods that capture the unique functional and structural characteristics of proteins. (arXiv, March 2021)
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SHARON LEVINE GRAFF whose telephone number is (571)317-0219. The examiner can normally be reached Mon - Fri 7:30 AM - 4 PM.
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.
/S.L.G./
Examiner, Art Unit 1687
/Karlheinz R. Skowronek/Supervisory Patent Examiner, Art Unit 1687
Prosecution Insights