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 .
Claim Status
Claims 1-24 are currently pending and under examination herein.
Claims 1-24 are rejected.
Priority
The instant application claims priority to 63314898 filed 28 February 2022 and 63390663 filed 20 July 2022. In this action, claims 1-24 are examined as though they had an effective filing date of 28 February 2022. In future actions, the effective filing date of one or more claims may change, due to amendments to the claims, or further analysis of the disclosure(s) of the priority application(s).
Information Disclosure Statement
No Information Disclosure Statement (IDS) was found amongst the submitted documents.
Drawings
The drawings filed 28 February 2023 with replacements filed 27 June 2023 are accepted.
Specification
The specification is objected to because it contains amino acid sequences without sequence IDs at least on Page 15 and 16 (Tables 2 and 3). All amino acid sequences of 4 amino acids or more must have sequence IDs accompanied by a sequence listing (see Nucleotide and/or Amino Acid Sequence Disclosures below).
Nucleotide and/or Amino Acid Sequence Disclosures
Summary of Requirements for Patent Applications Filed On Or After July 1, 2022, That Have Sequence Disclosures
37 CFR 1.831(a) requires that patent applications which contain disclosures of nucleotide and/or amino acid sequences that fall within the definitions of 37 CFR 1.831(b) must contain a “Sequence Listing XML”, as a separate part of the disclosure, which presents the nucleotide and/or amino acid sequences and associated information using the symbols and format in accordance with the requirements of 37 CFR 1.831-1.835. This “Sequence Listing XML” part of the disclosure may be submitted:
1. In accordance with 37 CFR 1.831(a) using the symbols and format requirements of 37 CFR 1.832 through 1.834 via the USPTO patent electronic filing system (see Section I.1 of the Legal Framework for Patent Electronic System (https://www.uspto.gov/PatentLegalFramework), hereinafter “Legal Framework”) in XML format, together with an incorporation by reference statement of the material in the XML file in a separate paragraph of the specification (an incorporation by reference paragraph) as required by 37 CFR 1.835(a)(2) or 1.835(b)(2) identifying:
a. the name of the XML file
b. the date of creation; and
c. the size of the XML file in bytes; or
2. In accordance with 37 CFR 1.831(a) using the symbols and format requirements of 37 CFR 1.832 through 1.834 on read-only optical disc(s) as permitted by 37 CFR 1.52(e)(1)(ii), labeled according to 37 CFR 1.52(e)(5), with an incorporation by reference statement of the material in the XML format according to 37 CFR 1.52(e)(8) and 37 CFR 1.835(a)(2) or 1.835(b)(2) in a separate paragraph of the specification identifying:
a. the name of the XML file;
b. the date of creation; and
c. the size of the XML file in bytes.
SPECIFIC DEFICIENCIES AND THE REQUIRED RESPONSE TO THIS NOTICE ARE AS FOLLOWS:
Specific deficiency - This application fails to comply with the requirements of 37 CFR 1.831-1.834 because it does not contain a “Sequence Listing XML” as a separate part of the disclosure. A “Sequence Listing XML” is required because amino acid sequences are listed in the specification (see objection to the specification above).
Required response - Applicant must provide:
• A “Sequence Listing XML” part of the disclosure, as described above in item 1. or 2.; together with
o A statement that indicates the basis for the amendment, with specific references to particular parts of the application as originally filed, as required by 37 CFR 1.835(a)(3);
o A statement that the “Sequence Listing XML” includes no new matter as required by 37 CFR 1.835(a)(4)
AND
• A substitute specification in compliance with 37 CFR 1.52, 1.121(b)(3), and 1.125 inserting the required incorporation by reference paragraph as required by 37 CFR 1.835(a)(2), consisting of:
o A copy of the previously-submitted specification, with deletions shown with strikethrough or brackets and insertions shown with underlining (marked-up version);
o A copy of the amended specification without markings (clean version); and
o A statement that the substitute specification contains no new matter.
Specific deficiency - Sequences appearing in the specification are not identified by sequence identifiers (i.e., “SEQ ID NO:X” or the like) in accordance with 37 CFR 1.831(c).
Required response – Applicant must provide:
A substitute specification in compliance with 37 CFR 1.52, 1.121(b)(3), and 1.125 inserting the required sequence identifiers, consisting of:
• A copy of the previously-submitted specification, with deletions shown with strikethrough or brackets and insertions shown with underlining (marked-up version);
• A copy of the amended specification without markings (clean version); and
• A statement that the substitute specification contains no new matter.
Claim Interpretation
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The claims in this application are given their broadest reasonable interpretation 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. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph:
(A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and
(C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.
Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts to entirely perform the recited function.
Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites function without reciting sufficient structure, material or acts to entirely perform the recited function.
Claim limitations in this application that use the word “means” (or “step”) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Conversely, claim limitations in this application that do not use the word “means” (or “step”) are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action.
This application includes one or more claim limitations that do not use the word “means,” but are nonetheless being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, because the claim limitation(s) uses a generic placeholder that is coupled with functional language without reciting sufficient structure to perform the recited function and the generic placeholder is not preceded by a structural modifier. Such claim limitation(s) is: semi-supervised learning module in claim 12.
Because this/these claim limitation(s) is/are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, it/they is/are being interpreted to cover the corresponding structure described in the specification as performing the claimed function, and equivalents thereof. However, no indication of the structure of the semi-supervised learning module could be found within the specification or the drawings (see 112(a) and 112(b) rejections).
If applicant does not intend to have this/these limitation(s) interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitation(s) to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitation(s) recite(s) sufficient structure to perform the claimed function so as to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph.
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
Claim 12 rejected under 35 U.S.C. 112(a) as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention. Claim 12 recites a semi-supervised learning module. No description of the structure was found within the claims, specification, or drawings (see 112(f) interpretation above).
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.
In Claim 12, the limitation “a semi-supervised learning module” invokes 35 U.S.C. 112(f) (see claim interpretation above). However, the written description fails to disclose the corresponding structure, material, or acts for performing the entire claimed function and to clearly link the structure, material, or acts to the function (see the 112(a) rejection above). No description of the structure of the module was found within the claims, specification, or drawings. Therefore, the claim is indefinite and is rejected under 35 U.S.C. 112(b).
Applicant may:
(a) Amend the claim so that the claim limitation will no longer be interpreted as a limitation under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph;
(b) Amend the written description of the specification such that it expressly recites what structure, material, or acts perform the entire claimed function, without introducing any new matter (35 U.S.C. 132(a)); or
(c) Amend the written description of the specification such that it clearly links the structure, material, or acts disclosed therein to the function recited in the claim, without introducing any new matter (35 U.S.C. 132(a)).
If applicant is of the opinion that the written description of the specification already implicitly or inherently discloses the corresponding structure, material, or acts and clearly links them to the function so that one of ordinary skill in the art would recognize what structure, material, or acts perform the claimed function, applicant should clarify the record by either:
(a) Amending the written description of the specification such that it expressly recites the corresponding structure, material, or acts for performing the claimed function and clearly links or associates the structure, material, or acts to the claimed function, without introducing any new matter (35 U.S.C. 132(a)); or
(b) Stating on the record what the corresponding structure, material, or acts, which are implicitly or inherently set forth in the written description of the specification, perform the claimed function. For more information, see 37 CFR 1.75(d) and MPEP §§ 608.01(o) and 2181.
For the purposes of examination, the module is interested as a generic computer processing system/software feature of a generic computer processing system.
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 therefore, subject to the conditions and requirements of this title.
Claims 1-24 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. 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 or natural law (Step 2A, Prong 1). Claims 1-17 are directed to methods and Claims 18-24 are directed to systems. In the instant application, the claims recite the following limitations that equate to an abstract idea:
Claim 1 recites the limitation - encoding, using a dilated convolutional encoder, a plurality of input protein sequences onto a latent space distribution; and decoding, using a decoder employing dilated causal convolutions, the latent space distribution to generate new protein sequences different from the input protein sequences. Based on the broadest reasonable interpretation, encoding and decoding the information encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 2 recites the limitation - generating protein sequences using a system including an encoder coupled to an autoregressive generator, and having been trained with a loss function that comprises reconstruction loss and a mutual information maximization term, the method comprising: encoding, using the encoder, a plurality of input protein sequences onto a latent space distribution; and decoding, using the autoregressive generator, the latent space distribution to generate new protein sequences different from the input protein sequences. Based on the broadest reasonable interpretation, encoding and decoding the information, and training with a loss function encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 3 recites the limitation - wherein the decoding step comprises decoding the latent space distribution using a dilated casual convolution autoregressive generator as the autoregressive generator. Based on the broadest reasonable interpretation, decoding the information encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 4 recites the limitation - wherein the decoding step comprises decoding the latent space distribution to generate the new protein sequences, which include sequences of different lengths. Based on the broadest reasonable interpretation, decoding the information encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 5 recites the limitation - wherein the encoding step comprises encoding the plurality of input protein sequences, which are unaligned. Based on the broadest reasonable interpretation, encoding the information encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 6 recites the limitation - wherein the encoding step comprises encoding the plurality of input protein sequences using a dilated convolutional neural network encoder. Based on the broadest reasonable interpretation, encoding the information encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 7 recites the limitation - wherein the encoding step comprises encoding the plurality of input protein sequences into a latent space embedding. Based on the broadest reasonable interpretation, encoding the information encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 8 recites the limitation - wherein the decoding step comprises predicting a next amino acid in a particular sequence, based on the particular sequence and a latent space embedding. Based on the broadest reasonable interpretation, predicting an amino acid encompasses equations and could practically be done by the human mind. This draws the limitation to a mathematical concept and a mental process, which classifies the limitation as an abstract idea.
Claim 9 recites the limitation - wherein the system was trained using the loss function, which further includes a semi-supervised loss. Based on the broadest reasonable interpretation, training with a semi supervised loss encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 10 recites the limitation - wherein the decoding step further comprises decoding the latent space distribution using the dilated casual convolution autoregressive generator, which incorporates residual and skip connections. Based on the broadest reasonable interpretation, decoding information encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 11 recites the limitation - training a system for generating protein sequences, the system including an encoder that encodes a plurality of input protein sequences onto a latent space distribution, and an autoregressive generator that decodes the latent space distribution to generate new protein sequences different from the input protein sequences, the method comprising: training the system with a loss function that comprises reconstruction loss and a mutual information maximization term. Based on the broadest reasonable interpretation, encoding and decoding information and training with a loss function encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 12 recites the limitation - a regression model with a set of training parameters that are learned by minimizing, for a subset of the latent space distribution, an error between outputs of the regression model and fitness values obtained from assay measurements; and the method further comprises training the system with a modified loss function that further includes term based on performance of the regression model. Based on the broadest reasonable interpretation, utilizing regression and training the system encompasses equations and could practically be done by the human mind. This draws the limitation to a mathematical concept and a mental process, which classifies the limitation as an abstract idea.
Claim 13 recites the limitation - wherein the term in the modified loss function is a mean-squared error term based on a ground truth and a predicted regression value of the regression model. Based on the broadest reasonable interpretation, utilizing the mean squared error encompasses equations and could practically be done by the human mind. This draws the limitation to a mathematical concept and a mental process, which classifies the limitation as an abstract idea.
Claim 14 recites the limitation - wherein the regression model is a neural network having weights as the training parameters, which are determined in the training step. Based on the broadest reasonable interpretation, determining weights within the model encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 15 recites the limitation - wherein the encoder is a dilated convolutional neural network encoder. Based on the broadest reasonable interpretation, the encoder encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 16 recites the limitation - wherein the encoder learns a latent space embedding. Based on the broadest reasonable interpretation, encoding encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 17 recites the limitation - wherein the training step comprises training the system with the loss function, which further includes a semi-supervised loss. Based on the broadest reasonable interpretation, training with a function encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 18 recites the limitation - an encoder configured to encode a plurality of input protein sequences onto a latent space distribution; and an autoregressive generator configured to decode the latent space distribution to generate new protein sequences different from the input protein sequences, wherein the system is trained with a loss function that includes reconstruction loss and a mutual information maximization term. Based on the broadest reasonable interpretation, encoding, decoding and training with a function encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 19 recites the limitation - wherein the autoregressive generator is a dilated casual convolution autoregressive generator. Based on the broadest reasonable interpretation, a dilated casual convolution autoregressive generator encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 20 recites the limitation - wherein the autoregressive generator is further configured to decode the latent space distribution to generate the new protein sequences, which include sequences of different lengths. Based on the broadest reasonable interpretation, decoding encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 21 recites the limitation - wherein the encoder is further configured to encode the plurality of input protein sequences, which are unaligned. Based on the broadest reasonable interpretation, encoding encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 22 recites the limitation - wherein the system was trained with the loss function, which further includes a semi-supervised loss. Based on the broadest reasonable interpretation, training with a function encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 23 recites the limitation - wherein the dilated casual convolution autoregressive generator incorporates residual and skip connections. Based on the broadest reasonable interpretation, utilizing residual and skip connections within the model encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
Claim 24 recites the limitation - generating protein sequences using a system including an encoder coupled to an autoregressive generator, and having been trained with a loss function that includes reconstruction loss and a mutual information maximization term, the method comprising: encoding, using the encoder of a variational autoencoder, a plurality of input protein sequences onto a latent space distribution; and decoding, using the autoregressive generator, the latent space distribution to generate new protein sequences different from the input protein sequences. Based on the broadest reasonable interpretation, encoding, decoding, and training with a function encompasses equations. This draws the limitation to a mathematical concept, which classifies the limitation as an abstract idea.
These limitations recite concepts of encoding, decoding, and predicting information and utilizing regression and training functions that are so generically recited that they can be practically performed in the human mind as claimed, which falls under the “Mental processes” and “Mathematical concepts” grouping of abstract ideas. A mathematical concept need not be expressed in mathematical symbols, because words used in a claim operating on data to solve a problem can serve the same purpose as a formula (MPEP 2106.04(a)(2)). Additionally, both product claims and process claims may recite mental processes, which can include a claim that requires a computer (MPEP 2106.04(a)(2)). Therefore, these limitations fall under the “Mental process” and “Mathematical concepts” groupings of abstract ideas. As such, claims 1-24 recite an abstract idea (Step 2A, Prong 1: YES).
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). These judicial exceptions are not integrated into a practical application because the claims do not recite an additional element that reflects an improvement to technology (MPEP § 2106.04(d)(1)). Rather, the claims provide insignificant extra-solution activity (MPEP § 2106.05(g)) and provide mere instructions to apply a judicial exception (MPEP § 2106.05(f)). Specifically, the claims recite the following additional elements:
Claim 12 recites a semi-supervised learning module (see 112(f) and 112(b) above).
Claim 24 recites a non-transitory computer-readable medium storing a program executed by processing circuitry.
There are no limitations that indicate that the claimed encoding, decoding, and predicting information and utilizing regression and training functions require anything other than generic computing systems. As such, these limitations equate to mere instructions to implement the abstract idea on a generic computer that the courts have stated does not render an abstract idea eligible. There is no indication that these steps are affected by the judicial exception in any way and thus do not integrate the recited judicial exception into a practical application. The claims do not recite improving a computer or its functioning. As such, claims 1-24 are directed to an abstract idea (Step 2A, Prong 2: NO).
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 conventional additional elements that equate to mere instructions to apply the recited exception in a generic way or in a generic computing environment. The claims also recite conventional additional elements that represent insignificant extra-solution activities.
As discussed above, there are no additional limitations to indicate that the claimed encoding, decoding, and predicting information and utilizing regression and training functions require anything other than generic computer components in order to carry out the recited abstract idea 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 or natural law eligible. MPEP 2106.05(f) discloses that mere instructions to apply the judicial exception cannot provide an inventive concept to the claims. As specified in MPEP 2106.05(g), extra-solution activities can be understood as incidental to the primary process or product that are merely a nominal or tangential addition to the claim. Insignificant extra-solution activities include mere data gathering, selecting a particular data source or type of data to be manipulated, and displaying information. Additionally, as can be seen from the art cited in the 35 USC 103 rejection below (Hawkins-Hooker et al., Shin et al., Bikard et al., and Alley et al.), using generic computing systems for modeling protein information with machine learning was well-understood, routine, and conventional at the time of the effective filling date.
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 1-24 are not patent eligible.
Claim Rejections - 35 USC § 103
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.
Claim 1 is rejected under 35 U.S.C. 103 as being unpatentable over Hawkins-Hooker et al. (2021, Plos Computational Biology, Vol. 17: 1-23), in view of Bikard et al. (US 20210193259 A1), and in further view of Shin et al. (2021, Nature Communications, Vol. 12: 1-11). Italicized text from reference art.
Applicable claims include:
Claim 1. A method of generating protein sequences, the method comprising: (Claim 1.i) encoding, using a dilated convolutional encoder, a plurality of input protein sequences onto a latent space distribution; and (Claim 1.ii) decoding, using a decoder employing dilated causal convolutions, the latent space distribution to generate new protein sequences different from the input protein sequences.
Regarding Claim 1, Hawkins-Hooker et al. teach (Claim 1.i) encoding, using an encoder, a plurality of input protein sequences onto a latent space distribution (Page 5, Paragraph 3: To handle raw sequences we therefore designed a model incorporating a convolutional encoder. the latent space was used). Hawkins-Hooker et al. teach (Claim 1.ii) decoding, using a decoder, the latent space distribution to generate new protein sequences different from the input protein sequences (Page 4, Figure 1: The decoder sequentially outputs predictions for the identity of the amino acid at each point in the sequence). Additionally, the architecture of Hawkins-Hooker et al. is interpreted to be causal, dilated, and convolutional through the implementation of the autoregressive generator (Page 4, Figure 1: a convolutional neural network (CNN) encoder and decoder which combined upsampling with autoregression).
Hawkins-Hooker et al. does not explicitly teach the dilated architecture (Claim 1).
Regarding Claim 1, Bikard et al. teach (Claim 1.i) encoding, using a encoder, a plurality of input protein sequences onto a latent space distribution (Paragraph 0015: encoder, the encoder making it possible to encode protein sequences in a latent space as latent codes). Bikard et al. teach (Claim 1.ii) decoding, using a decoder, the latent space distribution to generate new protein sequences different from the input protein sequences (Paragraph 0010: generating a protein sequence in an autoregressive neural network comprising an encoder and a decoder). Bikard et al. teach causal dilated convolutional neural network which include encoding and decoding to generate protein sequences (Paragraph 0238: The autoregressive processing is achieved by a series of dilated causal convolutions organized in the same manner as in a Wavenet).
Bikard et al. do not explicitly teach a dilated convolutional encoder (Claim 1).
Regarding Claim 1, Shin et al. teach causal dilated convolutional neural network which include encoding and decoding to generate protein sequences (Page 8, Column 1, Paragraph 2: we use a residual causal dilated convolutional neural network architecture).
It would have been obvious to one of ordinary skill in the art at the time of the effective filing date to combine the methods of Hawkins-Hooker et al., Bikard et al., and Shin et al. Hawkins-Hooker et al. teach modeling that was robust and experimentally validated to generate new proteins from sequence data (Page 12, Paragraph 2: Experimental validation confirmed that a significant fraction of variants of a target luxA protein generated by both models were functional, while confirming the strengths of the MSA model, which generated a set of variants which almost without exception retained high levels of luminescence), which is a major focus of Bikard et al., Shin et al., and the instant application. Bikard et al. teach their model architecture is accurate and efficient at generating new proteins from sequence data (Paragraph 0017: It is able to modify physical properties of existing proteins and to generate new ones having chosen properties with relative accuracy; Paragraph 0060: It has proven extremely efficient at learning high level features of complex datasets in an unsupervised fashion, and at generating novel samples with similar properties to the data they are trained on; Paragraph 0133: It is almost 100% accurate at predicting padding characters), which is a major focus of Hawkins-Hooker et al., Shin et al., and the instant application. Shin et al. teach their modeling architecture is state of the art and versatile with demonstrated applicability to designing proteins from unaligned sequence data (Page 2, Column 2, Paragraph 3: In addition to this state-of-the-art performance, our new alignment-free method is inherently more general. It can deal with a much larger class of sequences and take into account variable-length effects; Page 6, Column 2, Paragraph 1: we validated our model first on deep mutational scan data, with onpar performance with the best currently available model and demonstrated application to examples for which robust alignments cannot be constructed, such as sequences with multiple insertions, deletions, and substitutions, and cases for which protein structures and experimental data are not available) which is a major focus of Hawkins-Hooker et al., Bikard et al., and the instant application. Furthermore, one of ordinary skill in the art would predict that the methods could be readily combined with a reasonable expectation of success because both are within the same technical field – utilizing machine leaning models that take in sequence data to generate novel proteins.
Claims 2, 7, 11, 16, 18, and 24 are rejected under 35 U.S.C. 103 as being unpatentable over Hawkins-Hooker et al., as applied above to claim 1, in view of Bikard et al., as applied above to claim 1, and in further view of Chen et al. (2021, 2020 IEEE International Conference on Bioinformatics and Biomedicine: 251-256). Italicized text from reference art.
Applicable claims include:
Claim 2. A method of generating protein sequences using a system including (Claim 2.i) an encoder coupled to an autoregressive generator, and (Claim 2.ii) having been trained with a loss function that comprises reconstruction loss and a mutual information maximization term, the method comprising: (Claim 2.iii) encoding, using the encoder, a plurality of input protein sequences onto a latent space distribution; and (Claim 2.iv) decoding, using the autoregressive generator, the latent space distribution to generate new protein sequences different from the input protein sequences.
Claim 7. The method of claim 2, wherein the encoding step comprises encoding the plurality of input protein sequences into a latent space embedding.
Claim 11. A method of training a system for generating protein sequences, the system including (Claim 11.i) an encoder that encodes a plurality of input protein sequences onto a latent space distribution, and (Claim 11.i) an autoregressive generator that decodes the latent space distribution to generate new protein sequences different from the input protein sequences, the method comprising: (Claim 11.i) training the system with a loss function that comprises reconstruction loss and a mutual information maximization term.
Claim 16. The method of claim 11, wherein the encoder learns a latent space embedding.
Claim 18. A system for generating protein sequences, comprising: (Claim 18.i) an encoder configured to encode a plurality of input protein sequences onto a latent space distribution; and (Claim 18.ii) an autoregressive generator configured to decode the latent space distribution to generate new protein sequences different from the input protein sequences,(Claim 18.iii) wherein the system is trained with a loss function that includes reconstruction loss and a mutual information maximization term.
Claim 24. A non-transitory computer-readable medium storing a program that, when executed by processing circuitry, causes the processing circuitry to perform a method of generating protein sequences using a system including an encoder coupled to an autoregressive generator, and having been trained with a loss function that includes reconstruction loss and a mutual information maximization term, the method comprising: encoding, using the encoder of a variational autoencoder, a plurality of input protein sequences onto a latent space distribution; and decoding, using the autoregressive generator, the latent space distribution to generate new protein sequences different from the input protein sequences.
Regarding Claim 2, Hawkins-Hooker et al. teach (Claim 2.i) an encoder coupled to an autoregressive generator (Page 4, Figure 1: Encoder and Autoregressive model). Hawkins-Hooker et al. teach (Claim 2.iii) encoding, using the encoder, a plurality of input protein sequences onto a latent space distribution (Page 5, Paragraph 3: To handle raw sequences we therefore designed a model incorporating a convolutional encoder. the latent space was used). Hawkins-Hooker et al. teach (Claim 2.iv) decoding, using the autoregressive generator, the latent space distribution to generate new protein sequences different from the input protein sequences (Page 4, Paragraph 1: ensures that novel sequences can be generated; Page 13, Paragraph 4: we showed that a VAE with an autoregressive decoder could be used to generate realistic sequences).
Regarding Claim 7, Hawkins-Hooker et al. the encoding step comprises encoding the plurality of input protein sequences into a latent space embedding (Page 5, Paragraph 3: the latent space was used; Page 5, Paragraph 4: represent learned ‘embeddings’ of amino acid identity).
Regarding Claim 11, Hawkins-Hooker et al. teach (Claim 11.i) the system including an encoder that encodes a plurality of input protein sequences onto a latent space distribution (Page 5, Paragraph 3: To handle raw sequences we therefore designed a model incorporating a convolutional encoder. the latent space was used). Hawkins-Hooker et al. teach (Claim 11.ii) an autoregressive generator that decodes the latent space distribution to generate new protein sequences different from the input protein sequences (Page 13, Paragraph 4: we showed that a VAE with an autoregressive decoder could be used to generate realistic sequences).
Regarding Claim 16, Hawkins-Hooker et al. teach the encoder learns a latent space embedding (Page 5, Paragraph 3: the latent space was used; Page 5, Paragraph 4: represent learned ‘embeddings’ of amino acid identity).
Regarding Claim 18 and 24, Hawkins-Hooker et al. teach (Claim 18.i) an encoder configured to encode a plurality of input protein sequences onto a latent space distribution (Page 5, Paragraph 3: To handle raw sequences we therefore designed a model incorporating a convolutional encoder. the latent space was used). Hawkins-Hooker et al. teach (Claim 18.ii) an autoregressive generator configured to decode the latent space distribution to generate new protein sequences different from the input protein sequences (Page 4, Paragraph 1: ensures that novel sequences can be generated; Page 13, Paragraph 4: we showed that a VAE with an autoregressive decoder could be used to generate realistic sequences). Additionally, Hawkins-Hooker et al. teach the method is performed by a generic computer processing system which inherently contains program code, memory, including a non-transitory computer readable medium, and at least one processor to execute the method (Page 16, Paragraph 3: We used the EVCouplings python package). Claim 24 recites the limitations of claim 18 directed to a non-transitory computer readable medium.
Hawkins-Hooker et al. does not explicitly teach a loss function (Claim 2.ii, 11.iii, 18.iii).
Regarding Claim 2, Bikard et al. teach (Claim 2.i) an encoder coupled to an autoregressive generator (Paragraph 0037: FIG. 2a and FIG. 2b illustrate an example of the logical architecture of a conditional variational auto-encoder having an autoregressive decoder, used during the training phase and for generating new protein sequences, respectively). Bikard et al. teach (Claim 2.iii) encoding, using the encoder, a plurality of input protein sequences onto a latent space distribution (Paragraph 0037: FIG. 2a and FIG. 2b illustrate an example of the logical architecture of a conditional variational auto-encoder having an autoregressive decoder, used during the training phase and for generating new protein sequences, respectively). Bikard et al. teach (Claim 2.iv) decoding, using the autoregressive generator, the latent space distribution to generate new protein sequences different from the input protein sequences (Paragraph 0037: FIG. 2a and FIG. 2b illustrate an example of the logical architecture of a conditional variational auto-encoder having an autoregressive decoder, used during the training phase and for generating new protein sequences, respectively).
Regarding Claim 11, Bikard et al. teach (Claim 11.i) the system including an encoder that encodes a plurality of input protein sequences onto a latent space distribution (Paragraph 0037: FIG. 2a and FIG. 2b illustrate an example of the logical architecture of a conditional variational auto-encoder having an autoregressive decoder, used during the training phase and for generating new protein sequences, respectively). Bikard et al. teach (Claim 11.iii) training the system with a loss function that comprises reconstruction loss and a mutual information maximization term (Paragraph 0069: During learning, the parameters updated to minimize a loss function).
Regarding Claim 18 and 24, Bikard et al. teach (Claim 18.i) an encoder configured to encode a plurality of input protein sequences onto a latent space distribution (Paragraph 0015: encoder, the encoder making it possible to encode protein sequences in a latent space as latent codes). Bikard et al. teach (Claim 18.ii) an autoregressive generator configured to decode the latent space distribution to generate new protein sequences different from the input protein sequences (Paragraph 0037: FIG. 2a and FIG. 2b illustrate an example of the logical architecture of a conditional variational auto-encoder having an autoregressive decoder, used during the training phase and for generating new protein sequences, respectively). Bikard et al. teach (Claim 18.iii) trained with a loss function with multiple terms (Paragraph 0118: the loss function preferably comprises a cross-entropy term and a Kullback-Leibler divergence term). Additionally, Bikard et al. teach the method is performed by a generic computer processing system which inherently contains program code, memory, including a non-transitory computer readable medium, and at least on processor to execute the method (Paragraph 0267: The processing device may be a device such as a micro-computer, a workstation, or a highly parallel computer). Claim 24 recites the limitations of claim 18 directed to a non-transitory computer readable medium.
Bikard et al. teach does not explicitly teach loss function that comprises reconstruction loss and a mutual information maximization term (Claim 2.ii, 11.iii, 18.iii).
Regarding Claim 2, Chen et al. teach (Claim 2.ii) training with a loss function that comprises reconstruction loss and a mutual information maximization term (Page 253, Column 1, Paragraph 4: At the core of our framework is the maximization of the mutual information; Page 254, Column 2, Paragraph 2: parameters set for global mutual information loss and reconstruction loss).
Regarding Claim 7, Chen et al. teach encoding step comprises encoding the plurality of input into a latent space embedding (Page 252, Column 1, Paragraph 4: For instance, node v’s free latent embedding is obtained).
Regarding Claim 11, Chen et al. teach (Claim 11.iii) training the system with a loss function that comprises reconstruction loss and a mutual information maximization term (Page 253, Column 1, Paragraph 4: At the core of our framework is the maximization of the mutual information; Page 254, Column 2, Paragraph 2: parameters set for global mutual information loss and reconstruction loss).
Regarding Claim 16, Chen et al. teach the encoder learns a latent space embedding (Page 252, Column 1, Paragraph 4: For instance, node v’s free latent embedding is obtained).
Regarding Claim 18 and 24, Chen et al. teach (Claim 18.iii) trained with a loss function that includes reconstruction loss and a mutual information maximization term (Page 253, Column 1, Paragraph 4: At the core of our framework is the maximization of the mutual information; Page 254, Column 2, Paragraph 2: parameters set for global mutual information loss and reconstruction loss). Claim 24 recites the limitations of claim 18 directed to a non-transitory computer readable medium.
It would have been obvious to one of ordinary skill in the art at the time of the effective filing date to combine the methods of Hawkins-Hooker et al., Bikard et al., and Chen et al. Hawkins-Hooker et al. teach modeling that was robust and experimentally validated to generate new proteins from sequence data (Page 12, Paragraph 2: Experimental validation confirmed that a significant fraction of variants of a target luxA protein generated by both models were functional, while confirming the strengths of the MSA model, which generated a set of variants which almost without exception retained high levels of luminescence), which is a major focus of Bikard et al. and the instant application. Bikard et al. teach their model architecture is accurate and efficient at generating new proteins from sequence data (Paragraph 0017: It is able to modify physical properties of existing proteins and to generate new ones having chosen properties with relative accuracy; Paragraph 0060: It has proven extremely efficient at learning high level features of complex datasets in an unsupervised fashion, and at generating novel samples with similar properties to the data they are trained on; Paragraph 0133: It is almost 100% accurate at predicting padding characters), which is a major focus of Hawkins-Hooker et al. and the instant application. Chen et al. teach their methods, including generating loss functions, were effective over other modeling strategies when applied to modeling proteins within neural networks (Page 255, Column 2, Paragraph 1: Experimental results show that MMIDTI outperforms baseline methods on a real-world DTI (drug-protein interaction) prediction task, which verifies the effectiveness and necessity of our multi-level mutual information-aware framework), which is a major focus of Hawkins-Hooker et al., Bikard et al., and the instant application. Furthermore, one of ordinary skill in the art would predict that the methods could be readily combined with a reasonable expectation of success because Chen et al. apply their methods to an encoder decoder architecture utilizing loss functions, which are used by Hawkins-Hooker et al., Bikard et al., and the instant application.
Claims 1-11 and 15-24 are rejected under 35 U.S.C. 103 as being unpatentable over Hawkins-Hooker et al., as applied above to claim 1, 2, 7, 11, 16, 18, and 24, in view of Bikard et al., as applied above to claim 1, 2, 7, 11, 16, 18, and 24, and in further view of Chen et al., as applied above to claims 2, 7, 11, 16, 18, and 24, and Shin et al., as applied above to claim 1. Italicized text from reference art.
Applicable claims include:
Claims 1, 2, 7, 11, 16, 18, and 24 are presented above.
Claim 3. The method of claim 2, wherein the decoding step comprises decoding the latent space distribution using a dilated casual convolution autoregressive generator as the autoregressive generator.
Claim 4. The method of claim 2, wherein the decoding step comprises decoding the latent space distribution to generate the new protein sequences, which include sequences of different lengths.
Claim 5. The method of claim 2, wherein the encoding step comprises encoding the plurality of input protein sequences, which are unaligned.
Claim 6. The method of claim 2, wherein the encoding step comprises encoding the plurality of input protein sequences using a dilated convolutional neural network encoder.
Claim 8. The method of claim 2, wherein the decoding step comprises predicting a next amino acid in a particular sequence, based on the particular sequence and a latent space embedding.
Claim 9. The method of claim 2, wherein the system was trained using the loss function, which further includes a semi-supervised loss.
Claim 10. The method of claim 3, wherein the decoding step further comprises decoding the latent space distribution using the dilated casual convolution autoregressive generator, which incorporates residual and skip connections.
Claim 15. The method of claim 11, wherein the encoder is a dilated convolutional neural network encoder.
Claim 17. The method of claim 11, wherein the training step comprises training the system with the loss function, which further includes a semi-supervised loss.
Claim 19. The system of claim 18, wherein the autoregressive generator is a dilated casual convolution autoregressive generator.
Claim 20. The system of claim 18, wherein the autoregressive generator is further configured to decode the latent space distribution to generate the new protein sequences, which include sequences of different lengths.
Claim 21. The system of claim 18, wherein the encoder is further configured to encode the plurality of input protein sequences, which are unaligned.
Claim 22. The system of claim 18, wherein the system was trained with the loss function, which further includes a semi-supervised loss.
Claim 23. The system of claim 19, wherein the dilated casual convolution autoregressive generator incorporates residual and skip connections.
Regarding Claim 1, the limitations are taught by Hawkins-Hooker et al., Bikard et al., and Shin et al., as above.
Regarding Claims 2, 7, 11, 16, 18, and 24, the limitations are taught by Hawkins-Hooker et al., Bikard et al., and Chen et al., as above.
Regarding Claim 3 and 19, Hawkins-Hooker et al. teach the decoding step comprises decoding the latent space distribution using a autoregressive generator as the autoregressive generator (Page 13, Paragraph 4: we showed that a VAE with an autoregressive decoder could be used to generate realistic sequences). Additionally, the architecture of Hawkins-Hooker et al. is interpreted to be causal, dilated, and convolutional through the implementation of the autoregressive generator (Page 4, Figure 1: a convolutional neural network (CNN) encoder and decoder which combined upsampling with autoregression). Claim 19 recites the limitation of claim 3 directed to a system.
Regarding Claim 4 and 20, Hawkins-Hooker et al. teach the decoding step comprises decoding the latent space distribution to generate the new protein sequences, which include sequences of different lengths (Page 4, Paragraph 1: ensures that novel sequences can be generated; Page 13, Paragraph 4: we showed that a VAE with an autoregressive decoder could be used to generate realistic sequences; Page 5, Paragraph 3: the latent space was used). The generated sequences are different length (Page 8, Paragraph 1: 12 sequences from each model were selected for synthesis (S1 File), spanning a range of distances (17-48 total differences including substitutions and deletions)). Claim 20 recites the limitations of claim 4 directed to a system.
Regarding Claim 5 and 21, Hawkins-Hooker et al. teach the encoding step comprises encoding the plurality of input protein sequences, which are unaligned (Page 4, Figure 1: Unaligned sequences are input into an encoder). Claim 21 recites the limitations of claims 5 directed to a system.
Regarding Claim 6, Hawkins-Hooker et al. teach the encoding step comprises encoding the plurality of input protein sequences using a neural network encoder (Page 5, Paragraph 3: To handle raw sequences we therefore designed a model incorporating a convolutional encoder. the latent space was used). Additionally, the architecture of Hawkins-Hooker et al. is interpreted to be causal, dilated, and convolutional through the implementation of the autoregressive generator (Page 4, Figure 1: a convolutional neural network (CNN) encoder and decoder which combined upsampling with autoregression).
Regarding Claim 8, Hawkins-Hooker et al. teach the decoding step comprises predicting a next amino acid in a particular sequence, based on the particular sequence and a latent space embedding (Page 4, Figure 1: The decoder sequentially outputs predictions for the identity of the amino acid at each point in the sequence, conditioned on the upsampled latent representation together with the previous amino acids).
Hawkins-Hooker et al. does not explicitly teach the dilated architecture (Claim 3 and 15). Hawkins-Hooker et al. does not explicitly teach utilizing a semi supervised loss function (Claims 9, 17, and 22). Hawkins-Hooker et al. does not explicitly teach skip connections (Claims 10 and 23).
Regarding Claim 3 and 19, Bikard et al. teach the decoding step comprises decoding the latent space distribution using a dilated casual convolution autoregressive generator as the autoregressive generator (Paragraph 0238: The autoregressive processing is achieved by a series of dilated causal convolutions organized in the same manner as in a Wavenet). Claim 19 recites the limitation of claim 3 directed to a system.
Regarding Claim 6, Bikard et al. teach the encoding step comprises encoding the plurality of input protein sequences using a dilated convolutional neural network encoder (Paragraph 0238: The autoregressive processing is achieved by a series of dilated causal convolutions organized in the same manner as in a Wavenet).
Regarding Claim 8, Bikard et al. teach the decoding step comprises predicting a next amino acid in a particular sequence, based on the particular sequence and a latent space embedding (Paragraph 0061: To generate new samples, latent vectors are used to stimulate the autoregressive sequential generation of a new sample, one amino acid at a time).
Regarding Claim 9, 17, and 22, Bikard et al. teach the system was trained using the loss function, which further includes a semi-supervised loss (Paragraph 0128: An error term reflecting how well these predictions are made can then be added to the loss function. Training such a model could be described as semi-supervised). Claim 17 is directed the limitations of claim 9 directed to another method. Claim 22 is directed the limitations of claim 9 directed to a system.
Regarding Claim 10 and 23, Bikard et al. teach the decoding step further comprises decoding the latent space distribution using the dilated casual convolution autoregressive generator (Paragraph 0238: The autoregressive processing is achieved by a series of dilated causal convolutions organized in the same manner as in a Wavenet). Bikard et al. teach incorporating residual and skip connections (Paragraph 0087: used in skip connections; Paragraph 0115: these residual connections reduce the vanishing gradient problem, making training easier). Claim 23 recites the limitations of claims 10 directed to a system.
Regarding Claim 15, Bikard et al. teach the encoder is a dilated convolutional neural network encoder (Paragraph 0238: The autoregressive processing is achieved by a series of dilated causal convolutions organized in the same manner as in a Wavenet).
Bikard et al. does not explicitly teach the dilated architecture recited by Claim 3.
Regarding Claim 3 and 19, Shin et al. teach causal dilated convolutional neural network which includes encoding and decoding to generate novel sequences (Page 8, Column 1, Paragraph 2: we use a residual causal dilated convolutional neural network architecture). Claim 19 recites the limitation of claim 3 directed to a system.
Regarding Claim 4 and 20, Shin et al. suggest decoding step comprises decoding the latent space distribution to generate the new protein sequences, which include sequences of different lengths (Page 5, Column 1, Paragraph 1: Because the autoregressive model is not dependent on alignments, we can now learn mappings of sequences of high variability and diverse lengths). Claim 20 recites the limitations of claim 4 directed to a system.
Regarding Claim 5 and 21, Shin et al. suggest the encoding step comprises encoding the plurality of input protein sequences, which are unaligned (Page 3, Column 1, Paragraph 2: The autoregressive nature of this model obviates the need for a structural alignment). Claim 21 recites the limitations of claims 5 directed to a system.
Regarding Claim 6, Shin et al. teach causal dilated convolutional neural network which includes encoding and decoding to generate novel sequences (Page 8, Column 1, Paragraph 2: we use a residual causal dilated convolutional neural network architecture).
Regarding Claim 8, Shin et al. teach the decoding step comprises predicting a next amino acid in a particular sequence, based on the particular sequence and a latent space embedding (Page 3, Column 1, Paragraph 1: predicting the amino acid in a sequence using all of the amino acids that come before).
Regarding Claim 9, 17, and 22, Shin et al. suggest the system was trained using the loss function, which further includes a semi-supervised loss (Page 6, Column 2, Paragraph 1: we do not discount the utility of semi-supervised methods; Page 8, column 1, Paragraph 2: the loss had visibly converged). Claim 17 is directed the limitations of claim 9 directed to another method. Claim 22 is directed the limitations of claim 9 directed to a system.
Regarding Claim 15, Shin et al. teach the encoder is a dilated convolutional neural network encoder (Page 8, Column 1, Paragraph 2: we use a residual causal dilated convolutional neural network architecture).
Regarding Claim 9, 17, and 22, Chen et al. suggest the system was trained using the loss function, which further includes a semi-supervised loss (Page 1, Column 2, Paragraph 2: most successful GNNs are used in semi-supervised learning; Page 3, Column 1, Paragraph 1: loss function to be optimized). Claim 17 is directed the limitations of claim 9 directed to another method. Claim 22 is directed the limitations of claim 9 directed to a system.
It would have been obvious to one of ordinary skill in the art at the time of the effective filing date to combine the methods of Hawkins-Hooker et al., Bikard et al., Shin et al., and Chen et al. Hawkins-Hooker et al. teach modeling that was robust and expediently validated to generate new proteins from sequence data (Page 12, Paragraph 2: Experimental validation confirmed that a significant fraction of variants of a target luxA protein generated by both models were functional, while confirming the strengths of the MSA model, which generated a set of variants which almost without exception retained high levels of luminescence), which is a major focus of Bikard et al., Shin et al., and the instant application. Bikard et al. teach their model architecture is accurate and efficient at generating new proteins from sequence data (Paragraph 0017: It is able to modify physical properties of existing proteins and to generate new ones having chosen properties with relative accuracy; Paragraph 0060: It has proven extremely efficient at learning high level features of complex datasets in an unsupervised fashion, and at generating novel samples with similar properties to the data they are trained on; Paragraph 0133: It is almost 100% accurate at predicting padding characters), which is a major focus of Hawkins-Hooker et al., Shin et al., and the instant application. Shin et al. teach their modeling architecture is state of the art and versatile with demonstrated applicability to designing proteins from unaligned sequence data (Page 2, Column 2, Paragraph 3: In addition to this state-of-the-art performance, our new alignment-free method is inherently more general. It can deal with a much larger class of sequences and take into account variable-length effects; Page 6, Column 2, Paragraph 1: we validated our model first on deep mutational scan data, with onpar performance with the best currently available model and demonstrated application to examples for which robust alignments cannot be constructed, such as sequences with multiple insertions, deletions, and substitutions, and cases for which protein structures and experimental data are not available) which is a major focus of Hawkins-Hooker et al., Bikard et al., and the instant application. Chen et al. teach their methods, including generating loss functions, were effective over other modeling strategies when applied to modeling proteins within neural networks (Page 255, Column 2, Paragraph 1: Experimental results show that MMIDTI outperforms baseline methods on a real-world DTI (drug-protein interaction) prediction task, which verifies the effectiveness and necessity of our multi-level mutual information-aware framework), which is a major focus of Hawkins-Hooker et al., Bikard et al., and the instant application. Furthermore, one of ordinary skill in the art would predict that the methods could be readily combined with a reasonable expectation of success because all are within the same technical field – utilizing machine leaning models to make predictions on proteins.
Claims 1-24 are rejected under 35 U.S.C. 103 as being unpatentable over Hawkins-Hooker et al., as applied above to claim 1-11 and 15-24, in view of Bikard et al., as applied above to claims 1-11 and 15-24, and in further view of Chen et al., as applied above to claims 1-11 and 15-24, and Shin et al., as applied above to claims 1-11 and 15-24, and Alley et al. (2019, Nature Methods, Vol. 16: 1-12). Italicized text from reference art.
Applicable claims include:
Claims 1-11 and 15-24 are presented above.
Claim 12. The method of claim 11, wherein the system further includes a semi-supervised learning module including a regression model with a set of training parameters that are learned by minimizing, for a subset of the latent space distribution, an error between outputs of the regression model and fitness values obtained from assay measurements; and the method further comprises training the system with a modified loss function that further includes term based on performance of the regression model.
Claim 13. The method of claim 12, wherein the term in the modified loss function is a mean-squared error term based on a ground truth and a predicted regression value of the regression model.
Claim 14. The method of claim 12, wherein the regression model is a neural network having weights as the training parameters, which are determined in the training step.
Regarding Claims 1-11 and 15-24, these limitations are taught by Hawkins-Hooker et al., Bikard et al., Chen et al., and Shin et al. as indicated above.
Regarding Claim 12, Hawkins-Hooker et al. suggest including a regression model with a set of training parameters that are learned by minimizing, for a subset of the latent space distribution, an error between outputs of the regression model and fitness values obtained from assay measurements; and the method further comprises training the system with a modified loss function that further includes term based on performance of the regression model (Page 11, Paragraph 3: by learning the distribution of sequences in a particular family, the model captures patterns of sequence variation that underlie function. In practice, this is achieved by training the model in such a way as to maximize the likelihood of the sequences in the family, since these were arrived at by a process of natural selection) The models are trained to consider fitness. Additionally, Hawkins-Hooker et al. teach the methods are conducted by generic computers (semi-supervised learning module (see 112(f) claim interpretation)).
Hawkins-Hooker et al. do not explicitly teach the fitness modeling as recited by Claims 12-14.
Regarding Claim 12, Bikard et al. suggest including a regression model with a set of training parameters that are learned by minimizing, for a subset of the latent space distribution, an error between outputs of the regression model and fitness values obtained from assay measurements; and the method further comprises training the system with a modified loss function that further includes term based on performance of the regression model (Paragraph 0211: Training a simple multi-class logistic regression model on top of the latent representation of the 50,000 annotated samples makes it possible to classify proteins). Additionally, Bikard et al. teach the methods are conducted by generic computers (semi-supervised learning module (see 112(f) claim interpretation)).
Regarding Claim 13, Bikard et al. suggest the term in the modified loss function is an error term based on a ground truth and a predicted regression value of the regression model (Paragraph 0248: A model's reconstruction ability can be measured by the accuracy of its predictions of amino acid identities).
Bikard et al. do not explicitly teach the fitness modeling as recited by Claims 12-14.
Regarding Claim 12, Shin et al. suggest including a regression model with a set of training parameters that are learned by minimizing, for a subset of the latent space distribution, an error between outputs of the regression model and fitness values obtained from assay measurements; and the method further comprises training the system with a modified loss function that further includes term based on performance of the regression model (Page 3, Column 1, Paragraph 2: the model can be directly applied to generating fit proteins; Page 5, Column 1, Paragraph 2: We compare the fitness predictions calculated as log probabilities by the autoregressive model to experimental assays for the fitness of mutated biomolecules).
Regarding Claim 13, Shin et al. suggest the term in the modified loss function is an error term based on a ground truth and a predicted regression value of the regression model (Page 3, Column 1, Paragraph 2: the model can be directly applied to generating fit proteins; Page 5, Column 1, Paragraph 2: We compare the fitness predictions calculated as log probabilities by the autoregressive model to experimental assays for the fitness of mutated biomolecules).
Shin et al. do not explicitly teach the fitness modeling as recited by Claims 12-14.
Regarding Claim 12, Alley et al. teach including a regression model with a set of training parameters that are learned by minimizing, for a subset of the latent space distribution, an error between outputs of the regression model and fitness values obtained from assay measurements; and the method further comprises training the system with a modified loss function that further includes term based on performance of the regression model (Page 2, Figure 1: A top model (for example, a sparse linear regression) trained on top of the representation, which acts as a featurization of the input sequence, enables supervised learning on diverse protein informatics tasks; Page 6, Column 1, Paragraph 2: Using these trained unsupervised models we generated representations for the green fluorescent protein from avGFP variant sequences from Sarkisyan et al and trained simple sparse linear regression top models on each to predict avGFP brightness; Page 9, Column 1, Paragraph 4: Hyperparameters were tuned manually on a small number of weight updates and final parameters were selected based on the rate and stability of generalization loss decrease).
Regarding Claim 13, Alley et al. teach the term in the modified loss function is a mean-squared error term based on a ground truth and a predicted regression value of the regression model (Page 4, Figure 3: UniRep Fusion achieves statistically lower mean squared error).
Regarding Claim 14, Alley et al. teach the regression model is a neural network having weights as the training parameters, which are determined in the training step (Page 9, Column 1, Paragraph 4: Hyperparameters were tuned manually on a small number of weight updates and final parameters were selected based on the rate and stability of generalization loss decrease). Additionally, this is obvious given model training inherently includes adjusting weights (see other art citations related to training).
It would have been obvious to one of ordinary skill in the art at the time of the effective filing date to combine Alley et al. with Hawkins-Hooker et al., Bikard et al., Shin et al., and Chen et al. Alley et al. teach methods for modeling protein sequences, including the use of fitness data, that enhance the generation of novel proteins from sequence data (Page 1, Column 2, Paragraph 3: This method scalably leverages underused raw sequences to alleviate the data scarcity constraining protein informatics so far, and achieves generalizable, superior performance in critical engineering tasks from stability, to function, to design; Page 6, Column 2, Paragraph 2: since UniRep is learned from raw data, it is unconstrained by existing mental models for understanding proteins. By enabling rapid generalization to distant, unseen regions of the fitness landscape, UniRep may improve protein engineering workflows or, in the best case, enable the discovery of sequence variants inaccessible to purely experimental or structural approaches), which are the focus of Hawkins-Hooker et al., Bikard et al., Shin et al., and the instant application. Furthermore, one of ordinary skill in the art would predict that the methods could be readily combined with a reasonable expectation of success because all are within the same technical field - utilizing machine leaning models to make predictions on proteins.
Double Patenting
No double patenting was identified.
Conclusion
No claims are allowed.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BLAKE H ELKINS whose telephone number is (571)272-2649. The examiner can normally be reached Monday-Thursday 8-5PM.
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.
/B.H.E./Examiner, Art Unit 1687
/Karlheinz R. Skowronek/Supervisory Patent Examiner, Art Unit 1687
Prosecution Insights