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 1-15, 17-19, and 27-28 are pending.
Priority
This application is a 371 of PCT/CN2021/121878, filed 09/29/2021, The instant application has the effective filing date of 29 September 2021.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 03/08/2023 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement has been considered by the examiner.
Drawings
The drawings, submitted on 03/08/2023, are accepted by the examiner.
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 1-15, 17-19, and 27-28 are rejected under U.S.C 101 because the claimed invention is directed to abstract ideas without significantly more, as detailed in the analysis below.
Eligibility Step 1: Subject matter eligibility evaluation in accordance with MPEP § 2106:
Claims 1-15 and 17-19 are directed to a statutory category (method).
Claim 27 is directed to a statutory category (product).
Claim 28 is directed to a statutory category (apparatus).
Therefore, in accordance with MPEP § 2106.03, all claims have patent eligible subject matter.
[Eligibility Step 1: YES]
Eligibility Step 2A: This step determines whether a claim is directed to a judicial exception in accordance with MPEP § 2106.
Eligibility Step 2A -- Prong One: Limitations are analyzed to determine if the claims recite any concepts that could equate to a judicial exception (i.e. abstract idea, law of nature, or natural phenomenon). Possible judicial exceptions are explored below.
Recitations of Judicial Exceptions:
Claims 1 and 27-28: encoding the candidate RNA sequence to obtain an RNA vector sequence; encoding the candidate protein sequence to obtain a protein vector sequence; constructing a matching feature matrix according to the RNA vector sequence and the protein vector sequence; and performing feature extraction on the matching feature matrix, and determining, according to an extracted matching feature, an interaction between the candidate RNA sequence and the candidate protein sequence. (mathematical concept, mental process)
Claim 2: wherein the encoding the candidate RNA sequence to obtain an RNA vector sequence, comprising: converting the candidate RNA sequence into N base k-mer subsequences; and vectorizing each base k-mer subsequence of the N base k-mer subsequences to obtain the RNA vector sequence. (mathematical concept)
Claim 3: wherein the vectorizing each base k-mer subsequence of the N base k-mer subsequences to obtain the RNA vector sequence, comprising: encoding each base k-mer subsequence of the N base k-mer subsequences to obtain first vectors of the N base k-mer subsequences, and constituting the RNA vector sequence by the first vectors of the N base k-mer subsequences. (mathematical concept, mental process)
Claim 4: wherein the vectorizing each base k-mer subsequence of the N base k-mer subsequences to obtain the RNA vector sequence, comprising: encoding each base k-mer subsequence of the N base k-mer subsequences to obtain first vectors of the N base k-mer subsequences; and constituting the RNA vector sequence by the N base k-mer vectors. (mathematical concept, mental process)
Claim 5: wherein the vectorizing each base k-mer subsequence of the N base k-mer subsequences to obtain the RNA vector sequence, comprising: encoding each base k-mer subsequence of the N base k-mer subsequences to obtain first vectors of the N base k-mer subsequences; and performing operation on the first vectors of the N base k-mer subsequences by using a first mapping matrix to obtain second vectors of the N base k-mer subsequences, and constituting the RNA vector sequence by the second vectors of the N base k-mer subsequences. (mathematical concept)
Claim 6: wherein the vectorizing each base k-mer subsequence of the N base k-mer subsequences to obtain the RNA vector sequence, comprising: encoding each base k-mer subsequence of the N base k-mer subsequences to obtain first vectors of the N base k-mer subsequences; performing operation on the first vectors of the N base k-mer subsequences by using a first mapping matrix to obtain second vectors of the N base k-mer subsequences; and constituting the RNA vector sequence by the N base k-mer vectors. (mathematical concept, mental process)
Claim 7: wherein the encoding the candidate protein sequence to obtain a protein vector sequence, comprising: converting the candidate protein sequence into M amino acid k-mer subsequences; and vectorizing each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain the protein vector sequence. (mathematical concept)
Claim 8: wherein the vectorizing each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain the protein vector sequence, comprising: encoding each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain first vectors of the M amino acid k-mer subsequences, and constituting the protein vector sequence by the first vectors of the M amino acid k-mer subsequences. (mathematical concept, mental process)
Claim 9: wherein the vectorizing each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain the protein vector sequence, comprising: encoding each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain first vectors of the M amino acid k-mer subsequences; and constituting the protein vector sequence by the M amino acid k-mer vectors. (mathematical concept, mental process)
Claim 10: wherein the vectorizing each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain the protein vector sequence, comprising: encoding each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain first vectors of the M amino acid k-mer subsequences; and performing operation on the first vectors of the M amino acid k-mer subsequences by using a second mapping matrix to obtain second vectors of the M amino acid k-mer subsequences, and constituting the protein vector sequence by the second vectors of the M amino acid k-mer subsequences. (mathematical concept, mental process)
Claim 11: wherein the vectorizing each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain the protein vector sequence, comprising: encoding each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain first vectors of the M amino acid k-mer subsequences; performing operation on the first vectors of the M amino acid k-mer subsequences by using a second mapping matrix to obtain second vectors of the M amino acid k-mer subsequences; and constituting the protein vector sequence by the M amino acid k-mer vectors. (mathematical concept, mental process)
Claim 12: wherein the constructing a matching feature matrix according to the RNA vector sequence and the protein vector sequence, comprising: obtaining a calculated matching degree score by calculating a matching degree between a base k-mer vector in the RNA vector sequence and an amino acid k-mer vector in the protein vector sequence; and constructing the matching feature matrix by taking a-the calculated matching degree score as an element of the matching feature matrix. (mathematical concept, mental process)
Claim 13: wherein the calculating a matching degree between a base k-mer vector in the RNA vector sequence and an amino acid k-mer vector in the protein vector sequence, comprising: calculating the matching degree riI between the i-th base k-mer vector in the RNA vector sequence and the j-th amino acid k-mer vector o in the protein vector sequence according to [equation not recited herein, for brevity]. (mathematical concept, mental process)
Claim 15: wherein the performing feature extraction on the matching feature matrix, comprising: performing feature extraction on the matching feature matrix by using a feature extraction network to obtain an original feature; and performing operation on the original feature by using a third mapping matrix to obtain the extracted matching feature. (mathematical concept, mental process)
Claim 17: wherein the determining, according to an extracted matching feature, an interaction between the candidate RNA sequence and the candidate protein sequence, comprising: obtaining an interaction predicted value between the candidate RNA sequence and the candidate protein sequence according to the extracted matching feature; and determining the interaction between the candidate RNA sequence and the candidate protein sequence according to the interaction predicted value, wherein the determining the interaction between the candidate RNA sequence and the candidate protein sequence according to the interaction predicted value, comprising: determining, in response to the interaction predicted value meeting a preset threshold condition, that the interaction exists between the candidate RNA sequence and the candidate protein sequence. (mathematical concept, mental process)
Claim 18: wherein the obtaining an interaction predicted value between the candidate RNA sequence and the candidate protein sequence according to the extracted matching feature, comprising: and outputting a probability of the presence of the interaction between the candidate RNA sequence and the candidate protein sequence. (mathematical concept)
Claim 19: wherein the probability of the presence of the interaction between the candidate RNA sequence and the candidate protein sequence is [equation not recited herein, for brevity], wherein r represents the candidate RNA sequence, p represents the candidate protein sequence, C represents a first feature value in the extracted matching feature, and ci represents a second feature value in the extracted matching feature. (mathematical concept)
Step 2A – Prong One Analysis:
Analysis techniques constituting sequences and making mental determinations based on data, requiring nothing more than the human mind and pen/paper, read on observations, evaluations, judgments, and opinions, and thus fall under the mental process grouping of abstract ideas.
Analysis techniques such as encoding, vectorizing, calculating probability, and using classifiers in the form of equations and calculations recite mathematical calculations, functions, and relationships that fall under the mathematical concept grouping of abstract ideas.
Therefore, the claims are found to recite judicial exceptions.
[Eligibility Step 2A – Prong One: YES]
Eligibility Step 2A – Prong Two: A claim that integrates a judicial exception into a practical application will apply, rely on, or use the judicial exception in a manner that imposes a meaningful limit on the judicial exception. If the claim contains no additional claim elements beyond the abstract idea, the claim fails to integrate the abstract idea into a practical application (MPEP 2106.04(d)). Additional elements are recited, categorized, and analyzed below.
Data Gathering/Outputting Elements:
Claims 1 and 27-28: acquiring a candidate RNA sequence and a candidate protein sequence
Claim 4: inputting the first vectors of the N base k-mer subsequences sequentially into a pre-trained recurrent neural network, and outputting N base k-mer vectors
Claim 6: inputting the second vectors of the N base k-mer subsequences sequentially into a pre-trained recurrent neural network, outputting N base k-mer vectors
Claim 9: inputting the first vectors of the M amino acid k-mer subsequences sequentially into a pre-trained recurrent neural network, outputting M amino acid k-mer vectors,
Claim 11: inputting the second vectors of the M amino acid k-mer subsequences sequentially into a pre-trained recurrent neural network, outputting M amino acid k-mer vectors
Claim 18: inputting the extracted matching feature into a classifier
Computer Components Elements:
Claim 27: non-transitory computer-readable storage medium, having a computer program stored thereon, wherein the computer program, when performed by a processor
Claim 28: electronic device, comprising: a processor; and a memory for storing executable instructions for the processor
Step 2A – Prong Two Analysis:
Though claims 4, 6, 9, and 11 recite the use of a pre-trained recurrent neural network, the network itself merely acts as a tool to complete the judicial exceptions; and the claim limitations are directed to merely inputting and outputting data from the structure. As such, limitations of this nature are classified as insignificant extra solution activities that merely perform data gathering activities for the clamed invention per MPEP 2106.05(g).
Generic computer components and implementations provide mere instructions to implement the abstract ideas onto a technological environment per Alice Corp., 573 U.S. at 223, 110 USPQ2d at 1983. See also 573 U.S. at 224, 110 USPQ2d at 1984.
As such, the additional elements, when viewed separately and in the context of a whole claimed invention, do not integrate the judicial exceptions into practical application.
[Eligibility Step 2A – Prong Two: NO]
Eligibility Step 2B: Claim elements are probed for inventive concept equating to significantly more than the judicial exception (MPEP 2106.04(II)).
Step 2B Analysis:
The data gathering components of the generic computer environments are found well-understood, routine, and conventional per Mayo, 566 U.S. at 79, 101 USPQ2d at 1968; OIP Techs., Inc. v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1092-93 (Fed. Cir. 2015) of MPEP 2106.05(g).
The data gathering components which utilize neural networks are also found well-understood, routine, and conventional per Yan et al. (IEEE Access; vol. 8; 2020) which reviews RNA-Protein Binding Sites Predictions Based on Deep Learning that discloses inputting encoded sequences into recurrent neural networks.
The computer components are further found to be well-understood, routine, and conventional per Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); OIP Techs., 788 F.3d at 1363, 115 USPQ2d at 1092-93 for storing and retrieving information in memory and Alice Corp., 573 U.S. at 225, 110 USPQ2d at 1984 (MPEP 2106.05 (a)).
As such, the additional elements are further found to lack inventive concept.
[Eligibility Step 2B: NO]
Therefore, claims 1-15, 17-19, and 27-28 are directed to judicial exceptions without significantly more and are rejected under 35 U.S.C 101.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1-3, 5, 7-8, and 27-28 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Yuan et al. (Frontiers in Genetics; Vol. 11: 632861; 2021).
Yuan et al. describes DeCban, a method of predicting circular RNA-RNA binding protein interaction sites using double embeddings and cross-branch attention networks.
Claims 1 and 27-28 are directed to methods, computer readable mediums, and electronic devices that acquire candidate RNA and protein sequences; and encode them to obtain RNA and protein vector sequences.
Yuan et al. teaches collecting circRNAs and their interacting proteins (page 3, column 1); and converting input sequences, such as RNA and peptides (page 4, fig. 1) into numerical vectors with encoding schemes (page 3, column 2).
Yuan et al. further teaches DeCban outperforms the mainstream deep learning-based methods on computational efficiency the data sets (page 1, column 1); and the source code of this study is freely available (page 1, column 1).
Therefore, Yuan et al. teaches a method available as code on a computer readable medium and electronic device (computer).
Claims 1 and 27-28 are further directed to constructing a matching feature matrix according to the RNA and protein vector sequences; performing feature extraction on the matching feature matrix; and determining, according to an extracted matching feature, an interaction between the candidate RNA and protein sequence.
Yuan et al. teaches constructing the input matrix for each 7-mer fragment of a circRNA sequence by concatenating the RNA embedding and corresponding pseudo-peptide embedding (page 4, column 1); introducing self-attention modules to extract features of different abstract levels (page 4, column 2); and using the model for studying the interactions between RBP and circRNA (page 8, column 1).
Claim 2 is directed to encoding the candidate RNA sequence to obtain an RNA vector sequence, by converting the candidate RNA sequence into k-mer subsequences; and vectorizing each k-mer subsequence.
Yuan et al. teaches an RNA sequence is segmented into 7-mers, and each 7-mer is converted into an embedding vector (page 4, fig. 2).
Claim 3 is directed to vectorizing each k-mer subsequence to obtain the RNA vector sequence with the process of: encoding each base k-mer subsequence of the N base k-mer subsequences to obtain first vectors of the N base k-mer subsequences; and constituting the RNA vector sequence by the first vectors of the N base k-mer subsequences.
Yuan et al. teaches the first 7-mer “CACUAUA” contains the codons CAC, ACU, CUA, UAU, and AUA, which encode the amino acids H, T, L, Y, and I, respectively; then, the embedding vectors of “CACUAUA” and “HTLYI” are concatenated to represent the feature vector of “CACUAUA” (page 4, column 1).
Claim 5 is directed to vectorizing each k-mer subsequence to obtain the RNA vector sequence with the process of: encoding each base k-mer subsequence of the N base k-mer subsequences to obtain first vectors of the N base k-mer subsequences; performing operation on the first vectors of the N base k-mer subsequences by using a first mapping matrix to obtain second vectors of the N base k-mer subsequences; and constituting the RNA vector sequence by the second vectors of the N base k-mer subsequences.
Yuan et al. teaches performing pre-training of the word embeddings for k-mer RNA segments (page 3, column 2); treating the segmented k-mers as words; adopting the GloVe algorithm to train their embeddings; producing an embedding vector for the words by using a large corpus of text (page 3, column 2); and calling the feature extraction method as double embeddings (page 3, column 2).
Yuan et al. further teaches letting the RNA and peptide embedding vectors for wi be Ri and Pi, whose dimensions are p and q, respectively; defining the double embedding for wi as, Di=Ri㊉Pi,i∈{1,2,⋯,m}, where ㊉ denotes the concatenation operation; then representing the circRNA by a matrix of size (p + q) × m, [D1, D2, ⋯ , Dm] (page 4, column 2).
Claim 7 is directed to encoding the candidate protein sequence to obtain a protein vector sequence with the process of: converting the candidate protein sequence into M amino acid k-mer subsequences; and vectorizing each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain the protein vector sequence.
Yuan et al. teaches pre-training word embeddings for (k − 2)-mer peptides (page 3, column 2); and converting then into embedding vectors (page 4, fig. 2).
Claim 8 is directed to vectorizing each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain the protein vector sequence by the process of: encoding each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain first vectors of the M amino acid k-mer subsequences; and constituting the protein vector sequence by the first vectors of the M amino acid k-mer subsequences.
Yuan et al. teaches performing pre-training of the word embeddings for (k − 2)-mer peptides (page 3, column 2); and converting then into embedding vectors (page 4, fig. 2), such as “HTLYI” (page 4, column 1).
Claim 10 is directed to vectorizing each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain the protein vector sequence by the process of: encoding each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain first vectors of the M amino acid k-mer subsequences; and performing operation on the first vectors of the M amino acid k-mer subsequences by using a second mapping matrix to obtain second vectors of the M amino acid k-mer subsequences; and constituting the protein vector sequence by the second vectors of the M amino acid k-mer subsequences.
Yuan et al. teaches performing pre-training of the word embeddings for (k − 2)-mer peptides (page 3, column 2); converting then into embedding vectors (page 4, fig. 2), such as “HTLYI” (page 4, column 1); letting the RNA and peptide embedding vectors for wi be Ri and Pi, whose dimensions are p and q, respectively; defining the double embedding for wi as, Di=Ri㊉Pi,i∈{1,2,⋯,m}, where ㊉ denotes the concatenation operation; then representing the circRNA by a matrix of size (p + q) × m, [D1, D2, ⋯ , Dm] (page 4, column 2); processing the first layer output via a maximum pooling operation (page 4, column 2); and finally, obtaining the output O of the network through a FC layer (page 4, column 2).
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.
Claims 4, 6, 9, and 11 are rejected under 35 U.S.C. 103 as being unpatentable over Yuan et al. (Frontiers in Genetics; Vol. 11: 632861; 2021) in view of Iuchi et al. (Comp and Struct Biotech Journal; Vol. 19: 3198-3208; 2021).
Yuan et al. teaches a method of vectorizing and inputting k-mers of RNA and protein sequences into a model to predict RNA-protein interactions, as described above.
Claim 4 is directed to vectorizing each k-mer subsequence to obtain the RNA vector sequence with the process of: encoding each base k-mer subsequence of the N base k-mer subsequences to obtain first vectors of the N base k-mer subsequences; inputting the first vectors of the N base k-mer subsequences sequentially into a pre-trained recurrent neural network; outputting N base k-mer vectors; and constituting the RNA vector sequence by the N base k-mer vectors.
Yuan et al. teaches an RNA sequence is segmented into 7-mers, each 7-mer is converted into an embedding vector (page 4, fig. 2); mapping the 7-mer to a pseudo-peptide, which is also converted into an embedding vector (page 4, fig. 2); concatenating the two embedding vectors as a whole input (page 4, fig. 2) of a pre-trained, cross-branch attention neural network for classification (page 1, column 1).
Claim 6 is directed to vectorizing each k-mer subsequence to obtain the RNA vector sequence with the process of: encoding each base k-mer subsequence of the N base k-mer subsequences to obtain first vectors of the N base k-mer subsequences; performing operation on the first vectors of the N base k-mer subsequences by using a first mapping matrix to obtain second vectors of the N base k-mer subsequences; inputting the second vectors of the N base k-mer subsequences sequentially into a pre-trained recurrent neural network; outputting N base k-mer vectors; and constituting the RNA vector sequence by the N base k-mer vectors.
Yuan et al. teaches letting the RNA and peptide embedding vectors for wi be Ri and Pi, whose dimensions are p and q, respectively; defining the double embedding for wi as, Di=Ri㊉Pi,i∈{1,2,⋯,m}, where ㊉ denotes the concatenation operation; then representing the circRNA by a matrix of size (p + q) × m, [D1, D2, ⋯ , Dm] (page 4, column 2); and inputting the vectors into a pre-trained, cross-branch attention neural network for classification (page 1, column 1).
Claim 9 is directed to the vectorizing each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain the protein vector sequence by the process of: encoding each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain first vectors of the M amino acid k-mer subsequences; inputting the first vectors of the M amino acid l-mer subsequences sequentially into a pre-trained recurrent neural network; outputting M amino acid k-mer vectors; and constituting the protein vector sequence by the M amino acid k-mer vectors.
Yuan et al. teaches performing pre-training of the word embeddings for (k − 2)-mer peptides (page 3, column 2); converting then into embedding vectors (page 4, fig. 2), such as “HTLYI” (page 4, column 1); and concatenating the two embedding vectors as a whole input (page 4, fig. 2), representing the feature vector of “CACUAUA” (page 4, column 1) of a pre-trained cross-branch attention neural network for classification (page 1, column 1).
Claim 11 is directed to vectorizing each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain the protein vector sequence by the process of: encoding each amino acid k-mer subsequence of the M amino acid k-mer subsequences to obtain first vectors of the M amino acid k-mer subsequences; performing operation on the first vectors of the M amino acid k-mer subsequences by using a second mapping matrix to obtain second vectors of the M amino acid k-mer subsequences; inputting the second vectors of the M amino acid k-mer subsequences sequentially into a pre-trained recurrent neural network, outputting M amino acid k-mer vectors; and constituting the protein vector sequence by the M amino acid k-mer vectors.
Yuan et al. teaches letting the RNA and peptide embedding vectors for wi be Ri and Pi, whose dimensions are p and q, respectively; defining the double embedding for wi as, Di=Ri㊉Pi,i∈{1,2,⋯,m}, where ㊉ denotes the concatenation operation; then representing the circRNA by a matrix of size (p + q) × m, [D1, D2, ⋯ , Dm] (page 4, column 2).
Yuan et al. does not explicitly teach inputting the vectors into a recurrent neural network (claims 4, 6, 9, and 11).
Yuan et al. does teach DeCban is a convolutional neural network (page 6, column 1); but besides CNN models, we also consider the widely-used recurrent neural network (RNN), LSTM (page 6, column 2); in some cases, attention mechanism becomes a necessary part to enable the model focus on informative regions, thus the BiLSTM model with attention improves performance of LSTM significantly, even better than basic CNNs (page 7, column 1); and as can be seen, DeCban achieves the highest average F1 of 0.841, and the second best model is BiLSTM with attention, whose average F1 is 0.827 (page 5, column 2).
Therefore Yuan et al. teaches inputting the vectors into a recurrent neural network is a known technique that would yield success and predictable results for the purpose of constituting a classified sequence that gives insights into RNA-protein interactions. It further teaches that some RNN embodiments have benefits over convolutional neural networks. As such, it would be obvious to one of ordinary skill in the art to use a recurrent neural network for the claimed purpose as a substitution for the convolutional neural network, as it is among one of the known techniques that would yield predictable results for the constitution of the biological sequences in RNA-protein interaction prediction.
Yuan et al. also does not teach sequentially inputting the vectorized RNA and protein k-mer subsequences (claims 4, 6, 9, and 11).
Iuchi et al. describes representation learning applications in biological sequence analysis.
Iuchi et al. teaches in natural language processing methods, which tackle the issue of aptly analyzing substantial amounts of rapidly generated biological DNA/RNA/protein sequencing data, biological sequences are regarded as sentences while the single nucleic acids/amino acids or k-mers in these sequences represent the words; embeddings perform the conversion of these words into vectors; and the vectorized biological sequences can then be applied for function and structure estimation, or as input for other probabilistic models (page 1, column 1).
Iuchi et al. further teaches RNN and LSTM are developments of the classical autoregressive language models that have been primarily utilized for sequential tasks (page 3, column 2); and a comprehensive survey of representation learning applications in biological sequences includes BERT-RBP, which conducts RNA-RBP interaction prediction (page 6, table 1).
Therefore Iuchi et al. teaches a framework of inputting vectorized k-mers of biological sequences into probabilistic deep learning models, such as those capable of predicting RNA-protein interactions. Iuchi et al. further provides motivation for one of the ordinary skill in the art to sequentially input sequence representations into recurrent neural networks such as LSTMs, as it is a known function of the model.
Therefore Yuan et al. teaches RNNs, specifically LSTMs, can be successfully used for RNA-protein interaction prediction; and Iuchi et al. teaches LSTMs are also well suited for predictions of sequential input, such as RNA and protein sequences. As such, it would be obvious to one of ordinary skill in the art to input k-mer RNA and protein vectors sequentially into RNNs with a reasonable expectation of success in RNA-protein interaction prediction.
Claim 17 is rejected under 35 U.S.C. 103 as being unpatentable over Yuan et al. (Frontiers in Genetics; Vol. 11: 632861; 2021) in view of Iuchi et al. (Comp and Struct Biotech Journal; Vol. 19: 3198-3208; 2021) in view of Yan et al. (IEEE Access; vol. 8; 2020).
Yuan et al. in view of Iuchi et al. teach a method of vectorizing and inputting k-mers of RNA and protein sequences into recurrent neural networks to predict RNA-protein interactions, as described above.
Claim 17 is directed to determining an interaction between the candidate RNA sequence and the candidate protein sequences, according to an extracted matching feature, by: obtaining an interaction predicted value between the candidate RNA sequence and the candidate protein sequence according to the extracted matching feature; determining the interaction between the candidate RNA sequence and the candidate protein sequence according to the interaction predicted value; and determining, in response to the interaction predicted value meeting a preset threshold condition, that the interaction exists between the candidate RNA sequence and the candidate protein sequence.
Yuan et al. further teaches the attention layer is used to integrate the outputs of the three branches; then, the feature embeddings learned by the three layers are concatenated and fed to a fully connected layer to yield the final output (page 4, fig. 1) of a binary prediction model for each RNA binding protein (page 3, column 2).
Yuan et al. further teaches the high prediction accuracy makes it a useful tool for studying circRNA-RBP interactions (page 7, column 2); and predicting the binding relationship between RNA-binding-proteins and circRNAs (page 7, column 2).
Though Yan et al. teaches using a binary prediction model, it does not explicitly teach the interaction predicted value meeting a preset threshold condition to predict presence of an interaction between the RNA and protein sequence.
Yan et al. reviews RNA-protein binding site predictions that utilize deep learning.
Yan et al. teaches for balanced classification, the main evaluation criteria are Accuracy, Precision, Recall and f-score, which can be easily calculated by the confusion matrix (page 13, column 1), in which the statistics need to set a threshold to convert probability into binary, which is usually set manually; and in this paper, we generally choose 0.5 as the threshold, that is, the probability of classification results is more than 0.5, which is considered as a positive sample, otherwise it is a negative sample (page 13, column 1).
Therefore Yuan et al. teaches a binary prediction model; and Yan et al. teaches determining a preset threshold is necessary for such functions. As such, it would be obvious to one of ordinary skill in the art to apply the known technique of determining a preset threshold for classification in order to yield predictable results.
Claims 12-15 and 18-19 are rejected under 35 U.S.C. 103 as being unpatentable over Yuan et al. (Frontiers in Genetics; Vol. 11: 632861; 2021) in view of Iuchi et al. (Comp and Struct Biotech Journal; Vol. 19: 3198-3208; 2021) and Yan et al. (IEEE Access; vol. 8; 2020), as applied to claims 4, 6, 9, 11, and 17 previously, and in further view of Akiyama et al. (bioRxviv; 66874249; 2021).
Yuan et al. in view of Iuchi et al. and Yan et al. teach a method of vectorizing and inputting k-mers of RNA and protein sequences into recurrent neural networks to predict RNA-protein interactions, as described above.
Claim 12 is directed to constructing a matching feature matrix according to the RNA and protein vector sequence by calculating a matching degree between a base k-mer vector in the RNA vector sequence and an amino acid k-mer vector in the protein vector sequence; and constructing the matching feature matrix by taking the calculated matching degree score as an element of the matching feature matrix.
Yuan et al. teaches constructing the input matrix by using pre-trained RNA and amino acid word embeddings (page 4, column 1); and representing the circRNA by a matrix of size (p + q) × m, i.e., [D1, D2, ⋯ , Dm] (page 4, column 2).
Yuan et al. does not teach constructing the matrix via calculating a matching degree between a base k-mer vector in the RNA vector sequence and an amino acid k-mer vector in the protein vector sequence; nor taking the calculated matching degree score as an element of the matching feature matrix (claim 12).
Akiyama et al. describes a deep representation learning approach to informative RNA-base embedding.
Akiyama et al. teaches token embedding generates a 120-dimensional vector representing four RNA bases as a vector; position embedding generates a 120-dimensional vector that represents position information in an RNA sequence (page 3, column 1); and the element-wise sum of token embedding and position embedding for each base in the given RNA sequence is the input vector to the transformer layer (page 3, column 1), in which the self-attention mechanism is the central component (page 3, column 1).
Akiyama et al. further teaches defining the 𝛺 matrix, for a pairwise alignment of two RNA sequences, intended to be used as a score matrix when calculating the pairwise alignment by letting 𝑍 = [𝑧1,...,𝑧𝑛] and 𝑍′ = [𝑧1 ′,...,𝑧𝑚′ ] denote the embedded representations output from the transformer layer for the input of two RNA sequences of length n and m (page 4, column 1); and putting the match score in position (𝑖,𝑗) is 𝜔𝑖𝑗 of the score matrix 𝛺 (page 5, column 1).
Akiyama et al. further teaches performing RNA family clustering as the second evaluation test to confirm the quality of informative base embedding by defining a new measure of similarity between two RNA sequences with respect to soft symmetric alignment as letting 𝑍 = [𝑧1,...,𝑧n] and 𝑍′ = [𝑧1 ′,...,𝑧𝑚′ ] denote the embedded representations output from the transformer layer for the input of a pair of RNA sequences (page 5, column 1) of length n and m by defining the similarity 𝑠̂ between two RNA sequences as a weighted sum of the normalized inner product between all 𝑧𝑖 and 𝑧𝑗 ′ pairs (page 6, column 1); and using the RIBOSUM 85-60 score matrix as a match score between two bases, for RNA-sequence alignment (page 6, column 1).
Claim 13 is directed to calculating a matching degree between a base k-mer vector in the RNA vector sequence and an amino acid k-mer vector in the protein vector sequence by a process of: calculating the matching degree m(hR, hP) between the i-th base k-mer vector in the RNA vector sequence and the j-th amino acid k-mer vector hP in the protein vector sequence according to:
PNG
media_image1.png
71
317
media_image1.png
Greyscale
where hRi represents a length of hRi and hPj represents a length of hPj.
Akiyama et al. teaches each element 𝜔𝑖𝑗 in the 𝛺 matrix is defined to be the normalized inner product between z𝑖 and 𝑧𝑗 ′ with equation:
PNG
media_image2.png
76
179
media_image2.png
Greyscale
(page 4, column 1); and defining the similarity 𝑠̂ between two RNA sequences is defined to be a weighted sum of the normalized inner product between all 𝑧𝑖 and 𝑧𝑗 ′ pairs, with the following formula (page 6, column 1)
PNG
media_image3.png
221
614
media_image3.png
Greyscale
Claim 14 is directed to constructing a matching feature matrix according to the RNA vector sequence and the protein vector sequence to obtain the matching feature matrix by a process of: performing a dot-product operation on the RNA vector sequence and the protein vector sequence.
Akiyama et al. teaches the self-attention mechanism is described as mapping a query and a set of key-value pairs to an output sequence, where the query, key, and value are all matrices: query 𝑄𝑖 = [𝑞1 𝑖,...,𝑞𝑛 𝑖], key 𝐾𝑖 = [𝑘1 𝑖,...,𝑘𝑛 𝑖] and value 𝑉𝑖 = [𝑣1 𝑖,...,𝑣𝑛 𝑖]; these matrices are the inner products of the learnable weight matrix (page 3, column 1); in the scaled dot-product attention mechanism, each ℎ𝑒𝑎𝑑 calculates the next hidden states by computing the attention-weighted sum of the value vector 𝑣; and an attention coefficient is the output of the softmax function applied to the dot product of the query and key (page 4, column 1).
Claim 15 is directed to performing feature extraction on the matching feature matrix by using a feature extraction network to obtain an original feature; and performing an operation on the original feature using a third mapping matrix.
Akiyama et al. teaches finally, 𝐻 ℎ𝑒𝑎𝑑’s, calculated by different set of {𝑊𝑖 𝑄,𝑊𝑖 𝐾,𝑊𝑖 𝑉} are concatenated; the inner product of this concatenation matrix and 𝑊𝑂 yields the output sequence C (page 4, column 1); and after the transformer layer process including multi-head attention is performed 6 times, the informative base embedding denoted 𝑍 is obtained (page 4, column 1).
Claim 18 is directed to obtaining an interaction predicted value between the candidate RNA sequence and the candidate protein sequence according to the extracted matching feature by a process of: inputting the extracted matching feature into a classifier, and outputting a probability of the presence of the interaction between the candidate RNA sequence and the candidate protein sequence.
Yuan et al. teaches the attention layer is used to integrate the outputs of the three branches; then, the feature embeddings learned by the three layers are concatenated and fed to a fully connected layer to yield the final output (page 4, fig. 1).
Yuan et al. does not teach outputting a probability of the presence of the interaction between the RNA and protein sequence.
Akiyama et al. teaches in this training model, a classification layer is built on top of the output of the transformer layer; and finally, the output probability of each base is calculated using the softmax function (page 4, column 1).
Claim 19 is directed to wherein the probability of the presence of the interaction between the candidate RNA sequence and the candidate protein sequence is
PNG
media_image4.png
86
288
media_image4.png
Greyscale
wherein represents the candidate RNA sequence, p represents the candidate protein sequence, C. represents a first feature value in the extracted matching feature, and ci represents a second feature value in the extracted matching feature.
Akiyama et al. teaches the output probability of each base is calculated using the softmax function (page 4, column 1).
The equation provided represents a generic softmax function.
Therefore Akiyama et al. teaches methods of using a self-attention mechanism within a deep learning context to calculate, apply, and extract features from a matrix that includes RNA sequence embeddings, in the form of base embeddings. Akiyama et al. further teaches that similarly, base embedding can be applied to RNA interactome, RNA-protein interaction, and RNA-RNA interaction in which the RNA secondary structure acts on the interaction between molecules (page 11, column 1). As such, it would be obvious to one of ordinary skill in the art to apply the matrix feature extraction and prediction techniques of Akiyama et al. to a deep learning method of RNA-protein interaction prediction, with a reasonable expectation of success.
Conclusion
No claims are currently allowed.
Correspondence
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Milana Thompson whose telephone number is (571)272-8740. The examiner can normally be reached Monday - Friday, 9:00-6:00 ET.
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) 272-1113. 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.
/M.K.T./Examiner, Art Unit 1687
/Karlheinz R. Skowronek/Supervisory Patent Examiner, Art Unit 1687
Prosecution Insights