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 .
Status of the Claims
Claims 1-25 are pending.
Claims 1-25 are examined on the merits.
Priority
Applicant claims priority to provisional application 62/921,452 filed 06/18/2019. Applicant’s claim for the benefit of a prior-filed application under 35 U.S.C. 119(e) or under 35 U.S.C. 120, 121, 365(c), or 386(c) is acknowledged.
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 07/15/2022, 03/31/2025 and 09/26/2025 are acknowledged. A signed copy of the corresponding 1449 form has been included with this Office action.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1 and 11 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Gjoneska et al. (2015), Conserved epigenomic signals in mice and humans reveal immune basis of Alzheimer’s disease, Nature, Vol 518, pp 365-369 (cited in the IDS filed 07/15/2022).
Regarding claim 1, Gjoneska teaches a method for generating, by a data processing system (page 365, col 2, para 3: “We used ChromHMM [a data processing system: (http://compbio.mit.edu/ChromHMM/)] to learn a chromatin state model ... defined by recurrent combinations of histone modifications, consisting of promoters, enhancers"), data representing a synthetic genetic sequence (abstract: "we profile transcriptional and chromatin state dynamics across early and late pathology in the hippocampus of an inducible mouse model of AD-like neurodegeneration. We find a coordinated downregulation of synaptic plasticity genes and regulatory regions, and upregulation of immune response genes and regulatory regions", methods, page 2, col 1, para 2-3 "Mouse hippocampus was collected immediately after euthanasia. Chromatin immunoprecipitation was then performed ... H3K4me1 [antibodies, and antibodies to 6 other modified chromatin marks] were used to immunoprecipitate endogenous proteins. ... Sequencing libraries were prepared from ...1.C5 ng ChiP (or input) DNA... Reads were mapped to the mm¢9 reference mouse genome ... roughly 55.C60 million unique reads were obtained for each histone modification", methods page 1, col 1, para 6 - col 2, para 1 "We used ChromHMM to learn combinatorial chromatin states ... ChromHMM was trained using all seven chromatin marks ... Reads from replicate data sets were pooled before learning states. ... read counts were computed in non- overlapping 200-bp bins across the entire genome; each bin was discretized into two levels, 1 indicating enrichment, and 0 indicating no enrichment. The binarization was performed by comparing ... read counts to corresponding whole-cell extract control read counts within each bin", methods, page 2, col 2 para 3-4 "putative binding sites based on transcription factor binding site motifs were identified ... transcription factor binding sites were further clustered based on similarity ... The real transcription factor binding site motifs in the category of interest were compared [to] shuffled control motifs that preserved nucleotide content ... After identifying significant transcription factor binding sites in categories or regulatory regions, we collapsed the results into clusters of almost identical motifs, representing families ... A total of 14 oligonucleotide gBlocks ... ranging in 500.C1,000 nucleotides in length, and corresponding to 10 enhancer regions were synthesized. Each gBlock contained a constant ... region, for direct cloning into an EcoRV ... linearized minimal promoter firefly luciferase vector ... Transfections into BV-2 and N2a cells were performed"), configured for labeling at least one cell type by causing expression of a marker in the at least one cell type (methods page 2, col 2 para 4 "10 enhancer regions were synthesized ... Transfections into BV-2 and N2a cells were performed", page 366, col 2, para 6 "To verify whether the increased-level putative enhancer regions were indeed functional, we used a luciferase reporter assay to evaluate their ability to drive in vitro gene expression in immortalized murine microglial {i.e. immune cells] (BV-2) and neuroblastoma (N2a) cell lines. Eight of the nine increased-level human orthologues tested were indeed able to drive in vitro reporter expression. Two of these, BIN1 and ZNF710, were active in both cell types, while the remaining six showed a BV-2-cell-specific increase in luciferase expression"), the method comprising: receiving training data associating at least one feature of a genetic sequence of the at least one cell type with expression of a hallmark of the at least one cell type (page 365, col 2, para 3 "We used ChromHMM... to learn a chromatin state model ... defined by recurrent combinations of histone modifications, consisting of promoters, enhancers”, methods, page 1, col 1, para 6 “ChromHMM was trained using all seven chromatin marks", methods page 2, col 1 para 4 "To estimate computationally the relative composition of the neural and immune cell types we compared ... expression patterns in our data set to a set of established cell-type-specific-markers"); training, based on the training data, a model configured to generate data representing a synthetic genetic sequence (methods page 1, col 1, para 6 - col 2, para 1 "ChromHMM was trained using all seven chromatin marks ... Reads from replicate data sets were pooled before learning states. ... read counts were computed in non-overlapping 200-bp bins across the entire genome; each bin was discretized into two levels, 1 indicating enrichment, and 0 indicating no enrichment ... We trained several models ... We decided to use a 14-state model for all further analyses as it captured all the key interactions between the chromatin marks ... we used the ChromHMM package to compute the overlap and neighborhood enrichments of each state relative to coordinates of known gene annotations. The trained model was then used to compute the posterior probability of each state for each genomic bin ... The regions were labelled using the state with the maximum posterior probability", methods, page 2, col 2 para3 “putative binding sites based on transcription factor binding site motifs were identified ... transcription factor binding sites were further clustered based on similarity ... After identifying significant transcription factor binding sites in categories or regulatory regions, we collapsed the results into clusters of almost identical motifs, representing families”) receiving input data comprising the at least one feature [e.g., a cell-specific enhancer]; and generating the synthetic genetic sequence in response to receiving the input data comprising the at least one feature (methods page 2, col 2 para 4 "10 enhancer regions were synthesized ... Transfections into BV-2 and N2a cells were performed", page 366, col 2, para 6 "To verify whether the increased-level putative enhancer regions were indeed functional, we used a luciferase reporter assay to evaluate their ability to drive in vitro gene expression in immortalized murine microglial [i.e. immune cells] (BV-2) and neuroblastoma (N2a) cell lines. Eight of the nine increased- level human orthologues tested were indeed able to drive in vitro reporter expression. Two of these, BIN1 and ZNF710, were active in both cell types, while the remaining six showed a BV-2-cell-specific increase in luciferase expression").
Regarding claim 11, Gjoneska teaches a method for generating, by a data processing system, data representing a synthetic genetic sequence active in a cell type (e.g. immune cells), as discussed for claim 1. Gjoneska further teaches generating a [vector] nucleic acid sequence comprising the synthetic genetic sequence (methods page 2, col 2 para 3-4 "putative binding sites based on transcription factor binding site motifs were identified ... transcription factor binding sites were further clustered based on similarity ... The real transcription factor binding site motifs in the category of interest were compared [to] shuffled control motifs that preserved nucleotide content ... After identifying significant transcription factor binding sites in categories or regulatory regions, we collapsed the results into clusters of almost identical motifs, representing families ... A total of 14 oligonucleotide gBlocks ... ranging in 500.1,000 nucleotides in length, and corresponding to 10 enhancer regions were synthesized. Each gBlock contained a constant 5'-GCTAGCCTCGAGGAT and 3'- ATCAAGATCTGGCCT region, for direct cloning into an EcoRV ... linearized minimal promoter firefly luciferase vector").
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 2-6, 8-10, 16-18, and 20-24 are rejected under 35 U.S.C. 103 as being obvious over Gjoneska et al. (2015) as applied to claims 1 and 11 above, in view of Wang et al. (April 2019), Synthetic Promoter Design in Escherichia coli based on Generative Adversarial Network, bioRxiv, pp 1-10 (cited in the IDS filed 07/15/2022).
Gjoneska et al. is applied to claims 1and 11 above.
Regarding claim 2, Gjoneska does not expressly teach that the model comprises a generative adversarial network (GAN). Wang teaches design of synthetic promoters using said GAN (page 2, col 1, para 2 "Here, we proposed a GAN-based approach for de nova promoter sequence design and validated the activities of the generated promoters in vivo. Taking natural promoter sequences as inputs, our model automatically extracted the important promoter sequence features. Guided by the extracted features, the generator designed novel artificial sequences automatically from the huge sequence space ... More than 45% of the artificial sequences selected by our prediction model could be experimentally validated as functional promoters, and a number of them showed comparable or even higher activities than strong constitutive promoters as well as their strongest mutants.", see Fig. 1).
Regarding claim 3, Gjoneska and Wang teach a method for generating, by a data processing system comprising data used in a generative adversarial network (GAN), to produce a synthetic genetic sequence, as discussed for claim 2. Wang further teaches that the GAN is a deep convolutional GAN (DCGAN) (page 3, col 1, para 2 "we introduced a deep convolutional GAN (DCGAN) based network structure into synthetic promoter design (Figure S2(a))").
Regarding claim 4, Gjoneska and Wang teach a method for generating, by a data processing system comprising data used in a generative adversarial network (GAN), to produce a synthetic genetic sequence, as discussed for claim 2. Although Wang does not recite a generator network configured to receive latent random noise data as the input data, Wang teaches a generator network configured to receive latent random data as the input data and generate the synthetic genetic sequence; and a discriminator network configured to generate a probability value representing whether the input sequence is drawn from the synthetic genetic sequence from the generator network or from a distribution of natural genetic sequences (page 2, col 1, para 3 - col 2 para 1 "The GAN model has two main adversarial components, the discriminator and generator. The basic workflow of GAN goes as follows: The generator takes samples from the low-dimensional latent variable to generate fake promoters, and the discriminator evaluates the divergence between the current generated fake promoters and the natural promoters. By training the discriminator and generator iteratively, the discriminator gradually learns to act as a more meticulous critic of the current fake promoters. And the generator learns to improve its skills of mimicking the naturally occurring promoters in order to the deceive the discriminator. By producing promoters from the finally trained generator, GAN could help us navigate the potential sequence space to generate new synthetic promoters."). Thus, although Wang does not recite the term latent noise data, Wang teaches the equivalent workflow utilizing the term low-dimensional latent variable, in which the generator iteratively improves latent [noise] variable derived promoters and the discriminator ranks or excludes possible sequences (see claim 1).
Regarding claim 5, Gjoneska and Wang teach a method for generating, by a data processing system comprising data used in a generative adversarial network (GAN), to produce a synthetic genetic sequence, as discussed for claim 4. Wang further teaches training the generator network and the discriminator network of the GAN by alternating the input data between synthetic sequence derived from the latent noise data [latent variables] input to the generator network and natural genetic sequence data [e.g. E. Coli genome ] (see Fig. 1 of Wang, page 2, col 1, para 2-col 2, para 1 "The GAN model has two main adversarial components, the discriminator and generator. The basic workflow of GAN goes as follows: The generator takes samples from the low-dimensional latent variable [i.e. latent noise data] to generate fake promoters, and the discriminator evaluates the divergence between the current generated fake promoters and the natural promoters. By training the discriminator and generator iteratively, the discriminator gradually learns to act as a more meticulous critic of the current fake promoters. And the generator learns to improve its skills of mimicking the naturally occurring promoters in order to the deceive the discriminator. By producing promoters from the finally trained generator, GAN could help us navigate the potential sequence space to generate new synthetic promoters.").
Regarding claim 6, Gjoneska and Wang teach a method for generating, by a data processing system comprising data used in a generative adversarial network (GAN), to produce a synthetic genetic sequence, as discussed for claim 5, Gjoneska further teaches receiving input data representing endogenous genomic sequences underlying previously profiled open chromatin regions [activated promoters marked by chromatin modification] of at least two cell types (page 365, col 2, para 3 "For epigenome analysis, we used chromatin immunoprecipitation sequencing (ChIP-seq) to profile ... chromatin marks9: histone 3 Lys 4 trimethylation (H3K4me3; associated primarily with active promoters); H3K4me1(enhancers);H3K27 acetylation (H3K27ac; enhancer/ promoter activation) ... We used ChromHMM ... to learn a chromatin state model ... defined by recurrent combinations of histone modifications, consisting of promoters, enhancers", page 368, col 2, para 1 "These results are consistent with a model in which increased immune susceptibility to environmental factors during ageing and cognitive decline is mediated by interactions between genetically driven immune cell dysregulation and environmentally driven epigenomic alteration in neuronal cells", methods page 2, col 1, para 5 "We also compared our data to a published study of microglial activation in another mouse model of’), and updating the discriminator network for distinguishing between the at least two cell types (methods, page 1, col 1, para 6 "ChromHMM was trained using all seven chromatin marks", methods page 2, col 1 para 4 “To estimate computationally the relative composition of the neural and immune cell types we compared the changing expression patterns in our data set to a set of established cell -type-specific-markers").
Regarding claim 8, Gjoneska and Wang teach a method for generating, by a data processing system comprising data used in a generative adversarial network (GAN), to produce a synthetic genetic sequence, as discussed for claim 5. Wang further teaches optimizing a feature of the synthetic genetic sequence [a functional promoter] (page 2, col 1, para2 “Here, we proposed a GAN-based approach for de nova promoter sequence design and validated the activities of the generated promoters in vivo. Taking natural promoter sequences as inputs, our model automatically extracted the important promoter sequence features. Guided by the extracted features, the generator designed novel artificial sequences automatically from the huge sequence space ... More than 45% of the artificial sequences selected by our prediction model could be experimentally validated as functional promoters, and a number of them showed comparable or even higher activities than strong constitutive promoters as well as their strongest mutants.”), by applying a class label [classify output, fake from real] to both the generator network and the discriminator network (page 7, col 2, para 3 "More specifically, GAN model contains two parts, the generator and the discriminator. The aim of generator is to generate fake samples that could not be distinguished ... [from] real samples by the discriminator, whereas the aim of discriminator is learning to classify the fake samples and the real samples", page 5 col 2, para 2 "Specifically, the generator samples from low-dimensional random variables to generate fake promoters. Under the guidance of discriminator, the generator could progressively learn the effective generation strategy to transform random variables into functional promoters.”).
Regarding claim 9, Gjoneska and Wang teach a method for generating, by a data processing system comprising data used in a generative adversarial network (GAN), to produce a synthetic genetic sequence, as discussed for claim 8. Wang further teaches that the class label is configured to force a first probability value for a first input type and a second probability value for a second input type that is different from the first input type [i.e. assigns a probability to true promoters and generated promoters] (page 3, col 2, para 3 "the WGAN-GP [Wasserstein generative adversarial network with gradient penalty] model uses the earth-moving distance ... to measure the distance between the probability distribution of true samples and generated samples ... it could progressively train the model until the generated promoter distribution converge to the natural one. By using earth-moving distance, the motif logo of -10 and -35 region was nearly perfectly learned by the WGAN-GP model (Figure 3(a) (b)- compare Fig 3a and 3b)").
Regarding claim 10, Gjoneska and Wang teach a method for generating, by a data processing system comprising data used in a generative adversarial network (GAN), to produce a synthetic genetic sequence, as discussed for claim 8. Wang further teaches that the feature represents an enhancer of an activity [e.g. increases expression from a promoter] (page 2, col 1, para 2 "Here, we proposed a GAN- based approach for de nova promoter sequence design and validated the activities of the generated promoters in vivo. Taking natural promoter sequences as inputs, our model automatically extracted the important promoter sequence features. Guided by the extracted features, the generator designed novel artificial sequences automatically from the huge sequence space ... More than 45% of the artificial sequences selected by our prediction model could be experimentally validated as functional promoters, and a number of them showed comparable or even higher activities than strong constitutive promoters as well as their strongest mutants.").
Regarding claim 16, Gjoneska teaches receiving results data representing a delivery of a nucleic acid including the synthetic genetic sequence to cell types of an organism, the results data representing a successful labeling of the at least one cell type (methods, page 2, col 2 para 3-4 “putative binding sites based on transcription factor binding site motifs were identified ... transcription factor binding sites were further clustered based on similarity ... The real transcription factor binding site motifs in the category of interest were compared [to} shuffled control motifs that preserved nucleotide content ... After identifying significant transcription factor binding sites in categories or regulatory regions ...A total of 14 oligonucleotide gBlocks ... ranging in 500.1,000 nucleotides in length, and corresponding to 10 enhancer regions were synthesized ... for direct cloning into an EcoRV ... linearized minimal promoter firefly luciferase vector.", page 366, col 2, para 6 "To verify whether the increased-level putative enhancer regions were indeed functional, we used a luciferase reporter assay to evaluate their ability to drive in vitro gene expression in immortalized murine microglial (BV-2) and neuroblastoma (N2a) cell lines. Eight of the nine increased-level human orthologues tested were indeed able to drive in vitro reporter expression. Two of these, BIN1 and ZNF710, were active in both cell types, while the remaining six showed a BV-2-cell-specific increase in luciferase expression ... confirming both functional conservation and tissue specificity of increased-level enhancer regions"). Although Gjoneska does not expressly recite updating the model using the results data; and generating an updated synthetic genetic sequence based on the updated model, Gjoneska teaches using models to identify the state of promoters/enhancers (active or repressed) (methods page 1, col 1, para 6 - col 2, para 4 ChromHMM was trained using all seven chromatin marks in virtual concatenation mode across all conditions. Reads from replicate data sets were pooled before learning states. The ChromHMM parameters used are as follows ... read counts were computed in non- overlapping 200-bp bins across the entire genome; each bin was discretized into two levels, 1 indicating enrichment, and 0 indicating no enrichment ... The trained model was then used to compute the posterior probability of each state for each genomic bin in each condition.”).
Regarding claims 17 and 18, Gjoneska does not expressly teach receiving training data including the hallmark representing marker positive results marker negative results, or both; extracting at least one feature corresponding to the marker positive result or to the marker negative result; and adding the at least one feature to the model. Wang teaches a model for generating a synthetic promoter sequence using features (hallmarks) from natural promoters (page 2, col 1, para 1 “Taking natural promoter sequences as inputs, our model automatically extracted the important promoter sequence features. Guided by the extracted features, the generator designed novel artificial sequences automatically from the huge sequence space. These Al-generated promoters mimic key characteristics of the natural promoters such as k-mer frequency, .10 and .35 motifs and their spacing constraint, while show low global sequence similarity to the natural promoters and the E. coli genome. More than 45% of the artificial sequences selected by our prediction model could be experimentally validated as functional promoters, and a number of them showed comparable or even higher activities than strong constitutive promoters as well as their strongest mutants") wherein the at least one feature is a set of k-mer ... counts, and wherein extracting the at least one feature comprises scanning the sequence to determine the set of k-mer counts that form the sequence (page 3, col 1, para 1 "The feature learning ability of DCGAN model was analyzed by ... k-mer frequency - the k-mer frequency represents the high order dependency of nucleotides in promoter sequences").
Regarding claim 20, Gjoneska does not expressly teach that the model comprises a neural network. Wang teaches a model for generating a functional synthetic promoter sequence using features (hallmarks) from natural promoters, comprising a neural network (page 2, col 1, para 2 "For the aim of generating functional synthetic promoters, we introduced a deep learning-based framework (Figure 1), including a GAN network for de nova promoter generation and a deep convolutional neural network (CNN) for promoter strength prediction. Then activities of the generated artificial promoters were tested by fluorescent protein expression level in E. coli.", page 2, col 1, para 1 “More than 45% of the artificial sequences selected by our prediction model could be experimentally validated as functional promoters, and a number of them showed comparable or even higher activities than strong constitutive promoters as well as their strongest mutants").
Regarding claim 21, Gjoneska and Wang teach a method for generating, by a data processing system comprising a neural network, data representing synthetic promoter/enhancer sequences, as discussed for claim 20. Wang further teaches that the neural network is a convolutional neural network (page 2, col 1, para 2 "we introduced a deep learning-based framework (Figure 1), including ... a deep convolutional neural network (CNN) for promoter strength prediction").
Regarding claim 22, Gjoneska and Wang teach a method for generating, by a data processing system, data representing synthetic promoter/enhancer sequences, as discussed for claim 20. Although Wang does not expressly recite that the neural network comprises one or more weight values each associated with a feature of the synthetic genetic sequence, Wang teaches extracting the important features (page 2, col 1, para 2 "Taking natural promoter sequences as inputs, our model automatically extracted the important promoter sequence features. Guided by the extracted features, the generator designed novel artificial sequences", page 3, col 1, para 1 “we introduced a deep convolutional GAN (DCGAN) based network structure into synthetic promoter design ... Previous studies on genome sequence pattern learning have shown that convolutional layers could extract motif features and multilayer network could learn the motif combinatorial pattern of specific genomic regions"). Based on Wang's teaching, the DCGAN extracts the promoter sequence features, which is equivalent to weighting said features with the neural network convolutional layers, thus, it would have been obvious to an artisan of ordinary skill in the art to use Wang's DCGAN as a means to weight features of the synthetic genetic sequence to produce synthetic promoter/enhancer sequences
Regarding claim 23, Gjoneska and Wang teach a method for generating, by a data processing system, data representing synthetic promoter/enhancer sequences, as discussed for claim 20. Wang further teaches that the feature comprises a set of k-mer counts (e.g. frequency) of the genetic sequence (page 3, col 1, para 1 "The feature learning ability of DCGAN model was analyzed by ... k-mer frequency - the k-mer frequency represents the high order dependency of nucleotides in promoter sequences").
Regarding claim 24, Gjoneska and Wang teach a method for generating, by a data processing system, data representing synthetic promoter/enhancer sequences, as discussed for claim 20. Gjoneska further teaches that the feature represents a transcription factor binding motif (methods page 2, col 2, para 3 “putative binding sites based on transcription factor binding site motifs were identified independent of conservation ... real transcription factor binding site motifs in the category of interest were compared shuffled control motifs that preserved nucleotide content ... After identifying significant transcription factor binding sites in categories or regulatory regions, we collapsed the results into clusters of almost identical motifs, representing families.").Case for prima facie obviousness:
The claims are directed to creating a synthetic sequence configured for labelling parvalbumin (PV) positive and negative cell types by utilizing a training DCGAN model as a means of improving methods of genomic assays to isolate for cell type-specific mechanisms.
Claim 2 recites the model of claim 1 to be a generative adversarial network (GAN). Since Wang teaches that GAN-designed promoter (or enhancers) may be produced from a large sequence space, such as the reads produced in the method of Gjoneska, with ability to automatically extract the important promoter sequence features, it would have been obvious to an artisan of ordinary skill in the art to combine GAN with the method of Gjoneska to capture complex, high-dimensional features that would otherwise be missed by bulk tissue genomic assays.
Claim 16 recites “updating the model” using the results. Based on Gjoneska's teaching, it would have been obvious to an artisan of ordinary skill in the art to include updating the trained model with results from testing the ability of identified promoters/enhancers to exhibit cell-specific expression, because said updating would include relevant or exclude irrelevant promoters/enhancers.
Claims 17 and 18 recite extracting at least one feature corresponding to PV+ or PV- and adding it to the model where the feature is a set of k-mer counts. Claims 21-23 similarly recite the CNN and k-mer counts as a feature of the sequence. Since Wang's model (introducing a deep convolutional GAN based network structure into synthetic promoter design (Figure S2(a))) allows generating, by a data processing system, a functional synthetic promoter sequence, as well as predicting promoter strength, it would have been obvious to experiment with combining the methods and programs of Wang with the method of Gjoneska to improve identification of functional, cell-type specific promoters and enhancers.
In KSR Int 'l v. Teleflex, the Supreme Court, in rejecting the rigid application of the teaching, suggestion, and motivation test by the Federal Circuit, indicated that “The principles underlying [earlier] cases are instructive when the question is whether a patent claiming the combination of elements of prior art is obvious. When a work is available in one field of endeavor, design incentives and other market forces can prompt variations of it, either in the same field or a different one. If a person of ordinary skill can implement a predictable variation, § 103 likely bars its patentability.” KSR Int'l v. Teleflex lnc., 127 S. Ct. 1727, 1740 (2007).
This combination of Gjoneska and Wang would have been expected to yield the predictable result of a more comprehensive synthetic genetic sequence as they are both in the same field of synthetic promoter generation. Therefore, the invention would have been prima facie obvious to one of skill in the art before the effective filing date of the application, absent evidence to the contrary.
Claims 12-15 are rejected under 35 U.S.C. 103 as being unpatentable over Gjoneska and Wang applied to claims 1 and 11 above, and further in view of US 2012/0124685 A1 hereinafter Henikoff, (cited in the IDS filed 07/15/2022).
Regarding claim 12, Gjoneska teaches that the synthetic sequence is operably linked to a nucleotide sequence encoding a reporter plasmid (methods, page 2, col 2 para 4 "After identifying significant transcription factor binding sites in categories or regulatory regions ... A total of 14 oligonucleotide gBlocks ... ranging in 500.1,000 nucleotides in length, and corresponding to 10 enhancer regions were synthesized. Each gBlock contained a constant 5'-GCTAGCCTCGAGGAT and 3'- ATCAAGATCTGGCCT region, for direct cloning into an EcoRV ... linearized minimal promoter firefly luciferase vector"), but does not expressly teach that the synthetic genetic sequence is operably linked to a nucleotide sequence encoding a marker.
Henikoff teaches vectors for labeling cell types wherein a promoter sequence is inserted (abstract "compositions, methods, and kits for generating ... tagged nuclei of specific cell types ... under the control of a cell type-specific promoter.", para [0167] " the present invention provides a vector ... for selectively labeling nuclei in a cell type of interest", para [0016] " In some embodiments, the cell type of interest is a neuron, such as a post-mitotic neuron." para [0195] “expression cassettes are adapted to receive a promoter ... to be operationally linked to the sequence encoding the fusion protein ... For example, the expression cassette can include an insertion site flanked by one or more restriction enzyme recognition sequences for insertion of a promoter sequence”).
Regarding claim 13, Gjoneska and Henikoff teach a method for generating synthetic genetic sequences used with tagged (marker) proteins to label specific cell types, as discussed for claim 12. Henikoff further teaches that the marker is a tagged Sun 1. fusion polypeptide (para [0079] "a nucleic acid construct used to transgenically express a nuclear tagging fusion (NTF) protein based on the Sun- 1 protein and containing a domain encoding SUN domain (SD) to embed the protein in the inner nuclear membrane INM of the nuclear envelope ... a domain encoding GFP at the 3' end relative to the SUN-encoding domain results in an affinity reagent binding region and visualization tag being C-terminal to the SUN domain and localizing in the lumen (L) of the nuclear membrane").
Regarding claims 14 and 15, Gjoneska teaches that the synthetic sequence is operably linked to a nucleotide sequence encoding a reporter plasmid (methods, page 2, col 2 para 4 “After identifying significant transcription factor binding sites in ‘categories or regulatory regions ... 10 enhancer regions were synthesized ... for direct cloning into an EcoRV ... linearized minimal promoter firefly luciferase vector"), but does not expressly teach that the synthetic genetic sequence vector comprises a lentivirus nucleic acid sequence. Henikoff teaches lentivirus vector for use in cloning promoter/enhancers (para [0167] " the present invention Provides a vector ... for selectively labeling nuclei in a cell type of interest", para [0156] “the term “vector" is a nucleic acid molecule, preferably self-replicating, which transfers and/or replicates an inserted nucleic acid molecule into and/or between host cells. Exemplary vectors include ... viral vectors. An example of viral vector is a Lentiviral vector.", para [0195] “expression cassettes are adapted to receive ‘a promoter ... to be operationally linked to the sequence encoding the fusion protein").
Case for prima facie obviousness:
The instant application claims 12-15 recite linking the synthetic genetic sequence to a marker (Sun1 fusion polypeptide) and that the nucleic acid sequence comprises lentivirus.
Applying the KSR standard of obviousness to Gjoneska and Wang in view of Henikoff, one of ordinary skill would be motivated to combine prior art reference teachings to arrive at the claimed invention.
Based on Henikoff's teaching, it would have been obvious to an artisan of ordinary skill in the art to use vectors and cassettes of Henikoff to label specific cells with promoter/enhancers identified using the method of Gjoneska because Henikoff's vectors can easily receive the restriction fragments comprising the sequences of Gjoneska, and can be used to label specific cell types. Therefore, the invention would have been prima facie obvious to one of skill in the art before the effective filing date of the application, absent evidence to the contrary.
Claims 7 and 25 are rejected under 35 U.S.C. 103 as being unpatentable over Gjoneska and Wang as applied to claims 1-6 above, further in view of Henikoff and Woodruff et al. (2007), Networks of parvalbumin-positive interneurons in the basolateral amygdala, Journal of Neuroscience, Vol 27(3), pp 553-563 (cited in the IDS filed 07/15/2022).
Regarding claim 7, Gjoneska and Wang do not expressly teach that the at least two cell types include parvalbumin positive (PV+) and parvalbumin negative (PV-) neurons: and wherein the discriminator network is configured to distinguish between the PV+ and PV- neurons based on the input data. Henikoff teaches vectors comprising neuron-specific promoters (para [0167] "the present invention provides a vector ... for selectively labeling nuclei in a cell type of interest", para [0016] "In some embodiments, the cell type of interest is a neuron, such as a post- mitotic neuron.”, para [0195] “expression cassettes are adapted to receive a promoter ... the expression cassette can include an insertion | site flanked by one or more restriction enzyme recognition sequences for insertion of a promoter sequence") and Woodruff teaches labeling neurons with a parvalbumin promoter expressing a detectable label (abstract "The amygdala is a temporal lobe structure that is required for processing emotional information. Polymodal sensory information enters the amygdala at the level of the basolateral amygdala (BLA) ... Using mice in which EGFP (enhanced green fluorescent protein) is expressed under the control of the parvalbumin promoter, we - characterized the properties of parvalbumin-positive interneurons in the BLA. By making recordings from interneuron-interneuron and interneuron-principal neuron pairs, we analyzed the intrinsic circuitry of the BLA. We show that parvalbumin-positive interneurons can be divided into four subtypes as defined by their firing properties"),
Regarding claim 25, Gjoneska does not expressly teach that a synthetic sequence (promoter) is configured to distinguish between parvalbumin positive (PV+) and parvalbumin negative (PV-) neurons. Henikoff teaches vectors comprising neuron-specific promoters (para [0167] "the present invention provides a vector ... for selectively labeling nuclei in a cell type of interest" para [0016] "In some embodiments, the cell type of interest is a neuron, such as a post-mitotic neuron.", para [0195] "expression cassettes are adapted to receive a promoter ... the expression cassette can include an insertion site flanked by one or more restriction enzyme recognition sequences for insertion of a promoter sequence") and Woodruff teaches labeling neurons with a parvalbumin promoter expressing a detectable label (abstract "The amygdala is a temporal lobe structure that is required for processing emotional information. Polymodal sensory information enters the amygdala at the level of the basolateral amygdala (BLA) ... Using mice in which EGFP (enhanced green fluorescent protein) is expressed under the control of the parvalbumin promoter, we characterized the properties of parvalbumin-positive interneurons in the BLA. By making recordings from interneuron-interneuron and interneuron-principal neuron pairs we analyzed the intrinsic circuitry of the BLA. We show that parvalbumin-positive interneurons can be divided into four subtypes as defined by their firing properties"),
Case for prima facie obviousness:
Claim 7 is directed to specifying the cell types as parvalbumin positive (PV+) and parvalbumin negative (PV-) and claim 25 is directed to specifying that the method of claim 1 is configured to distinguish between the above cell types.
Applying the KSR standard of obviousness of Gjoneska and Wang in view of Henikoff, one of ordinary skill would be motivated to combine prior art reference teachings to arrive at the claimed invention.
Since Gjoneska and Wang teach identifying synthetic promoters/enhancers that are expressed in specific cell types from input data (as discussed for e.g. claims 1, 2, 6), Henikoff teaches vectors for neuron-specific expression, and Woodruff teaches a neuronal GFP expression vector driven by the parvalbumin promoter that labels parvalbumin positive (PV+) cells, it would have been obvious to an artisan of ordinary skill in the art to generate and test promoters/enhancers based on, for example, input data from the parvalbumin promoter/enhancer sequence of Woodruff that is expressed in parvalbumin positive (PV+) neurons (i.e. is fused to a label) and could be used to discriminate parvalbumin positive (PV+) (i.e. co-labels with Woodruff's GFP label) and parvalbumin negative (PV-) neurons (i.e. is expressed in basolateral amygdala neurons not labeled with GFP), because said promoter would be useful in research and or therapeutic applications. Therefore, the invention would have been prima facie obvious to one of skill in the art before the effective filing date of the application, absent evidence to the contrary.
Claim 19 is rejected under 35 U.S.C. 103 as being unpatentable over Gjoneska and Wang as applied to claim 1 above, in view of Meng et al. (2017), Construction of precise support vector machine based models for predicting promoter strength, Quantitative Biology, Vol. 5, pp 90-98 and US 2012/0254077 A1 (hereinafter Porikli); (cited in the IDS filed 07/15/2022).
Regarding claim 19, Gjoneska and Wang do not expressly teach that the model comprises a support vector machine, however, Meng teaches use of a support vector machine (SVM) for determining promoter strength (abstract "The prediction of the prokaryotic promoter strength based on its sequence is of great importance not only in the fundamental research of life sciences but also in the applied aspect of synthetic biology. Much advance has been made to build quantitative models for strength prediction ... As one of the most important machine learning methods, support vector machine (SVM) is more powerful to learn knowledge from small sample dataset and thus supposed to work in this problem.... we constructed SVM based models to quantitatively predict the Promoter strength. A library of 1UU promoter sequences and strength values was randomly divided into two datasets, including a training set (10 sequences) for model training and a test set (10 sequences) for model test ... The results indicate that the prediction performance increases with an increase of the size of training set, and the best performance was achieved at the size of 90 sequences. ... Our results demonstrate the SVM-based models can be employed for the quantitative prediction of promoter strength.”). Meng does not expressly recite a feature space; and a support vector representing a classification border in the feature space, wherein the method comprises adding, to the support vector of the support vector machine, a given set of k-mer counts within a predefined distance of the classification border in the feature space of the support vector machine, however, Porikli teaches SVM for binary classification utilizing boundaries to separate binary classes [e.g. a sequence that forms or is part of a functional promoter] (abstract "Frequency features to be used for binary | classification of data”, [0058) "Here, we describe a novel method that drastically speeds up testing ... and allows us to convert the training and evaluation of a kernel machine into the corresponding operations of a linear machine by mapping data into a relatively low- dimensional feature space that is determined by the distributions of the binary class labels.", para [0059] "We first generate a set of frequency features that are rich enough to approximate the separating boundary between two classes of data points"). Porikli does not recite using k-mer counts [frequency] to aid in classification, however, Wang teaches k-mers (page 3, col 1, para 1 “The feature learning ability of DCGAN model was analyzed by ... k-mer frequency - the k-mer frequency represents the high order dependency of nucleotides in promoter sequences”).
Case for prima facie obviousness:
Claim 19 recites a support vector machine model that creates a cell type classification border to sort data and adding k-mer counts.
Applying the KSR standard of obviousness of Gjoneska and Wang in view of Meng and Porikli, one of ordinary skill would be motivated to combine prior art reference teachings to arrive at the claimed invention.
Since Meng teaches that SVM are useful for classifying promoter strength, Porikli teaches SVM for binary classification in a feature space, utilizing boundaries, and Wang teaches using k-mer frequency to specifically analyze nucleotides in promoter sequences, it would have been obvious to an artisan of ordinary skill in the art to experiment with using SVM, including mapping data (e.g. k-mer frequency) to a feature space having boundaries that allow separation between 2 classes, such as a functional or non- functional promoter.
Conclusion
Claims 1-25 are rejected.
Inquiries
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KRITHIKA R KARUNAKARAN whose telephone number is (571)272-5527. The examiner can normally be reached M-F 8am-5pm EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Larry Riggs can be reached on (571) 270-3062. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of 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.
/K.R.K./Examiner, Art Unit 1686
/LARRY D RIGGS II/Supervisory Patent Examiner, Art Unit 1686