Sequence-based predictive models for identifying regulatory motifs

Abstract

The present invention relates to computer implemented methods for identifying enhancer elements, the method comprising: (a) receiving a first candidate sequence; (b) generating candidate sequences sequence using single base pair saturation mutagenesis; (c) for each promoter sequence generating a plurality of input sequences, each comprising a candidate sequence and the promoter sequence; (d) inputting the input sequences into two or more machine learning models; (e) receiving a predicted chromatin accessibility measurement for each input sequence from the two or more machine learning models; (f) synthesizing a composite measurement for each candidate sequence (g) selecting a top performing sequence; and (h) if the top performing sequence matches the first candidate sequence, identifying the top performing sequence as the enhancer element in the plant, and if the top performing sequence does not match the first candidate sequence, updating the top performing sequence and repeating steps (b) – (g).

Description

Docket No.: 165362002340SEQUENCE-BASED PREDICTIVE MODELS FOR IDENTIFYING REGULATORY MOTIFSCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to and benefit of U.S. Provisional Application No. 63 / 741,765, filed January 3, 2025, the entirety of which is incorporated herein by reference.REFERENCE TO AN ELECTRONIC SEQUENCE LISTING

[0002] The content of the electronic sequence listing (165362002340seqlist.xml; Size:178,633 bytes; and Date of Creation: December 31, 2025) is herein incorporated by reference in its entirety.FIELD OF THE INVENTION

[0003] In some aspects, the present invention relates to computer implemented methods for identifying enhancer elements in plants using machine learning models for predicting chromatin accessibility from a nucleotide sequence.BACKGROUND

[0004] A thorough investigation of the organization, function, and evolution of plant genes can be paramount to ascertaining and manipulating certain complex plant biological processes allowing development of plants with improved traits. In many instances, ascertaining and manipulating such complex plant biological processes may often be performed by determining and manipulating, for example, the genes and regulatory mechanisms controlling these biological processes.

[0005] Manipulation of gene regulatory elements, in some instances, allows for understanding and controlling of regulatory mechanisms. Gene regulatory elements may be classified in terms of their structure, such as cis regulatory sequences, genomic location, and the tissue context in which they are active. For example, an enhancer may be a linear nucleotide fragment of noncoding DNA located adjacent to or in a transcribed DNA strand may interact with other elements, such as promoters to upregulate transcription in one or more tissues. Therefore, developing tools for designing novel enhancers elements in plants has the possibility of allowing for rational engineering of plants to produce plants with beneficial traits.1MF-364940674Docket No.: 165362002340BRIEF SUMMARY

[0006] The present disclosure provides computer implemented methods of identifying enhancer elements in plants using machine learning models trained to predict chromatin accessibility from a nucleotide sequence. The methods described herein eliminate the need to independently design enhancer elements to test in individual genes of interest in a plant or plant part thereof of interest. Rather, the methods disclosed herein can be used to design ubiquitous enhancer elements that can upregulate a gene of interest in any plant tissue. In some aspects, the present disclosure provides methods for identifying an enhancer element in silico and validating the activity of the same.

[0007] Provided herein are computer implemented methods for designing synthetic enhancer elements that may be used to precisely control and finetune increase of gene expression of a gene of interest by altering the number of copies of the enhancer positioned in tandem (either with or without the presence of a linker in between each copy) as will be described and exemplified further herein. The present disclosure provides methods for generating synthetic enhancer elements that may be used to precisely control and finetune gene expression of a gene of interest by further altering the distance of the insertions of the one or more copies of the synthetic enhancer element from the transcription start site of the gene of interest as will be described and exemplified further herein.

[0008] Provided herein are computer implemented methods for identifying an enhancer element in a plant, the method comprising: by one or more computing devices comprising one or more processors and memory: (a) receiving a first candidate sequence from the memory; (b) generating a plurality of candidate sequences from the first candidate sequence using single base pair saturation mutagenesis; (c) for each promoter sequence in a plurality of promoter sequences, generating a plurality of input sequences, each comprising a candidate sequence from the plurality of candidate sequences and the promoter sequence; (d) inputting the plurality of input sequences into two or more machine learning models trained to predict a measurement of chromatin accessibility in one or more different plant tissues from a sequence; (e) receiving, from the two or more machine learning models, a predicted chromatin accessibility measurement for each of the plurality of input sequences; (f) synthesizing a composite measurement for each candidate sequence related to the predicted chromatin accessibility measurement for each input sequence in the plurality of input sequences comprising the candidate sequence from each of the two or more machine learning2MF-364940674Docket No.: 165362002340models; (g) selecting a top performing sequence from the plurality of candidate sequences based on the composite measurement; and (h) if the top performing sequence matches the first candidate sequence, identifying the top performing sequence as the enhancer element in the plant, and if the top performing sequence does not match the first candidate sequence, updating the first candidate sequence to be top performing sequence and repeating steps (b) -(g)-

[0009] In some aspects, the method further comprise experimentally validating the enhancer element increases expression of one or more genes in the plant.

[0010] In some aspects, the first candidate sequence comprises a polynucleotide sequence. In some aspects, the first candidate sequence comprises a randomly generated polynucleotide sequence comprising a uniform distribution of adenine, guanine, cytosine, and thymine nucleotides.

[0011] In some aspects, the first candidate sequence has a length of N nucleotides. In some aspects, N is between 10 and 60. In some aspects, plurality of candidate sequences comprises 3 x IV + 1 candidate sequences.

[0012] In some aspects, the first candidate sequence comprises a polynucleotide sequence comprising a known enhancer motif flanked by one or more variable nucleotides. In some aspects, the first candidate sequence comprises a polynucleotide sequence comprising a known enhancer motif repeated one or more times.

[0013] In some aspects, the known enhancer motif is a k-mer associated with a promoter of a plurality of highly expressed genes. In some aspects, the k-mer comprises a polynucleotide sequence identified by identifying a plurality of promoters associated with highly expressed genes in the plant; and identifying a polynucleotide sequence enriched in sequences for the plurality of promoters. In some aspects, the plurality of promoters have been identified as accessible in a plurality of cell types using snATAC-seq. In some aspects, k-mer comprises a polynucleotide sequence less than about 12 nucleotides.

[0014] In some aspects, the known enhancer motif is held constant during the single base pair saturation mutagenesis. In some aspects, the plurality of candidate sequences comprises 3 x M + 1 candidate sequences, wherein M is the number of variable nucleotides in the first candidate sequence. In some aspects, single base pair saturation mutagenesis comprises separately substituting the nucleotide at each position in the first candidate sequence with an adenine, a thymine, a guanine, and a cytosine nucleotide.3MF-364940674Docket No.: 165362002340

[0015] In some aspects, a promoter sequence in the plurality of promoter sequences is a known promoter sequence element in the genome of the plant.

[0016] In some aspects, each input sequence comprises the candidate sequence at the 5’ end of the promoter sequence. In some aspects, each input sequence comprises the candidate sequence repeated about 1, about 2, or about 3 times.

[0017] In some aspects, the two or more machine learning models have each been trained to predict a measurement of chromatin accessibility a plant tissue based on ATAC-seq data collected from the plant tissue. In some aspects, the two or more machine learning models have been trained to predict chromatin accessibility in different plant tissues. In some aspects, the one or more plant tissues comprise plant tissues from a dicot. In some aspects, the one or more plant tissues comprise plant tissues from a dicot and the plant is a dicot. In some aspects, the one or more plant tissues comprise plant tissues from the same species as the plant.

[0018] In some aspects, a machine learning model of the two or more machine learning models comprises a convolutional neural network (CNN). In some aspects, the CNN is a dilated CNN.

[0019] In some aspects, the one or more plant tissues are selected from a group consisting of bud, cotyledon, flower, flower bud, hypocotyl, leaf, pod, and root.

[0020] In some aspects, synthesizing the composite measurement comprises averaging the predicted chromatin accessibility measurements for each input comprising the candidate sequence from one of the two or more machine learning models and averaging the predicted chromatin accessibility measurements for each candidate sequence across the two or more machine learning models. In some aspects, the composite measurement is based on the chromatin accessibility of the candidate sequence when the candidate sequence is linked to the plurality of promoters in two or more tissues. In some aspects, the selecting the top performing sequence comprises selecting the candidate sequence with the highest composite measurement among the composite measurements for the plurality of candidate sequences.

[0021] In some aspects, the repeating steps (b)-(g) increases a probability the top performing sequence matches the first candidate sequence in a subsequent repetition of the steps. In some aspects, the methods comprise repeating steps (b)-(g) about 10 times.

[0022] In some aspects, selecting the top performing sequence comprises selecting two or more candidate sequences with the highest composite measurement among the composite 4MF-364940674Docket No.: 165362002340measurements for the plurality of candidate sequences and performing steps (b)-(g) with each of the two or more candidate sequences as the first candidate sequence. In some aspects, the top performing sequence matches the candidate sequence if the nucleotide sequence of the top performing sequence is the same as the nucleotide sequence of the candidate sequence.

[0023] In some aspects, the plant is soybean. In some aspects, the ATAC-seq data comprises soybean data collected from Williams 82 soybeans.

[0024] In some aspects, the methods comprise experimentally validating the enhancer element comprises a reporter assay. In some aspects, the reporter assay comprises transforming a polynucleotide vector comprising, the enhancer element operably linked to an endogenous promoter, a transcription start site, and a reporter gene into a plant or part thereof, wherein the reporter gene is a luciferase gene, and testing for expression of the luciferase gene. In some aspects, the polynucleotide vector comprises the enhancer element repeated three times.

[0025] In some aspects, wherein the reporter assay comprises: transforming a first polynucleotide vector comprising, the enhancer element operably linked to an endogenous promoter a transcription start site, and a reporter gene into a first plant or part thereof from the plant; transforming a second polynucleotide vector comprising, a control sequence element operably linked to the endogenous promoter, the transcription start site, and the reporter gene into a second plant or part thereof from the plant; and measuring the expression of the reporter gene in the first plant or part thereof and in the second plant or part thereof, wherein increased expression of the reporter gene in the first plant or part thereof compared to the second plant or part thereof indicates validation of the enhancer element. In some aspects, the first polynucleotide vector comprises the enhancer element repeated three times.

[0026] In some aspects, the control sequence element comprises an octopine synthase (OCS) enhancer element. In some aspects, wherein the control sequence element comprises a G-box element.

[0027] In some aspects, the endogenous promoter is promoter for a target gene. In some aspects, the target gene is a gene selected from a group consisting of AlPlOa, AlPlOb, AML4, CRN, HB-1, RIC1, RIC2, RPF1, JAG1, JAG2, KHZ1, PP2C, TCP5-L, FTla, BS1, BS2, TFLlb, CYP76C-1, CYP76C-2, and NF-YC4.

[0028] In some aspects, the plant or part thereof is a protoplast.5MF-364940674Docket No.: 165362002340

[0029] In some aspects, the methods comprise synthesizing a plurality of polynucleotide vectors, wherein each polynucleotide vector in the plurality of polynucleotides comprises the enhancer element, a different promoter, and a reporter gene; transforming each polynucleotide vectors into two or more plant tissues; measuring expression of the reporter gene in the two or more plant tissues, wherein expression of the reporter gene in the two or more plant tissues and with different promoters indicates validation of the enhancer element.

[0030] In some aspects, wherein the reporter gene is luciferase.

[0031] In some aspects, experimentally validating the enhancer element comprises inserting the enhancer element within about 1 kb of a gene in a plant, or part thereof and measuring expression of the gene, wherein increased expression of the gene compared to endogenous expression of the gene indicates validation of the enhancer element. In some aspects, the plant or part thereof is a protoplast.

[0032] In some aspects, experimentally validating the enhancer element comprises inserting the enhancer element within about 1 kb of a gene in two or more plant tissues and measuring expression of the gene in the two or more plant tissues, wherein increased expression of the gene compared to endogenous expression of the gene in at least two or the two or more plant tissues indicates validation of the enhancer element.

[0033] In some aspects, the enhancer element is inserted into a noncoding genomic region upstream of the gene. In some aspects, the enhancer element is inserted into a noncoding genomic region downstream of the gene. In some aspects, the enhancer element is inserted into a noncoding genomic region within the gene.

[0034] In some aspects, the methods comprise measuring endogenous expression of the gene without inserting the enhancer element. In some aspects, measuring expression comprises performing qPCR or RNA-sequencing.

[0035] Provided herein are methods for identifying an enhancer element in a plant, the method comprising: (a) obtaining a first candidate sequence; (b) generating a plurality of candidate sequences from the first candidate sequence using single base pair saturation mutagenesis; (c) for each promoter sequence in a plurality of promoter sequences, generating a plurality of input sequences, each comprising a candidate sequence from the plurality of candidate sequences and the promoter sequence; (d) inputting the plurality of input sequences into two or more machine learning models trained to predict a measurement of chromatin accessibility in one or more plant tissues from a sequence; (e) receiving, from 6MF-364940674Docket No.: 165362002340the two or more machine learning models, a predicted chromatin accessibility measurement for each of the plurality of input sequences; (f) synthesizing a composite measurement for each candidate sequence related to the predicted chromatin accessibility measurements for each input sequence in the plurality of input sequences comprising the candidate sequence from each of the two or more machine learning models; (g) selecting a top performing sequence from the plurality of candidate sequences based on the composite measurement; (h) if the top performing sequence matches the first candidate sequence, identifying the top performing sequence as the enhancer element in a plant, and if the top performing sequence does not match the first candidate sequence, updating the first candidate sequence to be top performing sequence and repeating steps (b) - (g); (i) synthesizing a polynucleotide vector comprising the candidate enhancer sequence operably linked to an endogenous promoter, a transcription start site, and a reporter gene; j) transforming a plant or part thereof with the polynucleotide vector; (k) transforming a second polynucleotide vector comprising, a control element operably linked to the endogenous promoter, the transcription start site, and the reporter gene into a second plant or part thereof; and (1) measuring the expression of the reporter gene in the first plant or part thereof and in the second plant or part thereof, wherein increased expression of the reporter gene in the first plant or part therefor compared to the second plant or part thereof indicates that the candidate sequence comprises an enhancer element.

[0036] Provided herein are systems for identifying an enhancer element in a plant, the system comprising: one or more processors; a user input device and a memory communicatively coupled to the one or more processors and configured to store instructions that, when executed by the one or more processors, cause the system to: (a) receive a first candidate sequence from the user input device; (b) generate a plurality of candidate sequences from the first candidate sequence using single base pair saturation mutagenesis; (c) for each promoter sequence in a plurality of promoter sequences, generate a plurality of input sequences, each comprising a candidate sequence from the plurality of candidate sequences and the promoter sequence; (d) input the plurality of input sequences into two or more machine learning models trained to predict a measurement of chromatin accessibility in one or more plant tissues from a sequence; (e) receive, from the two or more machine learning models, a predicted chromatin accessibility measurement for each of the plurality of input sequences; (f) synthesize a composite measurement for each candidate sequence related to the predicted chromatin accessibility measurements for each input sequence in the plurality of7MF-364940674Docket No.: 165362002340input sequences comprising the candidate sequence from each of the two or more machine learning models; (g) select a top performing sequence from the plurality of candidate sequences based on the composite measurement; (h) if the top performing sequence matches the first candidate sequence, identify the top performing sequence as the enhancer element in the plant, and if the top performing sequence does not match the first candidate sequence, update the first candidate sequence to be top performing sequence and repeating steps (b) -(g)-

[0037] Provided herein are methods of training a machine learning model to predict a measurement of chromatin accessibility in a dicot tissue from a sequence; obtaining training data comprising ATAC-seq read coverage for the dicot tissue, wherein the ATAC-seq read coverage represents a measurement of chromatin accessibility in the tissue from the dicot; selecting a plurality of positive genomic windows, wherein a positive genomic window is a region of Wm82v4 with ATAC-seq read coverage in the training data; selecting a plurality of negative genomic windows, wherein a negative genomic window is region of Wm82v4 without ATAC-seq read coverage in the training data; and training the machine learning model to predict a measurement of chromatin accessibility from a sequence wherein the training is based on the ATAC-seq read coverage at the plurality of positive genomic windows, ATAC-seq read coverage at the plurality negative genomic windows, the sequence in Wm82v4 corresponding to each of the positive genomic windows, the sequence in Wm82v4 corresponding to each of the negative genomic windows.

[0038] In some aspects, the dicot tissue is a soybean tissue. In some aspects, training the machine learning model comprises optimizing parameters of the machine learning model.

[0039] Also provided herein are methods of enhancing the expression level of a gene in a plant or part thereof, the method comprising: (a) identifying an enhancer element according to any of the computer implemented method described herein; and (b) introducing the identified enhancer element into a plant or part thereof, wherein the enhancer element is operably linked to the gene, and wherein the enhancer element is inserted upstream of a transcription start site associated with the gene, wherein the expression level of the gene in the plant or part thereof comprising the enhancer element is enhanced as compared to the expression level of the gene in a control plant or part thereof lacking the enhancer element.

[0040] Also provided herein are methods of enhancing the expression level of a gene in a plant or part thereof, the method comprising: (a) obtaining a first candidate sequence; (b)8MF-364940674Docket No.: 165362002340generating a plurality of candidate sequences from the first candidate sequence using single base pair saturation mutagenesis; (c) for each promoter sequence in a plurality of promoter sequences, generating a plurality of input sequences, each comprising a candidate sequence from the plurality of candidate sequences and the promoter sequence; (d) inputting the plurality of input sequences into two or more machine learning models trained to predict a measurement of chromatin accessibility in one or more plant tissues from a sequence; (e) receiving, from the two or more machine learning models, a predicted chromatin accessibility measurement for each of the plurality of input sequences; (f) synthesizing a composite measurement for each candidate sequence related to the predicted chromatin accessibility measurements for each input sequence in the plurality of input sequences comprising the candidate sequence from each of the two or more machine learning models; (g) selecting a top performing sequence from the plurality of candidate sequences based on the composite measurement; (h) if the top performing sequence matches the first candidate sequence, selecting the top performing sequence as an enhancer element, and if the top performing sequence does not match the first candidate sequence, updating the first candidate sequence to be top performing sequence and repeating steps (b) - (g); and (i) introducing the enhancer element into a plant or part thereof, wherein the enhancer element is operably linked to the gene, and wherein the enhancer element is inserted upstream of a transcription start site associated with the gene, wherein the expression level of the gene in the plant or part thereof comprising the enhancer element is enhanced as compared to the expression level of the gene in a control plant or part thereof lacking the enhancer element.BRIEF DESCRIPTION OF THE DRAWINGS

[0041] FIG. 1A depicts a strategy to achieve overexpression of a gene. An enhancer element is inserted one, two, or three times in a promoter region upstream of a transcription start site in a gene. The insertion of the enhancer element in the promoter region alters the expression level of the gene (“Edited line expression level”) relative to a wild-type promoter with no enhancer element inserted (“WT expression level”). FIG. IB depicts a strategy of using ATAC-seq data to train a neural network to predict chromatin accessibility within a genome (left panel), and a strategy of interacting with a trained neural network to inspect the sequence features driving chromatin accessibility predictions (right panel).9MF-364940674Docket No.: 165362002340

[0042] FIG. 2A depicts a method of using single base pair saturation mutagenesis to generate candidate enhancer sequences starting with an initial random motif and using trained machine learning models to predict measurements of chromatin accessibility and composite measurements for the candidate enhancer sequences in different promoter contexts. FIG. 2B depicts the predicted chromatin accessibility measurements for eight rounds of the steps of FIG. 2A. The dot or star at the end of each line in the graph indicates convergence, where no single base pair substitution is able to improve the score of the candidate enhancer under investigation. Each line is an evolution trajectory starting from a different random initial motif. The sequences in the legend are the final converged sequences. The sequence marked with a star corresponds to Enhancer 100 (SEQ ID NO: 100) which converged after 3 steps and was selected for subsequent validation experiments.

[0043] FIG. 3A depicts the G-box motif together with 2 base pair (bp) long flanking regions (shaded). FIG. 3B depicts the use of single base pair saturation mutagenesis to generate 256 variations of the G-box motif and 2 bp flanking regions, where the G-box motif is held constant and the 2 bp flanking regions are variable. FIG. 3C depicts a method of using single base pair saturation mutagenesis to generate candidate enhancer sequences based on the G-box motif and using trained machine learning models to predict measurements of chromatin accessibility and composite measurements (scores) for the candidate enhancer sequences in different promoter contexts. FIG. 3D depicts a box plot of the predicted chromatin accessibility measurements for every candidate enhancer sequence (“motif’) tested in every promoter context. Each dot represents a measurement per candidate enhancer sequence per promoter context. FIG. 3E depicts an excerpt of FIG. 3D to show candidate enhancer sequences with the lowest predicted chromatin accessibility measurements across all 5,438 promoters. FIG. 3F depicts an excerpt of FIG. 3D to show candidate enhancer sequences with the highest predicted chromatin accessibility measurements across all 5,438 promoters.

[0044] FIG. 4A depicts a polynucleotide vector for use in a dual luciferase assay to assess activity of the OCS enhancer, the CaMV 35S enhancer, or a candidate enhancer sequence. FIG. 4B depicts control, 3x OCS, 35S enhancer, and experimental enhancer vectors. The relative luminescence ratio of firefly luciferase (FLUC) to Renilla luciferase (RLUC) can be measured from each vector.

[0045] FIG. 5 depicts the relative luminescence ratio of firefly luciferase (FLUC) to Renilla luciferase (RLUC) using the vectors of FIGS. 4A-4B, where FLUC expression is driven by a 10MF-364940674Docket No.: 165362002340promoter containing one copy of the OCS enhancer (lx OCS), three copies of the OCS enhancer inserted in tandem (3x OCS), the CaMV 35S enhancer (35S Enhancer), three copies of the OCS enhancer inserted with spacing (3x OCS w / spacers), three copies of the cCRE enhancer (3x CCRE), 3 copies of the G-box enhancer (3x GBOX), or the wild-type TFLlb promoter (WT Promoter).

[0046] FIG. 6 depicts the relative luminescence ratio of firefly luciferase (FLUC) to Renilla luciferase (RLUC) for dual luciferase assays conducted with a vector containing the wildtype TFLlb promoter (“WT TFLlb control”), a vector containing three copies of the OCS enhancer inserted in tandem (“3x OCS control”), and a vector containing three copies of candidate Enhancer 100 (SEQ ID NO: 100) (“3x_12bp_2666”).

[0047] FIG. 7A depicts the relative luminescence ratio of firefly luciferase (FLUC) to Renilla luciferase (RLUC) for dual luciferase assays conducted with a vector containing the wild-type TFLlb promoter (“WT TFLlb control”), a vector containing three copies of the OCS enhancer inserted in tandem (“3x OCS control”), a vector containing three copies of candidate Enhancer 100 (single copy SEQ ID NO: 100; three copies SEQ ID NO: 138) (“3x_12bp_2666”), a vector containing three copies of candidate Enhancer 115 (single copy SEQ ID NO: 115; three copies SEQ ID NO: 140) (“3x_12bp_2681”), and a vector containing three copies of candidate Enhancer 107 (single copy SEQ ID NO: 107; three copies SEQ ID NO: 139) (“3x_12bp_2673”). FIG. 7B depicts the data of FIG. 7A on a log scale.

[0048] FIGS. 8A-8B depict the relative luminescence ratio of firefly luciferase (FLUC) to Renilla luciferase (RLUC) for dual luciferase assays conducted with a vector containing the wild-type TFLlb promoter (“TFLlb - WT”), a vector containing three copies of the OCS enhancer and the TFLlb promoter (“TFLlb - 3x OCS”), a vector containing the wild-type promoter of a medium-expression control gene (a sample gene known to exhibit medium expression levels within the protoplast assay system as determined by RNA sequencing data) (“MID - WT”), a vector containing three copies of the OCS enhancer and the promoter of the medium-expression control gene (“MID - 3x OCS”), a vector containing the wild-type promoter of a high-expression control gene (a sample gene known to exhibit high expression levels within the protoplast assay system as determined by RNA sequencing data) (“HIGH -WT”), a vector containing three copies of the OCS enhancer and the promoter of the high-expression control gene (“HIGH - 3x OCS”). In FIG. 8A, the dual luciferase assays were conducted with each vector at 55 ng / pL. In FIG. 8B, the dual luciferase assays were conducted with each vector at 110 ng / pL.11MF-364940674Docket No.: 165362002340

[0049] FIG. 9 depicts a computer implemented method for identifying an enhancer element according to some of the embodiments described herein.

[0050] FIG. 10 depicts a method for training a machine learning model to predict a measurement of chromatin accessibility according to some of the embodiments described herein.

[0051] FIG. 11 depicts a computing system for use according to some of the methods and systems of the present application.

[0052] FIGS. 12A-12C depict dual luciferase vectors. FIG. 12A depicts a dual luciferase vector used to test Enhancer 100 (single copy SEQ ID NO: 100; three copies SEQ ID NO: 138). FIG. 12B depicts a dual luciferase vector used to test Enhancer 115 (single copy SEQ ID NO: 115; three copies SEQ ID NO: 140). FIG. 12C depicts a dual luciferase vector used to test Enhancer 107 (single copy SEQ ID NO: 107; three copies SEQ ID NO: 139).

[0053] FIG. 13 depicts dual luciferase assays with each vector at a concentration of 110 ng / pL, illustrating the increase in FLUC / RLUC luminescence for vectors containing Enhancer 100.

[0054] FIG. 14 depicts dual luciferase assays with each vector at a concentration of 110 ng / pL, containing Enhancer 100 in various copies (2x, 3x, 4x, 5x, 6x), illustrating a dependency on copy number for this enhancer element.

[0055] FIG. 15 depicts dual luciferase assays with each vector at a concentration of 110 ng / pL, with 3x_12bp_2666 (SEQ ID NO: 138) and 3x OCS cloned at various distances from the transcription start site of the luciferase reporter gene, illustrating a dependency on distance from the TSS of the enhancer element.DETAILED DESCRIPTION

[0056] Provided herein are method and systems that can be used to identify enhancer elements in plants that are capable of upregulating genes linked to diverse promoters in multiple plant tissues. The methods take advantage of employing machine learning models trained to predict chromatin accessibility from a nucleotide sequence over multiple iterations of testing. These methods allow for testing of candidate sequences connected to multiple promoters and in multiple tissues at a large scale. The identification of enhancer elements that increase expression in diverse tissues and promoter contexts allows for reliable and12MF-364940674Docket No.: 165362002340reproducible control of gene expression and the generation of stable plant phenotypes for valuable agronomic traits.

[0057] Previous approaches to identify enhancers using machine learning models trained to predict chromatin accessibility from an input sequence are limited in their ability to identify enhancer sequences that will likely be active when operably linked to multiple enhancers and in multiple tissues. For example, Taskiran et al., Cell-type-directed design of synthetic enhancer, Nature, vol. 626, pp. 212-220 (2024), describe the use of machine learning models trained using chromatin accessibility measurements from Drosophila melanogaster and human cells to predict tissue specific enhancers. These methods result in “overfitting” of a specific enhancer to a specific sequence context and in a specific tissue.

[0058] The methods and systems extend this work to identify enhancers that function across a wide range of tissues and sequence contexts. Conventional methods for enhancer identification often rely on heuristic or single-tissue-focused models, which are computationally inefficient and fail to generalize across different tissues or promoter contexts. The methods and systems disclosed herein overcome these technical limitations by employing novel integrations of chromatin accessibility predicted from machine learning models trained in diverse tissue types. The methods and systems improve the efficiency, scalability and prediction accuracy of enhancer identification in plants while also reducing computational demands.

[0059] The disclosed methods address the technical limitations of conventional machine learning-based enhancer identification, particularly when applied at large scales. Iterative use of multiple machine learning models to generate a tissue - promoter-agnostic enhancer in a plant would likely cause computational time-outs, resulting in a failure of the system to produce results. Even if computational bandwidth were sufficient, conventional approaches would waste substantial computer resources due to the exponential growth of the combinations required by the analysis. Additionally, these traditional methods often require manual interventions to compare and interpret results across multiple tissue models, introducing inefficiencies and increasing the potential for human error.

[0060] The methods described herein comprise synthesizing predicted chromatin accessibility measurements from machine learning models trained on different tissues using input sequences comprising diverse promoter elements. This synthesis, combined with the use of greedy selection for iteration, and convergence criteria achieves several technical13MF-364940674Docket No.: 165362002340benefits: (1) minimizes the number of individual combinations requiring evaluation (e.g. tissue- specific combinations); (2) reduced the risk of overtraining the one or more machine learning models; and (3) eliminating the need for human intervention. The approaches described herein thus mitigate the probability a computer time outs by enabling large-scale analysis without proportional increasing the number of inputs into the system at each iteration.

[0061] The disclosed methods represent a transformative advance over prior approaches by enabling the identification of enhancers in a computationally efficient manner, solving a technical problem associated with large-scale genomic analyses. For example, the ability to process tissue-agnostic enhancer predictions without exhausting system resources or requiring manual intervention significantly expands the applicability of machine learning in genomic research. This advance directly impacts the development of plant genetic engineering strategies by enabling rapid and reliable identification of tissue-independent regulatory elements.

[0062] In some embodiments, the methods described herein can be used both for de novo generation of enhancers and variant-effect prediction tasks. For de novo enhancer generation, the models can learn to identify the parts of a DNA sequence that are relevant for predicting chromatin accessibility in a context-sensitive way and without relying on motif databases. For variant-effect prediction the models can learn to map genomics sequences to chromatin accessibility and thus, one can in-silico edit the input DNA sequence to predict its effect on chromatin accessibility without requiring in-vivo plant editing.

[0063] All references cited herein are hereby incorporated by reference in their entirety. Definitions

[0064] The use of the terms “a” and “an” and “the” and “at least one” and similar language in the context of describing embodiments of the disclosure (especially in the context of the following claims) are to be construed to cover both the singular and the plural, unless otherwise indicated herein or clearly contradicted by context. The use of the term “at least one” followed by a list of one or more items (for example, “at least one of A and B”) is to be construed to mean one item selected from the listed items (A or B) or any combination of two or more of the listed items (A and B), unless otherwise indicated herein or clearly contradicted by context. The terms “comprising,” “having,” “including,” and “containing” are14MF-364940674Docket No.: 165362002340to be construed as open-ended terms (i.e., meaning “including, but not limited to,”) unless otherwise noted.

[0065] The phrase “allelic variant” as used herein refers to a polynucleotide or polypeptide sequence variant that occurs in a different strain, variety, or isolate of a given organism.

[0066] The term "and / or" where used herein is to be taken as specific disclosure of each of the two specified features or components with or without the other. Thus, the term and / or" as used in a phrase such as "A and / or B" herein is intended to include "A and B," "A or B," "A" (alone), and "B" (alone). Likewise, the term "and / or" as used in a phrase such as "A, B, and / or C" is intended to encompass each of the following embodiments: A, B, and C; A, B, or C; A or C; A or B; B or C; A and C; A and B; B and C; A (alone); B (alone); and C (alone).

[0067] As used herein, “automatically” and its derivatives means “without human intervention,” unless expressly indicated otherwise or indicated otherwise by context.

[0068] As used herein, the term “complex” refers to two or more associated components, such as two or more associated nucleic acids and / or proteins. A complex may include two or more covalently linked nucleic acids and / or proteins, two or more non-covalently linked nucleic acids and / or proteins, or a combination thereof.

[0069] As used herein, the term “endogenous” refers to something that can be found in the organism prior to human intervention. An “endogenous sequence” refers to a DNA sequence located in the genome of the unedited organism.

[0070] As used herein, the term “exogenous” refers to something that cannot be found in the organism prior to human intervention. An “exogenous sequence” refers to a DNA sequence that is not located in the genome of the unedited organism. An exogenous sequence can be an edited sequence, a synthetic sequence, or a sequence from a different organism.

[0071] As used herein, the terms “heritable genetic modification”, “heritable edit”, and “heritable modification” refer to any insertion, substitution, or deletion of any number of nucleotides in the genomic sequence of a plant that is at least present in a meristem cell of the plant, such that at least one progeny of the plant possesses the same altered genomic sequence.

[0072] As used herein, the terms “include,” “includes,” and “including” are to be construed as at least having the features to which they refer while not excluding any additional unspecified features.15MF-364940674Docket No.: 165362002340

[0073] As used herein, “transforming,” “transformation,” or to “transform” refer to any method requiring human intervention which results in the transfer of an exogenous nucleic acid sequence, recombinant DNA construct, or polynucleotide vector into a cell, irrespective of the method used for transfer. This includes, but is not limited to, transfection, particle bombardment, biolistic transformation, or Agrobacterium- mediated transformation.

[0074] As used herein, the phrase “operably linked” refers to a juxtaposition wherein the components so described are in a relationship permitting them to function in their intended manner. For example, a promoter or enhancer is “operably linked” to a sequence to be expressed if the promoter or enhancer affects the transcription of expression of the sequence to be expressed. As used herein, elements may be “operably linked” irrespective of location.

[0075] As used herein, the terms “orthologous,” “ortholog,” or “orthologue” are used to describe genes or the RNAs or proteins encoded by those genes that are from different species but which have the same function. Orthologous genes will typically encode RNAs or proteins with some degree of sequence identity and can also exhibit conservation of sequence motifs, and / or conservation of structural features including RNA stem loop structures.

[0076] As used herein, the term “plant” includes a whole plant and any descendant, cell, tissue, or part of a plant. The term “plant parts” include any part(s) of a plant, including, for example and without limitation: seed (including mature seed and immature seed); a plant cutting; a plant cell; a plant cell culture; or a plant organ (e.g., intact nodal bud, shoot apex or shoot apical meristem, root apex or root apical meristem, lateral meristem, intercalary meristem, zygotic embryo, somatic embryo, ovule, pollen, microspore, anther, hypocotyl, cotyledon, leaf, petiole, stem, tuber, root, flowers, fruits, shoots, and explants). A plant tissue or plant organ may be a seed, protoplast, callus, or any other group of plant cells that is organized into a structural or functional unit. A plant cell or tissue culture may be capable of regenerating a plant having the physiological and morphological characteristics of the plant from which the cell or tissue was obtained, and of regenerating a plant having substantially the same genotype as the plant. Regenerable cells in a plant cell or tissue culture may be embryos, protoplasts, meristematic cells, callus, pollen, leaves, anthers, roots, root tips, silk, flowers, kernels, ears, cobs, husks, or stalks. In contrast, some plant cells are not capable of being regenerated to produce plants and are referred to herein as “non-regenerable” plant cells.16MF-364940674Docket No.: 165362002340

[0077] As used herein, the term “polynucleotide” refers to a nucleic acid molecule containing multiple nucleotides and encompasses both “oligonucleotides” (defined here as a polynucleotide molecule of between 2-25 nucleotides in length) and polynucleotides of 26 or more nucleotides. Polynucleotides are generally described as single- or double-stranded. Where a polynucleotide contains double-stranded regions formed by intra- or intermolecular hybridization, the length of each double- stranded region is conveniently described in terms of the number of base pairs. Aspects of this invention include the use of polynucleotides or compositions containing polynucleotides; embodiments include one or more oligonucleotides or polynucleotides or a mixture of both, including single- or double-stranded RNA or single-or double-stranded DNA or double- stranded DNA / RNA hybrids or chemically modified analogues or a mixture thereof. In various embodiments, a polynucleotide includes a combination of ribonucleotides and deoxyribonucleotides (e.g., synthetic polynucleotides consisting mainly of ribonucleotides but with one or more terminal deoxyribonucleotides or synthetic polynucleotides consisting mainly of deoxyribonucleotides but with one or more terminal dideoxyribonucleotides), or includes non-canonical nucleotides such as inosine, thiouridine, or pseudouridine. In embodiments, the polynucleotide includes chemically modified nucleotides (see, e.g., Verma and Eckstein Annu. Rev. Biochem. 1998, 67: 99-134); for example, the naturally occurring phosphodiester backbone of an oligonucleotide or polynucleotide can be partially or completely modified with phosphorothioate, phosphorodithioate, or methylphosphonate intemucleotide linkage modifications; modified nucleoside bases or modified sugars can be used in oligonucleotide or polynucleotide synthesis; and oligonucleotides or polynucleotides can be labelled with a fluorescent moiety (e.g., fluorescein or rhodamine or a fluorescence resonance energy transfer or FRET pair of chromophore labels) or other label (e.g., biotin or an isotope). Modified nucleic acids, particularly modified RNAs, are disclosed in U.S. Pat. No. 9,464,124, incorporated by reference in its entirety herein.

[0078] As used herein, the phrase “sequence identity” refers to the percent similarity of two polynucleotides or polypeptides. A polynucleotide or polypeptide has a certain percent “sequence identity” to another polynucleotide or polypeptide, meaning that, when aligned, that percentage of bases or amino acids are the same, and in the same relative position, when comparing the two sequences. Sequence similarity can be determined in a number of different manners. To determine sequence identity, sequences can be aligned using the methods and computer programs, including BLAST, available at ncbi[dot]nlm[dot]nih[dot]gov / BLAST.17MF-364940674Docket No.: 165362002340See, e.g., Altschul et al. Mol. Biol. 1990, 215:403-410. Another alignment algorithm is FASTA, available in the Genetics Computing Group (GCG) package, from Madison, Wis., USA, a wholly owned subsidiary of Oxford Molecular Group, Inc. Other techniques for alignment are described in Methods in Enzymology, vol. 266: Computer Methods for Macromolecular Sequence Analysis (1996), ed. Doolittle, Academic Press, Inc., a division of Harcourt Brace & Co., San Diego, Calif., USA. Of particular interest are alignment programs that permit gaps in the sequence. The Smith-Waterman is one type of algorithm that permits gaps in sequence alignments. See Meth. Mol. Biol., 70: 173-187 (1997). Also, the GAP program using the Needleman and Wunsch alignment method can be utilized to align sequences. See Mol. Biol., 48: 443-453 (1970).

[0079] As used herein, “or” is inclusive and not exclusive, unless expressly indicated otherwise or indicated otherwise by context. Therefore, herein, “A or B” means “A, B, or both,” unless expressly indicated otherwise or indicated otherwise by context. Moreover, “and” is both joint and several, unless expressly indicated otherwise or indicated otherwise by context. Therefore, herein, “A and B” means “A and B, jointly or severally,” unless expressly indicated otherwise or indicated otherwise by context.

[0080] Unless otherwise stated, nucleic acid sequences in the text of this specification are given, when read from left to right, in the 5' to 3' direction. Nucleic acid sequences may be provided as DNA or as RNA, as specified; disclosure of one necessarily defines the other, as well as necessarily defines the exact complements, as is known to one of ordinary skill in the art.

[0081] As used herein, “in tandem” is used to mean that there is no space between multiple copies of an element. For example, copies of an enhancer element that are “positioned in tandem” or that are “in tandem” are inserted adjacently and without space between each copy (i.e., the copies are separated by 0 base pairs).

[0082] Where a term is provided in the singular, the inventors also contemplate aspects of the invention described by the plural of that term.

[0083] To the extent to which any of the preceding definitions is inconsistent with definitions provided in any patent or non-patent reference incorporated herein by reference, any patent or non-patent reference cited herein, or in any patent or non-patent reference found elsewhere, it is understood that the preceding definition will be used herein.18MF-364940674Docket No.: 165362002340A. Computer implemented methods for identifying an enhancer element in a plant

[0084] Provided herein are computer implemented methods for identifying an enhancer element in a plant. FIG.9 depicts a non-limiting example of a computer implemented method for identifying an enhancer element in a plant. The computer implemented method, 900, comprises, by one or more computing devices comprising one or more processors and memory (a) receiving a first candidate sequence from the memory; (b) generating a plurality of candidate sequences from the first candidate sequence using single base pair saturation mutagenesis; (c) for each promoter sequence in a plurality of promoter sequences, generating a plurality of input sequences, each comprising a candidate sequence from the plurality of candidate sequences and the promoter sequence; (d) inputting the plurality of input sequences into two or more machine learning models trained to predict a measurement of chromatin accessibility in one or more different plant tissues from a sequence; (e) receiving, from the two or more machine learning models, a predicted chromatin accessibility measurement for each of the plurality of input sequences; (f) synthesizing a composite measurement for each candidate sequence related to the predicted chromatin accessibility measurement for each input sequence in the plurality of input sequences comprising the candidate sequence from each of the two or more machine learning models; (g) selecting a top performing sequence from the plurality of candidate sequences based on the composite measurement; and (h) if the top performing sequence matches the first candidate sequence, identifying the top performing sequence as the enhancer element in the plant, and if the top performing sequence does not match the first candidate sequence, updating the first candidate sequence to be top performing sequence and repeating steps (b) - (g).I. Candidate sequences and mutagenesis

[0085] At block 902, a first candidate sequence is received from the memory of the one or more computing devices. The first candidate sequence represents a first candidate sequence to be tested as a potential enhancer for a plant as described herein. Throughout the iterative steps of the methods, the first candidate sequence is updated according to the predicted performance of the first candidate sequence. The performance of the first candidate sequence relates to the predicated chromatin accessibility measurement for the first candidate sequence when linked to a plurality of promoter sequences as described herein and in different plant tissues.19MF-364940674Docket No.: 165362002340

[0086] The first candidate sequence comprises a polynucleotide sequence. The polynucleotide sequence may comprise adenine, guanine, cytosine, uracil and thymine nucleotides. In some embodiments, the polynucleotide sequence comprises between about 5 and about 105 nucleotides. In some embodiments, the polynucleotide sequence comprises about 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100, or 105 nucleotides. In some embodiments, the polynucleotide sequence comprises between about 10 and about 60 nucleotides. In some embodiments, the polynucleotide sequence comprises about 10, 20, 30, 40, 50, or 60 nucleotides. In some embodiments, the polynucleotide sequence comprises between 10 and 60, 10 and 50, 10 and 40, 10 and 30, 10 and 20 nucleotides.

[0087] In some embodiments, the polynucleotide sequence may be longer. Longer sequences may increase computational complexity based on alignments of the sequences to available chromatin accessibility models.

[0088] In some embodiments, the first candidate sequence comprises a randomly generated polynucleotide sequence comprising a uniform distribution of adenine, guanine, cytosine, and thymine nucleotides. In some embodiments, the methods comprise randomly generating the polynucleotide sequence. In some embodiments, the method comprises randomly generating the polynucleotide sequence by one or more processors and storing the randomly generated polynucleotide sequence in the memory to be received in the method.

[0089] At block 904, a plurality of candidate sequences are generated from the first candidate sequence received at block 902. Block 904 comprises generating a plurality of candidate sequences from the first candidate sequence using single base pair saturation mutagenesis. In some embodiments, the plurality of candidate sequences are generated by the one or more processors. Single base pair saturation mutagenesis may comprise generating a candidate sequence corresponding to every possible combination of single base pair mutations that are possible started from the first candidate sequence. In some embodiments, single base pair saturation mutagenesis comprises separately substituting the nucleotide at each position in the first candidate sequence with an adenine, a thymine, a guanine, and a cytosine nucleotide.

[0090] In some embodiments, the first candidate sequence has a length of N nucleotides. The length of N nucleotides may correspond to the number of nucleotides in the polynucleotide sequence of the first candidate sequence. In some embodiments, N is between20MF-364940674Docket No.: 165362002340about 10 and about 60 nucleotides. In some embodiments, N is between about 10, 20, 30, 40, 50, or 6O.h In some embodiments, N is between 10 and 60, 10 and 50, 10 and 40, 10 and 30, 10 and 20. In some embodiments, the polynucleotide sequence may be longer. Longer sequences may increase computational complexity based on alignments of the sequences with available chromatin accessibility models.

[0091] In some embodiments, the plurality of candidate sequences comprises 3 x IV + 1 candidate sequences. In some embodiments, the plurality of candidate sequences comprises up to 3 x IV + 1 candidate sequences, wherein N is any of the N values as described herein. In some embodiments, the plurality of candidate sequences comprises between 1 and3 x IV + 1 candidate sequences, wherein N is any of the N values as described herein.

[0092] In some embodiments, the first candidate sequence comprises a polynucleotide sequence comprising a known enhancer motif flanked by one or more variable nucleotides. The methods described herein for identifying an enhancer element in a plant wherein the first candidate sequence comprises a known enhancer motif can be used to identify improved enhancer elements compared to those previously described. In some embodiments, the improvement of the enhancer element is the ability of the enhancer element to up regulate a gene when operably linked to a plurality of promoters and / or in multiple tissues. The methods comprise a first candidate sequence with added nucleotides on either or both sides (e.g., flanking regions) of the known enhancer motif.

[0093] In some embodiments, the one or more variable nucleotides are located 3’ to the known enhancer motif, 5’ to the known enhancer motif or both 3’ and 5’ to the known enhancer motif. In some embodiments, the one or more variable nucleotides are randomly generated from a uniform distribution of adenine, guanine, cytosine, and thymine nucleotides. In some embodiments, the methods comprise randomly generating the first candidate sequence comprising the known enhancer motif. In some embodiments, the method comprises randomly adding variable nucleotides to either or both sides of the known enhancer motif by one or more processors and storing the first candidate sequence in the memory to be received in the method.

[0094] In some embodiments, the known enhancer motif comprises a known enhancer motif in the plant. In some embodiments, the known enhancer motif comprises a sequence element disclosed in the literature to increase gene expression in the plant. In some embodiments, the known enhancer motif comprises an OCS element, a G-Box motif. In some21MF-364940674Docket No.: 165362002340embodiments, the first candidate sequence comprises the known enhancer motif flanked by one or more variable nucleotides. In some embodiments, the first candidate sequence comprises the known enhancer motif flanked by 1, 2, 3, 4, or 5 variable nucleotides. As explained herein, during single base pair saturation mutagenesis the known enhancer motif may be held constant and the variable nucleotides flanking the know enhancer motif are separately substituted at each position with an adenine, a thymine, a guanine, and a cytosine nucleotide. The first candidate sequence comprising the known enhancer motif and the variable flanking nucleotides are used for generating input sequences as described herein. In some embodiments, the first candidate sequence comprises the known enhancer motif repeated 2 or more times. The known enhancer motifs may be repeated 2, 3, 4, 5, or more times. In some embodiments, the orientation of the repeated known enhancer motifs may be different, for example one of the known enhancer motifs in the repetition may be the reverse compliment of the known enhancer motif. It is appreciated that known enhancer motifs often comprise short palindromic sequences that may fold back on each other when inserted into a plant, decreasing the efficiency of the element as an enhancer. Thus, it may be important to separate the known enhancer motif within a candidate enhancer with nucleotides. The methods disclosed herein can be used to identify the nucleotides predicted to prevent the known enhancer motifs from folding in planta. In some embodiments, the known enhancer motifs are separated by variable nucleotides. In some embodiments, the known enhancer motifs are separated by 1, 2, 3, 4, or 5 variable nucleotides.

[0095] In some embodiments, the known enhancer motif is a k-mer, wherein k represents the number of nucleotides in the sequence or motif. In some embodiments, the first candidate sequence comprises the k-mer motif flanked by 1, 2, 3, 4, or 5 variable nucleotides. In some embodiments, the first candidate sequence comprises the k-mer repeated 2 or more times. The k-mer may be repeated 2, 3, 4, 5, or more times. In some embodiments, the copies of the k-mer are separated by variable nucleotides. In some embodiments, the k-mers are separated by 1, 2, 3, 4, or 5 variable nucleotides.

[0096] In some embodiments, the k-mer comprises a polynucleotide sequence of 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 nucleotides. In some embodiments, the k-mer comprises a polynucleotide sequence between 2 and 12, 2 and 11, 2 and 10, 2 and 9, 2 and 8, 2 and 7, 2 and 6, 2 and 5, 2 and 4, or 2 and 3 nucleotides. In some embodiments, the k-mer comprises a polynucleotide sequence between 2 and 12, 3 and 12, 4 and 12, 5 and 12, 5 and 12, 7 and 12,22MF-364940674Docket No.: 1653620023408 and 12, 9 and 12, 10 and 12, or 11 and 12 nucleotides. In some embodiments, the k-mer comprises a polynucleotide sequence less than about 12 nucleotides.

[0097] In some embodiments, the methods comprise identifying a k-mer. In some embodiments, the k-mer is identified by identifying a plurality of promoters associated with highly expressed genes in the plant; and identifying a polynucleotide sequence enriched in sequences for the plurality of promoters. Identifying a plurality of promoters associated with highly expressed genes may comprise identifying accessible chromatin regions in one or more cell types from the plant using single nucleus ATAC-seq (snATAC-seq). In some embodiments, the accessible chromatin regions are accessible in about 75% of cell types in the plant or cell types in the plant in the snATAC-seq data. The accessible chromatin regions may be defined by a snATAC-seq peaks. Highly expressed genes may be determined using RNA-seq data for the plant. Identifying highly expressed genes may comprise ranking genes according to RNA-seq expression into buckets, such as high, medium, and low expression buckets.

[0098] Identifying a k-mer comprises identifying a polynucleotide sequence enriched in sequences for the plurality of promoters. Identifying polynucleotide sequences enriched in the sequences for the plurality of promoters may comprise performing a motif enrichment analysis using the sequences of the plurality of promoters as input. The sequences for the plurality of promoters may comprise about 500 snATAC-seq peaks. In some embodiments, the top enriched sequences are identified as k-mers. In some embodiments, the candidate sequence comprises one or more k-mers, such as k-mers identified according to the methods describe herein. Including a known enhancer motif (e.g. k-mer) the candidate sequence may increase efficiency of the methods described herein because holding some nucleotides in a candidate sequence consistent during single base pair mutagenesis may reduce the number of inputs for the machine learning model downstream. Including a known enhancer motif in the candidate sequence reduces the computational power needed to identify an enhancer element, technically improving the methods in a biologically informed manner.

[0099] In some embodiments, the methods comprise generating a plurality of candidate sequences from the first candidate sequence using single base pair saturation mutagenesis, wherein single base pair saturation mutagenesis is performed on the variable nucleotides (e.g. variable nucleotides flanking known enhancer motifs and / or variable nucleotides between copies of known enhancer motifs). In some embodiments, the known enhancer motif is held constant during the single base pair saturation mutagenesis. In some embodiments, single 23MF-364940674Docket No.: 165362002340base pair saturation mutagenesis comprises separately substituting the nucleotide at each position of the variable nucleotides in the first candidate sequence with an adenine, a thymine, a guanine, and a cytosine nucleotide.

[0100] In some embodiments, the plurality of candidate sequences comprises 3 x M + 1 candidate sequences, wherein M is the number of variable nucleotides in the first candidate sequence. In some embodiments, the number of variable nucleotides comprises the sum of the variable nucleotides flanking a known enhancer motif and the variable nucleotides between copies of the known enhancer motif. In some embodiments, the number of variable nucleotides in the first candidate sequences is the number of nucleotides that were held constant during single base pair mutagenesis. In some embodiments, the constant nucleotides are the nucleotides of a known enhancer motif (e.g. k-mer). In some embodiments, M is a value of variable nucleotides and is the length of the known enhancer motif subtracted from the length of the first candidate sequence as described herein. In some embodiments, the plurality of candidate sequences comprises up to 3 x M + 1 candidate sequences, wherein M is any of the M values as described herein. In some embodiments, the plurality of candidate sequences comprises between 1 and 3 x M + 1 candidate sequences, wherein M is any of the M values as described herein.IL Generating input sequences

[0101] At block 906, input sequences are generated using the candidate sequences generated at block 904. Block 906 comprises generating a plurality of input sequences, each comprising a candidate sequence from the plurality of candidate sequences and a promoter. The method comprises generating input sequences for each promoter sequence in a plurality of promoter sequences. As generated according to the methods provided herein the plurality of input sequences may comprise each candidate sequence from the plurality of candidate sequences and each promoter sequence in the plurality of promoter sequences. The input sequences may represent the candidate sequences operably linked to a wide range of promoters. By generating input sequences with the candidate sequences and a wide range of promoters the methods described herein can be used identify enhancer elements that are likely to upregulate genes when operably linked to any promoter for the plant. The methods can be used to identify enhancers that can be used as tools for upregulating one or more genes in the plant without needing to individually test promoters.24MF-364940674Docket No.: 165362002340

[0102] In some embodiments, the plurality of input sequences comprises the number of candidate motifs multiplied by the number of promoters in the plurality of promoters. In some embodiments, the plurality of input sequences comprises greater than 200,000 sequences. In some embodiments, the plurality of input sequences comprises between about 50,000 and about 400,000 sequences. In some embodiments, the plurality of input sequences comprises between about 100,000 and about 400,000, between about 150,000 and about 400,000, between about 200,000 and about 400,000, between about 250,000 and about 400,000, between about 300,000 and about 400,000, or between about 350,000 and about 400,000 sequences. In some embodiments, the plurality of input sequences comprises between about 50,000 and about 350,000, between about 50,000 and about 300,000, between about 50,000 and about 250,000, between about 50,000 and about 200,000, between about 50,000 and about 150,000, or between about 50,000 and about 100,000 sequences.

[0103] In some embodiments, a promoter in the plurality of promoters is a known promoter sequence element in the genome of the plant, such as the plant the method is being used to identify an enhancer element for. In some embodiments, the promoter is a known promoter in the genome of a plant of a different species. In some embodiments, the promoter is a region about 75bp to 150bp upstream of a transcription start site (TSS) in a plant genome. In some embodiments, a promoter sequence is a region about 75 bp to 150 bp upstream of a TSS in the plant genome and wherein a promoter has been predicted with use of a functional genomic technology. The functional genomic technology may comprise cap analysis gene expression (CAGE) sequencing. The CAGE sequencing may have been performed to identify TSS and promoters in a plant tissue. In some embodiments, a promoter comprises a polynucleotide sequence. In some embodiments, the promoter comprises a polynucleotide sequence with a length of about 25, about 50, about 75, about 100, about 125, about 150, about 175, about 200, about 225, about 250, about 275, about 300, about 325 or 350 base pairs. In some embodiments, the promoter comprises a polynucleotide sequence with a length of between about 25 and 350, about 50 and 350, about 100 and 350, about 125 and 350, about 150 and 350, about 200 and 350, about 250 and 350, about 275 and 350, about 300 and 350 or about 325 and 350 base pairs. In some embodiments, the promoter comprises a polynucleotide sequence with a length of between about 25 and 350, about 25 and 325, about 25 and 300, about 25 and 275, about 25 and 250, about 25 and 225, about 25 and 200, about 25 and 175, about 25 and 150, about 25 and 125, about 25 and 100, about 25 and 75 or about 25 and 50 base pairs. In some embodiments, the promoter comprises a polynucleotide25MF-364940674Docket No.: 165362002340sequence with a length of between about 500 base pairs and 2000 base pairs. In some embodiments, the promoter comprises a polynucleotide sequence with a length of about 500, about 600, about, about 800, about 900, about 1000, about 1100, about 1200, about 1300, about 1400, about 1500, about 1600, about 1700, about 1800, about 1900, or about 2000 base pairs. In some embodiments, the promoter comprises a polynucleotide sequence with a length of between about 500 and 1750, about 500 and 1500, about 500 and 1250, about 500 and 1000, or about 500 and 750 base pairs. In some embodiments, the length of the promoter region may be based on epigenetic data collected for a gene. In some embodiments, the promoters in the plurality of promoters are of variable lengths and the input sequences are thus of variable lengths.

[0104] In some embodiments, the plurality of promoters comprises more than 5000 promoters. In some embodiments, the plurality of promoters comprises more than 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000, 2000, 3000, 4000, 5000, or 6000 promoters. In some embodiments, the plurality of promoters comprises between 100 and 6000, 200 and 6000, 300 and 6000, 400 and 6000, 500 and 6000, 600 and 6000, 700 and 6000, 800 and 6000, 900 and 6000, 1000 and 6000, 2000 and 6000, 3000 and 6000, 4000 and 6000, or 5000 and 6000 promoters. In some embodiments, the plurality of promoters comprises between 100 and 5000, 100 and 4000, 100 and 3000, 100 and 2000, 100 and 1000, 100 and 900, 100 and 800, 100 and 700, 100 and 600, 100 and 500, 100 and 400, 100 and 300 or 100 and 200 promoters.

[0105] In some embodiments, each input sequence comprises the candidate sequence within the promoter. In some embodiments, each input sequence comprises the candidate sequence at the 5’ end of the promoter. In some embodiments, each input sequence comprises the candidate sequence at the 3’ end of the promoter. In some embodiments, each input sequence comprises the candidate sequence once. In some embodiments, each input sequence comprises the candidate sequence repeated two times. In some embodiments, each input sequence comprises the candidate sequence repeated three times. In some embodiments, each input sequence comprises the candidate sequence repeated at least one time, at least two times or at least three times.

[0106] In some embodiments, generating an input sequence in the plurality of input sequences comprises adding one or more nucleotides to the 5’ and / or 3’ end of the candidate sequence. In some embodiments, the one or more nucleotides are random nucleotides.Adding random nucleotides to the 5’ and / or 3’ end of the candidate sequence may improve the methods disclosed herein because adding the random nucleotides may mimic insertion 26MF-364940674Docket No.: 165362002340scarring. In some embodiments, between about 1 and about 12 random nucleotides are added to the 5’ and / or 3’ end of the candidate sequence. In some embodiments, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 or 12 random nucleotides are added to the 5’ and / or 3’ end of the candidate sequence. In some embodiments, an addition of 12 random nucleotides on the 5’ and / or 3’ end of the candidate sequence may mimic a faulty inserted enhancer. It has been observed that insertion of a polynucleotide into a plant or plant part thereof according to the methods described herein often lead to imperfect repair of DNA around the insertion, causing random base substitution at insertion boundaries. This process can be modeled in silico by adding random nucleotides to the 5’ and / or 3’ end of the candidate sequences. In some embodiments, the same nucleotides are added at the 5’ and / or 3’ end of each candidate sequence in each input sequence. In some embodiments, different nucleotides are added at the 5’ and / or 3’ end of each candidate sequence.III. Using two or more machine learning models to synthesize a composite measurement for each candidate

[0107] At block 908, the input sequences from block 906 are input into two or more machine learning models trained to predict a measurement of chromatin accessibility from a sequence. The method at block 908 comprises inputting the plurality of input sequences, as described herein, into two or more machine learning models trained to predict a measurement of chromatin accessibility in one or more different plant tissues from a sequence. Inputting the plurality of input sequences in the two or more machine learning models allows for the methods to be used to identify enhancer elements that are likely to be accessible in multiple tissues. Chromatin accessibility likely differs between plant tissues because opening and closing chromatin is a gene regulatory mechanism that allows for gene expression differences between tissues. The methods described herein allow for identification of sequences that will likely be accessible in multiple tissues or are not under the control of tissue specific regulation. As described herein, enhancer elements that are accessible in multiple tissues can be used as tools to understand gene regulation without the need to identify and validate enhancers in different tissues independently.

[0108] In some embodiments, the measurement of chromatin accessibility is a predicted ATAC-seq coverage tracks. In some embodiments, ATAC-seq coverage tracks are the direct functional genomics assay readouts of chromatin profiling ATAC-seq assays. An ATAC-seq coverage track may comprise ATAC-seq reads mapped to a genomic location. ATAC-seq27MF-364940674Docket No.: 165362002340coverage tracks may be used to determine ATAC-seq peak shape and or peak height as a measurement of ATAC-seq signal. In some embodiments, the measurement of chromatin accessibility is a predicted ATAC-seq peak height. ATAC-seq peak height may relate to overall read count in a predicted ATAC-seq coverage track. In some embodiments, the measurement of chromatin accessibility is a predicted ATAC-seq peak shape. ATAC-seq peak shape may relate to genomic locations and a shape of an ATAC-seq peak in a predicted signal track window. In some embodiments, the measurement of chromatin accessibility relates to ATAC-seq peak height and ATAC-seq peak shape. As described herein, ATAC-seq peaks may relate to chromatin accessibility because an ATAC-seq derived peak corresponds to ATAC-seq read coverage and thus accessible chromatin.

[0109] The two or more machine learning models have been trained to predict a measurement of chromatin accessibility in one or more different plant tissues from a sequence. Model training and architecture are described herein. In some embodiments, the two or more machine learning models have been trained to predict chromatin accessibility in different plant tissues. In some embodiments, the plant tissues are selected from a group consisting of bud, cotyledon, flower, flower bud, hypocotyl, leaf, pod, and root. According to the methods described herein, one machine learning model may have been trained to predict chromatin accessibility in one tissue, (e.g., leaf) and another machine learning model may have been trained to predict chromatin accessibility on a different tissue (e.g., flower).

[0110] In some embodiments, the two or more machine learning models described herein comprise multi-task models. In some embodiments, the two or more machine learning models are configured as an ensemble model. In some embodiments, the ensemble model comprises model nodes trained independently to predict chromatin accessibility in a plant tissue such as cotyledon, flower, flower bud, hypocotyl, leaf, pod, or root. In some embodiments, the ensemble model the predictions from each of the nodes are used to predict a measurement of chromatin accessibility in one or more different plant tissues.

[0111] In some embodiments, a machine learning model has been trained to predict chromatin accessibility in, cotyledon, flower, flower bud, hypocotyl, leaf, pod, and root independently. In some embodiments, the methods comprise inputting the plurality of input sequences into each of the machine learning models trained to predict a measurement of chromatin accessibility in each of the plant tissues described herein. By inputting the plurality of input sequences into the 8 machine learning models, the methods described herein can be used to identify an enhancer with predicted pan-tissue activity for the plant.28MF-364940674Docket No.: 165362002340

[0112] In some embodiments, the one or more plant tissues comprise plant tissues from a dicot. In some embodiments, the one or more plant tissues comprise plant tissues from a dicot and the plant the method is being used to identify enhancer for is a dicot. In some embodiments, the one or more plant tissues comprise plant tissues from the same species as the plant that the method is being used to identify an enhancer for. Due to conservation of chromatin accessibility between plants in the same clade (e.g. dicots), machine learning modes trained using data from one dicot (e.g. soybean) may perform well for predicting chromatin accessibility in other dicots as well.

[0113] In some embodiments, the machine learning models have been trained with chromatin accessibility data, as described herein, from plant tissues from the same species as the plant. In some embodiments, the methods comprise use of machine learning models trained with chromatin accessibility data, as described herein, from collected from plant tissues from the same species as the plant the method is used to identify enhancers for.

[0114] In some embodiments, the two or more machine learning models have each been trained to predict a measurement of chromatin accessibility in a plant tissue, such as plant tissue described herein, based on chromatin accessibility data collected from the plant tissue. In some embodiments, the chromatin accessibility data collected from the plant tissue is ATAC-Seq data. In some embodiments, the ATAC-seq is bulk ATAC-seq, single cell ATAC-seq or single nuclei ATAC-seq. In some embodiments, additional functional genomic data collected from a plant tissue may be used to improve performance of the machine learning models described herein. In some embodiments, the additional functional genomic data comprises CAGE-seq, STRIPE-seq, RNA-seq, or STARR-seq.

[0115] In some embodiments, the methods comprise collected ATAC-seq data from two or more plant tissues and training two or more machine learning models to predict a measurement of chromatin accessibility in plant tissues according to the methods described herein. In some embodiments, collected ATAC-seq data from a plant tissue comprises harvesting DNA from cells of the plant tissue, incubating the DNA with a transposase enzyme that cuts open chromatin regions, and performing whole genome sequencing on the resulting DNA fragments. In some embodiments, the transposase enzyme is Tn5.

[0116] In some embodiments, the plant is soybean. In some embodiments, the two or more machine learning models are trained on chromatin accessibility data collected from soybean tissues. In some embodiments, the soybean tissues are selected from a group consisting of29MF-364940674Docket No.: 165362002340bud, cotyledon, flower, flower bud, hypocotyl, leaf, pod, and root. In some embodiments, the chromatin accessibility data comprises soybean data collected from Williams 82 soybeans. In some embodiments, the chromatin accessibility data comprises Wm82v4 ATAC-seq data. Wm82v4 ATAC-seq data is ATAC-seq data collected from one or more tissues from William 82 soybeans and mapped to the Wm82v4 reference genome.

[0117] In some embodiments, the machine learning model of the two or more machine learning models comprises a machine learning model as described herein. In some embodiments, the machine learning models comprises a neural network. In some embodiments, the neural network is a convolutional neural network (CNN). In some embodiments, the CNN is a dilated CNN.

[0118] At block 910, a predicted chromatin accessibility measurement is received from the two or more machine learning models at block 908. Block 910 comprises receiving, from the two or more machine learning models, a predicted chromatin accessibility measurement for each of the plurality of input sequences. The predicted chromatin accessibility measurements relate to the predicted chromatin accessibility of the input sequence comprising a candidate sequence of the plurality of candidate sequences and a promoter of the plurality of promoters for each of the plant tissues that the two or more machine learning models have been trained on. For example, if the two or more machine learning models have been trained on stem and bud independently, the method may comprise receiving a predicted chromatin accessibility measurement for each of the plurality of input sequences in stem and bud. Receiving predicted chromatin accessibility measurements for the plurality of input sequences rather than for the candidate sequences improves upon methods know in the art because the method can be used to identify candidate sequences that are not context dependent (e.g., dependent on a specific promoter).

[0119] At block 912, a composite measurement is synthesized for each candidate sequence using the predicted chromatin accessibility measurements from block 910. Block 912 comprises synthesizing a composite measurement for each candidate sequence related to the predicted chromatin accessibility measurement for each input sequence in the plurality of input sequences comprising the candidate sequence from each of the two or more machine learning models. In some embodiments, the composite measurement is based on the chromatin accessibility of the candidate sequence when the candidate sequence is linked to the plurality of promoters in two or more tissues.30MF-364940674Docket No.: 165362002340

[0120] Because each input sequence comprises a candidate sequence, predicted chromatin accessibility measurements outputted from the two or more machine learning models can be grouped according to the candidate sequence that was used to generate the input sequences and a composite measurement can be created. In some embodiments, the methods comprise synthesizing a composite measurement for a candidate sequence relating to the predicted chromatin accessibility measurement for the candidate sequence in the context of each of the plurality of promoters and across two or more tissues. In some embodiments, synthesizing a composite measurement for each candidate sequence provides a technical advantage of reducing the predicted chromatin accessibility measurements that must be stored in the memory of the computing device for each iteration of the methods. The composite measurements can be used according to the methods described herein to select a top performing candidate sequence for each iteration of the methods.

[0121] In some embodiments, the composite measurement is an average of the predicted chromatin accessibility measurements related to each input comprising the candidate sequence. In some embodiments, synthesizing the composite measurement comprises averaging the predicted chromatin accessibility measurements for each input comprising the candidate sequence from one of the two or more machine learning models and averaging the predicted chromatin accessibility measurements for each candidate sequence across the two or more machine learning models. In some embodiments, a composite measurement is synthesized for each candidate sequence by first averaging the predicted chromatin accessibility measurements for all the input sequences comprising the candidate sequence output from one of the two or more machine learning models and then averaging the composite measurement for the candidate sequences across the two or more machine learning modes.

[0122] In some embodiments, a composite measurement is related to predicted chromatin accessibility measurements for two or more tissues. In some embodiments, a composite measurement is related to predicted chromatin accessibility measurements for three or more tissues. In some embodiments, a composite measurement is related to predicted chromatin accessibility measurements for four or more tissues. In some embodiments, a composite measurement is related to predicted chromatin accessibility measurements for five or more tissues. In some embodiments, a composite measurement is related to predicted chromatin accessibility measurements for six or more tissues. In some embodiments, a composite measurement is related to predicted chromatin accessibility measurements for seven or more31MF-364940674Docket No.: 165362002340tissues. In some embodiments, a composite measurement is related to predicted chromatin accessibility measurements for eight or more tissues. In some embodiments, a composite measurement is related to predicted chromatin accessibility measurements for bud, cotyledon, flower, flower bud, hypocotyl, leaf, pod, and root.

[0123] In some embodiments, the composite measurement is a weighted average of the predicted chromatin accessibility measurements related to each input comprising the candidate sequence. In some embodiments, generating a weighted average may comprise weighting the predicted chromatin accessibility measurement according to metrics of the machine learning model. The metrics of the machine learning model may comprise data quality or model confidence for the machine learning model. In some embodiments, generating a weighted average may comprise weighting the predicted chromatin accessibility measurements based on one or more target tissues.

[0124] In some embodiments, the composite measurement is the median of the predicted chromatin accessibility measurements related to each input comprising the candidate sequence. In some embodiments, the composite measurement is a sum of the predicted chromatin accessibility measurements related to each input comprising the candidate sequence.

[0125] In some embodiments, synthesizing the composite measurement comprises incorporating data about a sequence. In some embodiments, synthesizing the composite measurement comprises incorporating data about a sequence generated from other sequence-to-function models trained on data modalities. In some embodiments, the other data modalities comprise histone marks, sequence conservation, transcription factor binding, and gene expression. Accounting for additional data related to the properties of the candidate sequence increases the flexibility in methods and allows for users to steer the sequence evolution toward a desired sequence in a biologically informed manner.

[0126] In some embodiments, the composite measurement is synthesized using the predicted chromatin accessibility generated from one or more machine learning models of an ensemble machine learning model as described herein.IV. Selecting top preforming and iteration criteria

[0127] At block 914, the composite metrics synthesized at block 912 are used for selecting a top performing sequence. Block 914 comprises selecting a top performing sequence from the plurality of candidate sequences based on the composite measurements and if the top 32MF-364940674Docket No.: 165362002340performing sequence matches the first candidate sequence, according to block 916, identifying the top performing sequence as the enhancer element in the plant, and if the top performing sequence does not match the first candidate sequence, according to block 916, updating the first candidate sequence to be top performing sequence at block 920 and repeating all of the steps of the block 904 - block 916.

[0128] In some embodiments, selecting a top performing sequence from the plurality of candidate sequences based on the composite measurements comprises greedy selection. Greedy selection, similar to greedy algorithms for machine learning training, rely on selection of a top performing candidate according to a use identified criteria and use the top performing candidate iteratively to identify a local maximum. Using greedy selection increases model efficiency by decreasing the data needed to be analyzed in a subsequence round. In some embodiments, greedy selection is implemented in the methods described herein by selecting a candidate sequence with the highest composite measurement. In some embodiments, greedy selection can be performed by choosing two or more of the candidate sequences with the highest composite measurements for a subsequent iteration of the method steps.

[0129] By updating the first candidate sequence if the top performing candidate does not match the first candidate sequence for further iterations of the method steps and identifying the top performing candidate sequence as the enhancer element in the plant, the method is designed to converge on a top performing candidate sequence. The method converges when there is not further mutation that can be made to the candidate sequence so that a mutated version will produce a higher composite measurement than the sequence with the highest composite measurement in the previous round.

[0130] In some embodiments, repeating all of the steps of the methods after receiving a first candidate sequence increases the probability the top performing sequence matches the first candidate sequence in a subsequent iteration of the method steps. In some embodiments, the method steps are repeated about 10 times. In some embodiments, the method steps are repeated between about 1 and 30, 1 and 25, 1 and 20, 1 and 15, or 1 and 10 times. In some embodiments, the method steps are repeated between about 5 and 30, 10 and 30, 15 and 30, 20 and 30 or 25 and 30 times. It is appreciated that the number of variable nucleotide positions in the candidate sequence (N and M as described herein) relates to the number of iterations that will likely be needed to reach convergence of the method. As M and N increase, the number of iterations may also increase.33MF-364940674Docket No.: 165362002340

[0131] The methods described herein comprise identifying the top performing sequence as the enhancer element in the plant, and if the top performing sequence does not match the first candidate sequence. In some embodiments, the identified enhancer comprises a polynucleotide sequence predicted to up regulate gene expression in a plant or part thereof. In some embodiments, the identified enhancer comprises a polynucleotide sequence predicted to upregulate gene expression in two or more tissues of a plant. In some embodiments, the identified enhancer comprises a polynucleotide sequence predicted to upregulate gene expression of a gene following introduction of the identified enhancer into the plant genome upstream of the gene, according to the methods described herein. In some embodiments, the identified enhancer comprises a known enhancer motif, such as a known enhancer motif as described herein. In some embodiments, the identified enhancer is experimentally validated using the methods described herein.

[0132] In certain embodiments, the identified enhancer may be generated and introduced to one or more crop seeds (e.g., com crop seeds, soybean crop seeds, rice crop seeds, wheat crop seeds, tomato crop seeds, citrus fruit crop seeds, cacao crop seeds, potato crop seeds, cotton crop seeds, cabbage crop seeds, mushroom crop seeds, canola crop seeds, papaya crop seeds, and so forth) to germinate one or more crops (e.g., corn crop, soybean crop, rice crop, wheat crop, tomato crop, citrus fruit crop, cacao crop, potato crop, cotton crop, cabbage crop, mushroom crop, canola crop, papaya crop, and so forth) in accordance with the identified enhancer. In some embodiments, enhancer has been identified in the plant species for which the identified enhancer is generated for and introduced into. In some aspects, an identified enhancer can be generated and introduced into a seed of a different species.B. Methods of validating an enhancer element in a plant or part thereof

[0133] In some aspects, the methods provided herein comprise experimental validation of enhancer elements. Experimental validation may comprise validating the enhancer element is capable of increasing expression of one or more genes across tissue contexts. In some embodiments, validating the enhancer element comprises a reporter assay. The validation may comprise performing a reporter assay, as described herein, to measure expression of a reporter gene in the two or more plant tissues, wherein expression of the reporter gene in the two or more plant tissues and with different promoters (e.g. two or more endogenous promoters) indicates validation of the enhancer element. In some embodiments, validating the34MF-364940674Docket No.: 165362002340enhancer comprises two or more reporter assays. In some embodiments, the two or more reporter assays are reporter assays as described herein, using different promoters.

[0134] The methods described herein include ways of validating enhancer element activity. As described herein, enhancer elements may be operably linked to a promoter or operably linked to a promoter and a gene, where the enhancer element participates in the regulation and expression of the gene along with the promoter. Enhancer elements may increase the expression or transcription of genes. It is well-understood in the art that reporter assays may be used to measure gene expression in biological systems, and to interrogate the impact of certain perturbations on gene expression. In many instances, the expression of a reporter gene can indicate the activity of an enhancer or other genetic modification. For example, the increased expression of a reporter gene in the presence of an enhancer element, compared to in the absence of an enhancer element, may indicate that the enhancer element increases gene expression.

[0135] In some aspects, the reporter assay comprises transforming a polynucleotide vector comprising, the enhancer element operably linked to an endogenous promoter (e.g. promoter sequence), a transcription start site, and a reporter gene into a plant or part thereof, wherein the reporter gene is a luciferase gene, and testing for expression of the luciferase gene. The reporter assay may comprise transformation of one or more polynucleotide vectors into a plant or part thereof and a comparison of the relative expression of the reporter genes in the one or more polynucleotide vectors.

[0136] In some aspects, the reporter assay comprises transforming a first polynucleotide vector comprising, the enhancer element operably linked to an endogenous promoter, a transcription start site, and a reporter gene into a first plant or part thereof from the plant; transforming a second polynucleotide vector comprising, a control sequence element operably linked to the endogenous promoter, the transcription start site, and the reporter gene into a second plant or part thereof from the plant; and measuring the expression of the reporter gene in the first plant or part thereof and in the second plant or part thereof, wherein increased expression of the reporter gene in the first plant or part thereof compared to the second plant or part thereof indicates that the enhancer element is effective for increasing transcription.35MF-364940674Docket No.: 165362002340I. Reporter Gene

[0137] In some embodiments, a reporter assay is used to experimentally validate an enhancer element. In some embodiments, the reporter assay uses at least one reporter gene. The reporter gene may include, but is not limited to, a luciferase gene, a fluorescence-related gene, or a gene for another detectable marker. The reporter gene may encode a luciferase protein, a fluorescent protein, or any protein that may produce a detectable readout. In some embodiments, the luciferase is a firefly luciferase, a Renilla luciferase, or a nanoluciferase. In some embodiments, the fluorescent protein is green fluorescent protein, red fluorescent protein, or yellow fluorescent protein. In some embodiments, the reporter assay uses two reporter genes. In some aspects, the vector contains two reporter genes. Each of the reporter genes may be a luciferase gene, a fluorescence gene, or another gene that produces a discernable readout that may be measured. In some aspects, the vector contains two reporter genes that are both luminescence genes. In some aspects, the two reporter genes are a firefly luciferase gene and a Renilla luciferase gene.IL Polynucleotide Vectors

[0138] Reporter assays may involve a polynucleotide vector, which may also be referred to as a plasmid, a construct, an expression construct, or an expression vector. In some embodiments, the polynucleotide vector contains one or more genes may be capable of being expressed under certain conditions. In some embodiments, the polynucleotide vector comprises one or more promoters selected through the method of an aspect of the present disclosure. In some embodiments, the vector comprises one or more enhancer elements, motifs, regulatory motifs, or k-mers selected through the method of an aspect of the present disclosure. In some embodiments, the one or more genes, one or more promoters, and one or more enhancer elements, motifs, regulatory motifs, or k-mers are operably linked within the polynucleotide vector.III. Transformation

[0139] In some instances, the reporter assay may involve transformation of a plant or part thereof with the vector. In some instances, the plant or part thereof is an isolated protoplast. The vector can be delivered to the plant, meristem cells of the plant, or plant protoplasts by particle mediated delivery, and any other direct method of delivery, such as but not limiting to, AgraZzacterm -mediated transformation, polyethylene glycol (PEG)-mediated36MF-364940674Docket No.: 165362002340transfection to protoplasts, whiskers mediated transformation, electroporation, particle bombardment, microprojectile-mediated delivery with a biolistic device, DNA injection, and / or by use of cell-penetrating peptides.

[0140] In some embodiments, the vector may be transformed into a plant or part thereof via at least one viral vector selected from the group consisting of adenoviruses, lentiviruses, adeno-associated viruses, retroviruses, gemini viruses, begomoviruses, tobamoviruses, potex viruses, comoviruses, wheat streak mosaic virus, barley stripe mosaic virus, bean yellow dwarf virus, bean pod mottle virus, cabbage leaf curl virus, beet curly top virus, tobacco yellow dwarf virus, tobacco rattle virus, potato virus X, and cowpea mosaic virus. In some embodiments, the vector may be transformed into a cell via at least one bacterial vector capable of transforming a plant cell and selected from the group consisting of Agrobacterium sp., Rhizobium sp., Sinorhizobium (Ensifer) sp., Mesorhizobium sp., Bradyrhizobium sp., Azobacter sp., and Phyllobacterium sp. In some embodiments, a viral vector may be delivered to a plant by transformation with Agrobacterium.

[0141] In some embodiments, a T-DNA vector may be used to deliver at least one expression vector to a plant or part thereof. In some embodiments, a T-DNA binary vector is used. In some embodiments, a T-DNA superbinary vector system is used. In other embodiments, a T-DNA ternary vector system is used. In some embodiments, the T-DNA system further comprises an additional virulence gene cluster. In some embodiments, the T-DNA system further comprises an accessory plasmid or virulence helper plasmid. In some embodiments, the T-DNA vector is an Agrobacterium vector.IV. Analysis of the reporter assay

[0142] In some aspects, the reporter gene or reporter genes contained within the polynucleotide vectors produce detectable readouts that may be used to analyze results of the reporter assay. In some aspects, the reporter assay readout is the amount of luminescence or fluorescence produced from a single luciferase or fluorescence reporter gene. In some aspects, the reporter assay readout is the amount of luminescence or fluorescence produced from two luciferase or fluorescence reporter genes. In some aspects, the reporter assay readout is the comparison between one reporter gene and a second reporter gene, wherein the first reporter gene is a luciferase gene, and the second reporter gene is a different luciferase gene. In some aspects, the reporter assay readout is the comparison between a firefly luciferase reporter gene and a Renilla luciferase reporter gene.37MF-364940674Docket No.: 165362002340V. Additional validation methods

[0143] In some aspects, validating an enhancer element may comprise inserting the enhancer element upstream of a gene in two or more plant tissues and measuring expression of the gene in the two or more plant tissues, wherein increased expression of the gene compared to endogenous expression of the gene in at least two or the two or more plant tissues indicates validation of the enhancer element. In some aspects, validating the enhancer may comprise measuring expression of the gene without inserting the enhancer. In some embodiments, an enhancer is validated if the expression of the gene is higher in two or more plant tissues wherein the enhancer has been inserted compared to exogenous expression of the gene without the inserted enhancer.

[0144] In some aspects, validating an enhancer elements according to the methods described herein may comprise the use of gene editing techniques to insert the enhancer into the genome or a plant or part thereof and measuring the expression (mRNA or protein) of a gene. The gene may be selected based on knowledge that the up regulation of the gene in a plant or part thereof confers a favorable phenotype.

[0145] In some aspects, measuring expression comprises measuring expression of mRNA transcribed from a gene. In some aspects, measuring expression comprises qPCR or RNA sequencing. The RNA sequencing may be whole genome or targeted to measure expression of the gene. In some aspects, measuring expression comprises measuring expression of a protein product of the gene. Measuring protein expression may comprise a western blot, ELISA, mass spectrometry, immunohistochemistry or flow cytometry.C. Promoters and Transcription Start Sites

[0146] The methods disclosed herein may involve, but are not limited to, the examination of enhancer elements in multiple promoter sequences, which may also be referred to as promoters. Promoters are well-understood in the art as nucleotide sequences that participate in gene function, regulation, and expression. The interaction of a promoter with a gene to facilitate these processes is commonly referred to as the promoter being operably linked to a gene. In many instances, a promoter is a nucleotide sequence located near the 5’ end of the coding region of a gene (see Villao-Uzho et al., Plant Promoters: Their Identification, Characterization, and Role in Gene Regulation, 14 Genes 1226 (2023)). A promoter may be located upstream or downstream of a transcription start site in a gene. A transcription start site is well-known in the art as the location where transcription begins within a gene. In many 38MF-364940674Docket No.: 165362002340instances, the transcription start site is located at the 5’ end of the gene, preceding the coding region, and the promoter is located upstream of the transcription start site. In some instances, the promoter itself may include or overlap with the transcription start site of the gene. In some embodiments, the promoter is a region about 5 base pairs to about 1000 base pairs upstream of a transcription start site in a plant genome.

[0147] In some instances, a promoter or promoter sequence may be described as a regulatory region that includes binding sites for transcription factors and transcription factor complexes. Transcription factors and transcription factor complexes may interact with core transcription machinery at promoters to initiate transcription and participate in gene regulation and expression. In some embodiments, RNA polymerase binds to the promoter for the purposes of initiating gene transcription. In some embodiments, the promoter is a constitutive promoter, a conditional promoter, an inducible promoter, or a temporally or spatially specific promoter (e.g., a tissue specific promoter, a developmentally regulated promoter, or a cell cycle regulated promoter).

[0148] A promoter or promoter sequence may be exogenous or endogenous. As used herein, an exogenous promoter or promoter sequence is understood as one that is not found in a species, cell, or genome prior to human intervention. As used herein, an endogenous promoter or promoter sequence is understood as one that can be found in a species, cell, or genome prior to human intervention. In some embodiments, the promoter or promoter sequence is an endogenous promoter or promoter sequence. In some embodiments, the promoter is an endogenous plant promoter for a target gene. In some embodiments, the promoter is a minimal 35S promoter. In some embodiments, the promoter is endogenous to a soybean plant genome. In some embodiments, the promoter is the AlPlOa promoter, the AlPlOb promoter, the AML4 promoter, the CRN promoter, the HB-1 promoter, the RIC1 promoter, the RIC2 promoter, the RPF1 promoter, the JAG1 promoter, the JAG2 promoter, the KHZ1 promoter, the PP2C promoter, the TCP5-L promoter, the FT la promoter, the BS1 promoter, the BS2 promoter, the TFLlb promoter, the CYP76C-1 promoter, the CYP76C-2 promoter, the NF-YC4 promoter, the MYB promoter, or the Hl_5 promoter.D. Enhancer Elements

[0149] The methods disclosed herein can be used to identify enhancer elements for use in plants. Enhancer elements are understood to include, but are not limited to, polynucleotide sequences that increase the expression of one or more genes. In some instances, transcription39MF-364940674Docket No.: 165362002340factors may bind to enhancer elements, and those transcription factors may interact with endogenous transcription machinery to ultimately increase gene expression. Endogenous enhancer elements are well-known in the art to exist in genomes and may be present in a variety of genomic locations (see Jores et al., Identification of Plant Enhancers and Their Constituent Elements by STARR-seq in Tobacco Leaves, 32 Plant Cell 2120 (2020)). In some instances, specific patterns of gene expression may be driven by combinations of transcription factors that activate specific enhancer elements see Taskiran et al., Cell-type-directed design of synthetic enhancer, Nature, vol. 626, pp. 212-220 (2024)).

[0150] It is well understood that some enhancer elements contain short nucleotide sequences that are important for enhancer element activity. These short nucleotide sequences may also be referred to as motifs, regulatory motifs, or k-mers where k represents the number of nucleotides in the sequence or motif. These short nucleotide sequences may, in some instances, be palindromic. For example, the G-box is one type of regulatory motif that is known to modulate transcription of many plant genes (see Ishige et al., A G-box motif (GCCACGTGCC) tetramer confers high-level constitutive expression in dicot and monocot plants, 18 The Plant Journal 443-448 (1999)). The G-box is understood to contain the sequence set forth in Enhancer 99 (Table 2) and may be flanked by different nucleotides. The octopine synthase (OCS) element is another type of regulatory motif that is known to modulate transcription of plant genes (see Bouchez et al., The ocs-element is a component of the promoters of several T-DNA and plant viral genes, 20 The EMBO Journal 4197-4204 (1989)).

[0151] In the present invention, computer implemented methods can be used to identify novel enhancer elements that are not endogenous to genomes. An enhancer element may be exogenous or endogenous. As used herein, an exogenous enhancer element is understood as one that is not found in a species, cell, or genome prior to human intervention. As used herein, an endogenous enhancer element is understood as one that can be found in a species, cell, or genome prior to human intervention. In some embodiments, the enhancer element is an exogenous enhancer element. In some embodiments, the enhancer element is a randomly generated polynucleotide sequence comprising a uniform distribution of adenine, guanine, cytosine, and thymine nucleotides.

[0152] In some embodiments, the enhancer element is an endogenous enhancer element. In some embodiments, the enhancer element comprises, at least in part, one or more known regulatory motifs. The regulatory motifs may be repeated 1, 2, 3, or 4 times. Multiple copies 40MF-364940674Docket No.: 165362002340of a regulatory motif may be positioned in tandem, or may be separated by at least 1, at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, or at least 12 base pairs. In some embodiments, the enhancer element comprises, at least in part, one or more k-mers. A k-mer may be repeated 1, 2, 3, or 4 times. Multiple copies of a k-mer may be positioned in tandem, or may be separated by 1, 2, 3, 4, 5, or 6 base pairs.

[0153] In some embodiments, the enhancer element comprises a G-box element. In some embodiments, the enhancer element comprises a G-box element flanked by 1, 2, 3, 4, or 5 nucleotides on either side of the G-box element. In some embodiments, the G-box element may be present in about 1, about 2, or about 3 copies. Multiple copies of a G-box element may be positioned in tandem, or may be separated by at least 1, at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, or at least 12 base pairs. In some embodiments, the enhancer element comprises an OCS element. In some embodiments, the enhancer element comprises an OCS element flanked by 1, 2, 3, 4, or 5 nucleotides on either side of the OCS element. In some embodiments, the OCS element may be present in 1, 2, or 3 copies. Multiple copies of an OCS element may be positioned in tandem, or may be separated by at least 1, at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, or at least 12 base pairs. In some embodiments, multiple copies of an OCS element are separated by a spacer sequence. In some embodiments, the spacer sequence comprises at least 1, at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, or at least 12 base pairs. In some embodiments, the spacer sequence comprises about 10 base pairs. In some embodiments, the spacer sequence comprises the nucleotide sequence set forth in SEQ ID NO: 145.

[0154] In some embodiments, the enhancer element comprises an OCS element comprising the nucleotide sequence set forth in SEQ ID NO: 141. In some embodiments, the enhancer element comprises an OCS element comprising the nucleotide sequence set forth in SEQ ID NO: 141, and as used herein the OCS element is present in one copy. In some embodiments, the enhancer element comprises an OCS element comprising the nucleotide sequence set forth in SEQ ID NO: 141, and as used herein the OCS element is present in two copies, three copies, four copies, five copies, or six copies. In some embodiments, the element comprises three copies of an OCS element and comprises the nucleotide sequence set forth in SEQ ID NO: 142, wherein each copy of the OCS element is present in tandem. In some embodiments, the element comprises three copies of an OCS element and comprises the nucleotide41MF-364940674Docket No.: 165362002340sequence set forth in SEQ ID NO: 144. In some embodiments, the three copies of the OCS element are separated by a spacer sequence. In some embodiments, the spacer sequence is about 10 base pairs in length. In some embodiments, the spacer sequence comprises the nucleotide sequence set forth in SEQ ID NO: 145.

[0155] In some embodiments, an enhancer element may be present in a single copy, that is, the enhancer element may be repeated only one time. In some embodiments, the enhancer element may be present in about two copies, that is, the enhancer element may be repeated two times. In the methods described herein, the two copies of the enhancer element may be positioned in tandem, or may be separated by a spacer sequence comprising at least 1, at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, or at least 12 base pairs. In some embodiments, the two copies of the enhancer element are separated by a spacer sequence that is about 10 base pairs in length. In some embodiments, the enhancer element may be present in three copies (i.e., repeated three times), in four copies (i.e., repeated four times), in five copies (i.e., repeated five times), or in six copies (i.e., repeated six times).. In the methods described herein, each copy of the enhancer element may be positioned in tandem, or each copy may be separated by a spacer sequence comprising at least 1, at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, or at least 12 base pairs. In some embodiments, each copy of the enhancer element is separated by a spacer sequence that is about 10 base pairs in length. In some embodiments, the spacer sequence comprises the nucleotide sequence set forth in SEQ ID NO: 145. In some embodiments, the presence of a spacer sequence between copies of a regulatory motif or enhancer element improves the ability of the regulatory motif or enhancer element to increase gene expression. For example, the inclusion of a spacer sequence may remove the palindromic nature of a regulatory motif, which increases genomic stability.

[0156] The enhancer elements identified from the methods described herein may be examined and validated in a variety of locations, including but not limited to: within a coding genomic region, within a noncoding genomic region, within a 5’ untranslated region of a gene, within a 3’ untranslated region of a gene, within an exon, within an intron, upstream of a promoter, within a promoter, downstream of a promoter, upstream of a gene, within a gene, downstream of a gene, upstream of a transcription start site, or downstream of a transcription start site. In some embodiments, the enhancer element is located within about 1,000 base pairs of a gene. In some embodiments, the enhancer is located within about 1,000 base pairs42MF-364940674Docket No.: 165362002340of a transcription start site. In some embodiments, the enhancer element is located within a promoter. In some embodiments, the enhancer is located within any one of the promoters for AlPlOa, AlPlOb, AML4, CRN, HB-1, RIC1, RIC2, RPF1, JAG1, JAG2, KHZ1, PP2C, TCP5-L, FTla, BS1, BS2, TFLlb, CYP76C-1, CYP76C-2, and NF-YC4. In some embodiments, the enhancer element is located upstream of a promoter, where the promoter is any one of the promoters AlPlOa, AlPlOb, AML4, CRN, HB-1, RIC1, RIC2, RPF1, JAG1, JAG2, KHZ1, PP2C, TCP5-L, FTla, BS1, BS2, TFLlb, CYP76C-1, CYP76C-2, and NF-YC4. In some embodiments, the enhancer element is located upstream of a transcription start site. In some embodiments, the enhancer element is located between about 5 base pairs and 1000 base pairs upstream of a transcription start site. In some embodiments, the enhancer element is located between about 5 base pairs and about 100 base pairs, between about 100 base pairs and about 200 base pairs, between about 200 base pairs and about 300 base pairs, between about 300 base pairs and about 400 base pairs, between about 400 base pairs and about 500 base pairs, between about 500 base pairs and about 600 base pairs, between about 600 base pairs and about 700 base pairs, between about 700 and about 800 base pairs, between about 800 base pairs and about 900 base pairs, or between about 900 base pairs and about 1000 base pairs. In some embodiments, the enhancer element is located between about 5 base pairs and about 200 base pairs upstream of a transcription start site. In some embodiments, the synthetic enhancer element is located between about 25 base pairs and about 100 base pairs upstream of a transcription start site. In some embodiments, the enhancer element is located about 5, about 10, about 15, about 20, about 25, about 30, about 35, about 40, about 45, about 50, about 55, about 60, about 65, about 70, about 75, about 80, about 85, about 90, about 95, about 100, about 105, about 110, about 115, about 120, about 125, about 130, about 135, about 140, about 145, about 150, about 155, about 160, about 165, about 170, about 175, about 180, about 185, about 190, about 195, or about 200 base pairs upstream of a transcription start site. In some embodiments, the enhancer element is located about 25 base pairs, about 50 base pairs, about 75 base pairs, about 100 base pairs, about 125 base pairs, about 150 base pairs, or about 200 base pairs upstream of a transcription start site. In some embodiments, the synthetic enhancer element is located about 25 base pairs, about 50 base pairs, about 75 base pairs or about 100 base pairs upstream of a transcription start site.

[0157] An enhancer element may also be operably linked to a promoter or operably linked to a promoter and a gene, where the enhancer element participates in the regulation and43MF-364940674Docket No.: 165362002340expression of the gene along with the promoter. The enhancer elements identified from the methods described herein may be used to increase the expression of one or more genes. In some embodiments, the enhancer element increases gene expression by enhancing chromatin accessibility in the promoter region upstream of the gene. In some embodiments, the enhancer element facilitates interactions with transcription factors and other regulatory elements to enhance gene expression. In some embodiments, the enhancer element constitutively activates gene transcription, enabling gene expression under conditions where it would otherwise be inactive. In some embodiments, the enhancer element elevates gene expression to levels exceeding expression levels observed in the absence of the enhancer element. In some embodiments, the enhancer element increases gene expression about 5 -fold, about 10-fold, about 15-fold, about 20-fold, about 25-fold, about 30-fold, about 35-fold, about 40-fold, about 45-fold, about 50-fold, about 55-fold, about 60-fold, about 65-fold, about 70-fold, about 75-fold, or about 80-fold above levels of expression in the absence of the enhancer element.

[0158] Gene expression may be increased when the enhancer element is located in a variety of locations, including but not limited to: within coding regions of DNA, within noncoding regions of DNA, within exons, within introns, upstream of promoters, within promoters, downstream of promoters, upstream of genes, within genes, downstream of genes, upstream of transcription start sites, or downstream of transcription start sites. In some embodiments, the one or more genes may be plant genes. In some embodiments, the one or more genes may be endogenous to a soybean plant genome. In some embodiments, the one or more genes may include an AlPlOa gene, an AlPlOb gene, an AML4 gene, a CRN gene, a HB-1 gene, a RIC1 gene, a RIC2 gene, a RPF1 gene, a JAG1 gene, a JAG2 gene, a KHZ1 gene, a PP2C gene, a TCP5-L gene, a FTla gene, a BS1 gene, a BS2 gene, a TFLlb gene, a CYP76C-1 gene, a CYP76C-2 gene, or a NF-YC4 gene.E. Gene editing techniques

[0159] The methods and systems described herein may be used to identify enhancer elements that may be introduced into the genome of a plant or part thereof. In some embodiments, a modified plant or part thereof comprises an enhancer element that is operably linked to an endogenous promoter in the modified plant or part thereof and an endogenous gene in the modified plant or part thereof. In some embodiments, the enhancer element is introduced into the genome of a plant or part thereof via a gene editing approach. In some44MF-364940674Docket No.: 165362002340embodiments, a gene editing approach involves the introduction of gene editing machinery (e.g., a genome editing reagent) into a modified plant or part thereof. This may include the introduction of a Cas nuclease, a TALEN, or a zinc-finger nuclease. In some embodiments, the enhancer element is delivered to a plant or part thereof along with gene editing machinery. In some embodiments, the modified plant or part thereof is soybean (e.g., Glycine max), com (e.g., maize, Zea mays), or wheat (e.g., common wheat, spelt, durum, einkom, emmer, kamut, Triticum aestivum, Triticum spelta, Triticum durum, Triticum urartu, Triticum monococcum, Triticum turanicum, Triticum spp.).

[0160] In some embodiments, the enhancer element or the enhancer element and a genome editing reagent are delivered into a soybean plant or part thereof. In some embodiments, the enhancer element or the enhancer element and a genome editing reagent are delivered into a maize plant or part thereof. In some embodiments, the enhancer element and a genome editing reagent are delivered into a plant cell of a rootstock, of a grafted scion, of a seed (including mature seed and immature seed), of a plant cutting, of a plant cell culture, of a plant organ (e.g., intact nodal bud, shoot apex or shoot apical meristem, root apex or root apical meristem, lateral meristem, intercalary meristem, zygotic embryo, somatic embryo, ovule, pollen, microspore, anther, hypocotyl, cotyledon, leaf, petiole, stem, tuber, root, flowers, fruits, shoots, and explants). In some embodiments, the enhancer element and a genome editing reagent are introduced to one or more crop seeds (e.g., corn crop seeds, soybean crop seeds, rice crop seeds, wheat crop seeds, tomato crop seeds, citrus fruit crop seeds, cacao crop seeds, potato crop seeds, cotton crop seeds, cabbage crop seeds, mushroom crop seeds, canola crop seeds, papaya crop seeds, and so forth) to germinate one or more crops (e.g., com crop, soybean crop, rice crop, wheat crop, tomato crop, citrus fruit crop, cacao crop, potato crop, cotton crop, cabbage crop, mushroom crop, canola crop, papaya crop, and so forth).

[0161] In some embodiments, one or more one chemical, enzymatic, or physical agent, separately or in combination with a genome editing reagent, is provided / applied at a location in the plant or plant part other than the plant location, part, or tissue from which the plant cell or plant protoplast is obtained or isolated. In some embodiments, the genome editing reagent is applied to adjacent or distal cells or tissues and is transported (e.g., through the vascular system or by cell-to-cell movement) to the meristem from which plant cells or plant protoplasts are subsequently isolated. In some embodiments, a genome editing reagentcontaining composition is applied by soaking a seed or seed fragment or zygotic or somatic45MF-364940674Docket No.: 165362002340embryo in the genome editing reagent-containing composition, whereby the genome editing reagent is delivered to the seed or seed fragment or zygotic or somatic embryo from which plant cells or plant protoplasts are subsequently isolated. In embodiments, a flower bud or shoot tip is contacted with a genome editing reagent-containing composition, whereby the genome editing reagent is delivered to cells in the flower bud or shoot tip from which plant cells or plant protoplasts are subsequently isolated. In embodiments, a genome editing reagent-containing composition is applied to the surface of a plant or of a part of a plant (e.g., a leaf surface), whereby the genome editing reagent is delivered to tissues of the plant from which plant cells or plant protoplasts are subsequently isolated. In embodiments a whole plant or plant tissue is subjected to particle- or nanoparticle-mediated delivery (e.g., Biolistics or carbon nanotube or nanoparticle delivery) of a genome editing reagent-containing composition, whereby the genome editing reagent is delivered to cells or tissues from which plant cells or plant protoplasts are subsequently isolated.

[0162] Genome editing reagents and compositions comprising enhancer elements can be delivered to the plant and / or meristem cells of the plant by any method of delivery, by particle mediated delivery, and any other direct method of delivery, such as but not limiting to, Agrobacterium-mediated transformation, polyethylene glycol (PEG) -mediated transfection to protoplasts, whiskers mediated transformation, electroporation, particle bombardment, and / or by use of cell-penetrating peptides. In some embodiments, the enhancer element may be delivered to the plant or part thereof by treatment with pressure, centrifugation, Biolistics or particle bombardment, microinjection, infiltration (e.g., with a syringe), or by direct application to the surface of the plant tissue.

[0163] CRISPR technology for editing the genes of eukaryotes is disclosed in U.S. Patent Application Publications 2016 / 0138008A1 and US2015 / 0344912A1, and in U.S. Pat. Nos.8,697,359, 8,771,945, 8,945,839, 8,999,641, 8,993,233, 8,895,308, 8,865,406, 8,889,418, 8,871,445, 8,889,356, 8,932,814, 8,795,965, and 8,906,616. Cpfl endonuclease and corresponding guide RNAs and PAM sites are disclosed in U. S. Patent Application Publication 2016 / 0208243 Al. Other CRISPR nucleases useful for editing genomes include C2cl and C2c3 (see Shmakov et al. (2015) Mol. Cell, 60:385-397) and CasX and CasY (see Burstein et al. (2016) Nature, doi:10.1038 / nature21059). Plant RNA promoters for expressing CRISPR guide RNA and plant codon-optimized CRISPR Cas9 endonuclease are disclosed in International Patent Application PCT / US2015 / 018104 (published as WO 2015 / 131101 and claiming priority to US Provisional Patent Application 61 / 945,700). Methods of using46MF-364940674Docket No.: 165362002340CRISPR technology for genome editing in plants are disclosed in in U. S. Patent Application Publications US 2015 / 0082478A1 and US 2015 / 0059010A1 and in International Patent Application PCT / US2015 / 038767 Al (published as WO 2016 / 007347 and claiming priority to U.S. Provisional Patent Application 62 / 023,246).

[0164] In some embodiments, the enhancer element may be transformed into a plant or part thereof via at least one viral vector selected from the group consisting of adenoviruses, lentiviruses, adeno-associated viruses, retroviruses, geminiviruses, begomoviruses, tobamoviruses, potex viruses, comoviruses, wheat streak mosaic virus, barley stripe mosaic virus, bean yellow dwarf virus, bean pod mottle virus, cabbage leaf curl virus, beet curly top virus, tobacco yellow dwarf virus, tobacco rattle virus, potato virus X, and cowpea mosaic virus. In some embodiments, the enhancer element may be transformed into a plant or part thereof via at least one bacterial vector capable of transforming a plant cell and selected from the group consisting of Agrobacterium sp., Rhizobium sp., Sinorhizobium (Ensifer) sp., Mesorhizobium sp., Bradyrhizobium sp., Azobacter sp., and Phyllobacterium sp. In some embodiments, a viral vector may be delivered to a plant by transformation with Agrobacterium.

[0165] In some embodiments, a T-DNA vector may be used to deliver at least one enhancer element to a plant or part thereof. In some embodiments, a T-DNA binary vector is used. In some embodiments, a T-DNA superbinary vector system is used. In other embodiments, a T-DNA ternary vector system is used. In some embodiments, the T-DNA system further comprises an additional virulence gene cluster. In some embodiments, the T-DNA system further comprises an accessory plasmid or virulence helper plasmid. In some embodiments, the T-DNA vector is an Agrobacterium vector.F. Machine learning model architecture and training

[0166] The methods and systems described herein comprising using machine learning models trained to predict chromatin accessibility for a plant tissue from a sequence. In some embodiments, the methods and systems comprise training a machine learning model to predict chromatin accessibility for a plant tissue from a sequence. As described herein multiple machine learning models can be used and trained to predict chromatin accessibility for multiple plant tissues. Using and training multiple machine learning models to predict chromatin accessibility for multiple plant tissues improves the performance of the methods47MF-364940674Docket No.: 165362002340and systems in the ability to identify an enhancer element in a plant that will likely be active regardless of tissue context.

[0167] In some embodiments, the methods are for training a machine learning model to predict a measurement of chromatin accessibility in a dicot tissue from a sequence and comprise: obtaining training data comprising ATAC-seq read coverage for the dicot tissue, wherein the ATAC-seq read coverage represents a measurement of chromatin accessibility in the tissue from the dicot; selecting a plurality of positive genomic windows, wherein a positive genomic window is a region of Wm82v4 with ATAC-seq read coverage in the training data; selecting a plurality of negative genomic windows, wherein a negative genomic window is region of Wm82v4 without ATAC-seq read coverage in the training data; and training the machine learning model to predict a measurement of chromatin accessibility from a sequence wherein the training is based on the ATAC-seq read coverage at the plurality of positive genomic windows, ATAC-seq read coverage at the plurality negative genomic windows, the sequence in Wm82v4 corresponding to each of the positive genomic windows, the sequence in Wm82v4 corresponding to each of the negative genomic windows.

[0168] In some embodiments, the methods comprise training two or more machine learning models using multi-task learning. In some embodiments, the two or more machine learning models are configured as an ensemble model. In some embodiments, the ensemble model comprises model nodes trained independently to predict chromatin accessibility in a plant tissue such as cotyledon, flower, flower bud, hypocotyl, leaf, pod, or root.

[0169] The methods described herein for predicting a measurement for chromatin accessibility from a sequence can be used to predict chromatin accessibility for species other than the species the training data was generated from. The methods described herein comprise obtaining training data from soybean tissues. Due to the conservation of chromatin accessibility in tissues from plants in the same clade, the trained models can be used to application outside of soybean, for example in other dicot species.I. Training data

[0170] In some embodiments, the machine learning model, such as a machine learning model at block 908 of FIG. 9, has been trained using training data representing at least chromatin accessibility in a plant tissue. In some embodiments, the training data comprises ATAC-seq read coverage. In some embodiments, the training data comprises ATAC-seq read coverage for a plurality of samples of the plant tissue. In some embodiments, the ATAC-seq 48MF-364940674Docket No.: 165362002340read coverage represents a measurement of chromatin accessibility in the tissue from the plant (e.g. dicot, soybean). In some embodiments, the ATAC-seq read coverage is based at least on ATAC-seq data. In some embodiments, additional functional genomic data for the plant tissue can be integrated with ATAC-seq data to improve the ability of the machine learning model to predict chromatin accessibility in the plant tissues. In some embodiments, the functional genomic data comprise Cap Analysis of Gene expression sequencing (CAGE-seq), Short- and long-read sequencing of Transcriptionally active chromatin sequencing (STRIPE-seq), RNA sequencing (RNA-seq), and / or Self-Transcribing Active Regulatory Region Sequencing (STARR-seq). Incorporating diverse data types such as CAGE-seq, STRIPE-seq, RNA-seq, or STARR-seq improves chromatin accessibility predictions because each dataset provides complementary information about gene regulation and chromatin dynamics. CAGE-seq can be used to identify transcription start sites (TSSs) and TSSs often correlate with accessible chromatin regions. STRIPE-seq can provide insights into the actively transcribed regions of the genome. RNA-seq can measures gene expression levels, which can serve as an indirect marker of accessible chromatin regions. Highly expressed genes tend to have promoters and enhancers located in accessible chromatin, making RNA-seq valuable for inferring chromatin states. STARR-seq can be used to identify enhancer sequences and associated activity of the enhancers sequences. Each data type captures different facets of chromatin dynamics / chromatin behavior. Including each data type can improved model training because incorporating diverse datasets ensures that machine learning models are exposed to a broad range of biological signals, reducing bias and increasing the generalizability of predictions made by the machine learning model.

[0171] In some embodiments, the training data comprises ATAC-seq read coverage based on ATAC-seq data. In some embodiments, training data is generated using ATAC-seq data collected from a plant tissue (e.g. dicot tissue, soybean). In some embodiments, ATAC-seq data comprises pair-end sequencing reads that can be mapped to a plant genome using mapping methods known in the art, such as but not limited to bowtie BWA, HISAT, STAR, or TopHat. The ATAC-seq reads represent cut sites where the ATAC-seq transpose was able to cut the open chromatin. Accordingly, ATAC-seq reads can be used to infer accessible chromatin in a plant tissue. In some embodiments, chromatin accessibility measurements are generated from ATAC-seq reads by applying a peak calling method, such as a peak calling method known in the art, such as but not limited to MACS2. In some embodiments, a measurement of chromatin accessibility is ATAC-seq coverage in a peak.49MF-364940674Docket No.: 165362002340

[0172] Data processing methods may be used to improve the quality of the chromatin accessibility measurements and thus improve the performance of the machine learning models described herein to predict chromatin accessibility. In some embodiments, improving the quality of the chromatin accessibility measurements comprise removing data from regions of the genome expected to generate a disproportionate number of reads in the ATAC-seq data through data processing. In some embodiments, data processing may comprise cleaning the ATAC-seq data to account for transposase (e.g., Tn5) enzyme bias, read shifts, and removing data in regions expected to be false positive. In some embodiments, data processing may comprise cleaning the ATAC-seq data to account for repetitive genome sequences. Repetitive sequences, due to their repetitive nature, they may falsely accumulate large numbers of ATAC-seq reads. In some embodiments, the repetitive sequences may be sequences incorporated into the genome that originate from mitochondria and chloroplasts, whether biologically relevant or introduced through assembly errors, which are known to accumulate a disproportionate number of reads and can therefore skew distributions for learning purposes.

[0173] In some embodiments, data processing may comprise normalization of data with data collected in additional ATAC-seq experiments. In some embodiments, the GC-content of a region is accounted for during mapping and peak calling of the ATAC-seq data.

[0174] In some embodiments, the ATAC-seq data is bulk ATAC-seq data. In some embodiments, the ATAC-seq data is single cell ATAC-seq (scATAC-seq) data. In some embodiments, the ATAC-seq data is single nucleus ATAC-seq (snATAC-seq) data.

[0175] In some embodiments, the training data comprises ATAC-seq read coverage at ATAC-seq derived peaks. In some embodiments, the training data comprises ATAC-seq derived peaks one or more chromosomes of the plant genome (e.g. dicot). In some embodiments, the training data comprises ATAC-seq derived peaks spanning about 500 base pairs, about 1000 base pairs, about 1500 base pairs, about 2000 base pairs, about 2500 base pairs, about 3000 base pairs, about 3500 base pairs, about 4000 base pairs, about 4500 base pairs or about 5000 base pairs of genomic sequence.

[0176] In some embodiments, the machine learning models described herein are trained in two stages and thus the training data may comprise ATAC-seq derived peaks for training at each stage. In some embodiments, the first stage may comprise training a bias model. In some embodiments, the training data for the bias model may comprise ATAC-seq derived peaks in50MF-364940674Docket No.: 165362002340low signal regions. In some embodiments, ATAC-seq derived peaks in low signal regions are matched for GC content with regions comprising ATAC-seq derived peaks in high signal regions. In some embodiments, the second stage of training comprises training an accessibility model. In some embodiments, the training data for the accessibility model comprises bias-model-subtracted signal. In some embodiments, the training data comprises ATAC-seq read coverage at ATAC-seq peaks with high signal and ATAC-seq read coverage at non-peak regions.

[0177] In some embodiments, the training data comprises ATAC-seq read coverage. In some embodiments, the training data comprises log2 transformed ATAC-seq read coverage. In some embodiments, the training data comprises ATAC-seq coverage spanning about 500 base pairs, about 1000 base pairs, about 1500 base pairs, about 2000 base pairs, about 2500 base pairs, about 3000 base pairs, about 3500 base pairs, about 4000 base pairs, about 4500 base pairs or about 5000 base pairs of genomic sequence.

[0178] In some embodiments, the ATAC-seq read coverage comprises ATAC-seq coverage tracks. As described herein, ATAC-seq coverage tracks can account for ATAC-seq read coverage and peak shape for a genomic region. In some embodiments, training data comprising ATAC-seq coverage tracks can be used to train the machine learning model to pay attention to positional information (the peak location and shape) as well as the read counts in the output window (the peak height). In some embodiments, ATAC-seq read coverage relates to the height of an ATAC-seq derived peak. In some embodiments, ATAC-seq read coverage relates to the shape of an ATAC-seq derived peak. In some embodiments, ATAC-seq read coverage relates to the height and shape of an ATAC-seq derived peak.

[0179] In some embodiments, training data comprises ATAC-seq data as described herein and matched sequences. The matched sequences may be the nucleotide sequence from the reference genome the ATAC-seq reads were mapped to. In some embodiments, the reference genome is WM82v4. In some embodiments, the matched sequences comprise a nucleotide sequence. The matched sequences may comprise the nucleotide sequences corresponding to the ATAC-seq data used for training the machine learning model, such as the genomic location wherein the ATAC-seq data maps to. In some embodiments, the matched sequences comprise about 500 nucleotides, about 1000 nucleotides, about 1500 nucleotides, about 2000 nucleotides, about 2500 nucleotides, about 3000 nucleotides, about 3500 nucleotides, about 4000 nucleotides, about 4500 nucleotides or about 5000 nucleotides.51MF-364940674Docket No.: 165362002340

[0180] In some embodiments, the training data comprises ATAC-seq read coverage for a plurality of samples of the plant tissue. In some embodiments, the plurality of samples may be from the same plant tissue or from multiple plant tissues. In some embodiments, the plurality of samples comprises at least 1000, 2000, 3000, 4000, or 5000 samples.

[0181] In some embodiments, the training data comprises ATAC-seq coverage data for between about 100,000 and about 200,000 ATAC-seq peaks per sample. In some embodiments, model may be trained with ATAC-seq coverage data for the between about 100,000 and about 200,000 ATAC-seq peaks for a single sample. In some embodiments, the between about 100,000 and about 200,000 ATAC-seq peaks may represent high signal and low signal regions as described herein.

[0182] In some embodiments, models trained with ATAC-seq derived peaks for multiple tissues outperforms machine learning models trained with ATAC-seq derived peaks from a single tissue. In some embodiments, the training data comprises ATAC-seq derived peaks for one plant tissue. In some embodiments, the training data comprises ATAC-seq derived peaks for one or more, two or more, three or more, four or more, five or more, six or more, seven or more or eight or more plant tissues, such as the plant tissues described herein.IL Machine learning model architectures and training

[0183] Machine learning algorithms can be advantageous for a number of reasons. For example, machine learning algorithms may automatically and quickly produce desired results like predicting chromatin accessibility data for a plant tissue. In some embodiments, machine learning algorithms may be robust than conventional algorithms. For example, a conventional algorithm may only be capable of predicting chromatin accessibility in ways that a programmer has specifically accounted for. A machine learning algorithm may be able to process data with unanticipated characteristics (e.g., new polynucleotide sequences) based on training data provided to the machine learning algorithm. In some embodiments, machine learning algorithms may be more accurate than conventional algorithms. For example, a machine learning algorithm may improve over time with reinforcement mechanisms (e.g., if a human operator periodically performs manual checks on machine learning outputs). In some embodiments, conventional algorithms may have to be re-written and / or modified to improve the algorithm’s accuracy.

[0184] FIG. 10 depicts a non-limiting example of a method for training a machine learning model to predict a measurement of chromatin accessibility. In some embodiments, the 52MF-364940674Docket No.: 165362002340method depicted in FIG. 10 is for training a machine learning to predict a measurement of chromatin accessibility in a plant tissue (e.g. dicot tissue) and comprises obtaining training data, selecting a plurality of positive genomic windows, selecting a plurality of negative genomic windows, and training the machine learning model with at least ATAC-seq read coverage at the plurality of positive and negative genomic windows and the reference genome sequence (e.g. Wm82v4) at each of the positive and negative genomic windows.

[0185] At block 1002, training data is obtained. The training data, as described herein, comprises ATAC-seq read coverage for the dicot tissue, wherein the ATAC-seq read coverage represents a measurement of chromatin accessibility in the tissue from the dicot.

[0186] At block 1004, the training data from block 1002 is used to select positive genomic window. Block 1004 comprises selecting a plurality of positive genomic windows, wherein a positive genomic window is a region with ATAC-seq read coverage in the training data. In some embodiments, the genomic windows are regions of the Wm82v4 genome. The positive genomic windows may be selected as ATAC-seq derived peaks I the training data with the highest ATAC-seq read coverage. The positive genomic windows represent regions of the genome likely to be open chromatin in the training data samples. The positive genomic window may comprise ATAC-seq derived peaks mapping to genes.

[0187] In some embodiments, a positive genomic window may be 500 nucleotides, about 1000 nucleotides, about 1500 nucleotides, about 2000 nucleotides, about 2500 nucleotides, about 3000 nucleotides, about 3500 nucleotides, about 4000 nucleotides, about 4500 nucleotides or about 5000 nucleotides. In some embodiments, a positive genomic window may comprise one or more ATAC-seq derived peaks.

[0188] At block 1006, the training data from block 1002 is used to select negative genomic window. Block 1006 comprises selecting a plurality of negative genomic windows, wherein a negative genomic window is a region without ATAC-seq read coverage in the training data. In some embodiments, the genomic windows are regions of the Wm82v4 genome. In some embodiments, a negative genomic window may be selected as a region of the genome without any ATAC-seq derived peaks. In some embodiments, a negative genomic window may be an intergenic region of the reference genome. In some embodiments, a negative genomic window may be selected by maximizing the genomic distance to a ATAC-seq derived peak or positive genomic window. In some embodiments, the negative genomic53MF-364940674Docket No.: 165362002340windows may be selected to match sequence characteristics of the positive genomic windows such as but not limited to, chromosome, length, or nucleotide composition (e.g. GC content).

[0189] In some embodiments, a negative genomic window may be 500 nucleotides, about 1000 nucleotides, about 1500 nucleotides, about 2000 nucleotides, about 2500 nucleotides, about 3000 nucleotides, about 3500 nucleotides, about 4000 nucleotides, about 4500 nucleotides or about 5000 nucleotides.

[0190] It is appreciated that selecting positive genomic windows at block 1004, and selecting negative genomic windows at block 1006 can be performed in any order. In some embodiments, block 1004 is performed before block 1006. In some embodiments, block 1006 is performed before block 1004. In some embodiments, block 1004 and block 1006 are performed simultaneously.

[0191] In some embodiments, the training data obtained at block 1002 comprises ATAC-seq read coverage data from a plurality of samples of the plant tissue. In some embodiments, the methods comprise filtering the training data to remove data from one or more of the plurality of sample of the plant tissue. In some embodiments, filtering the training data comprises filtering data from samples of the plant tissue with excessive ATAC-seq read coverage. Excessive ATAC-seq read coverage may be outlier ATAC-seq read coverage when the ATAC-seq read coverage is compared across the genome for the plurality of samples. Excessive ATAC-seq read coverage may be outlier ATAC-seq read coverage when the ATAC-seq read coverage is compared across the positive genomic windows for the plurality of samples. Filtering samples with excessive read coverage may ensure that one sample of the plurality of samples does not disproportionally contribute to training of the machine learning model. In some embodiments, the method comprise removing ATAC-seq read coverage data mapping to genomic regions with repetitive and / or high fractions of uncertain bases. In some embodiments, uncertain bases may be represented in a reference genome sequence as an “N” rather than “A”, “T”, “G”, or “C”. In some embodiments, the repetitive or high uncertainty regions are less informative in terms of sequence content for the model and removing them may improve model performance.

[0192] It can be beneficial to curate and / or pre-process a set of training data, such as the training data described herein. In some embodiments, processed training data may yield a more accurate and / or more robust convolutional neural network and may increase a speed at which a neural network increases its accuracy. In some embodiments, curating the training54MF-364940674Docket No.: 165362002340data comprises oversampling training data. In some embodiments, oversampling training data comprises oversampling ATAC-seq read coverage in positive genomic windows.

[0193] In some embodiments, curating the training data comprises augmenting the sequences by up to about 128 nucleotides on either side. In some embodiments, the methods may comprise increasing the size of the positive genomic windows and negative genomic window by up to 128 nucleotides on either side. The methods may comprise training the machine learning model with the reference genome (e.g. Wm82v4) sequence corresponding to the augmented genomic windows.

[0194] In some embodiments, curating the training data comprises augmenting the sequences using methods established in deep learning to improve model performance. In some embodiments, augmentation may comprise introducing random jitter and 50%-chance of random reverse-complementation. In some embodiments, random jitter comprises shifting the sequence windows and target sequences upstream or downstream by up to about 128 nucleotides. In some embodiments, adding random jitter incentives the model to identify sequence patterns that are robust to small random changes in the context and discourages it from memorizing peaks, thus improving performance.

[0195] In some embodiments, curating the training data may comprise inspecting statistical properties of the training data such as the ATAC-seq read coverage at the plurality of positive genomic windows and the plurality of negative genomic windows. The positive and negative genomic windows may be adjusted to account for GC content of the sequences corresponding to each of the windows. The positive and negative genomic windows may be adjusted to balance GC contents across the sequences corresponding to positive and negative genomic windows so that GC content doesn’t disproportionately contribute to training of the machine learning model.

[0196] In some embodiments, the methods comprise storing the training data in an Apache Arrow data format. The Apache Arrow dataset format improves the efficiency of model loading and training. Using the Apache Arrow dataset format allows for iterative training and scalability of model training as additional training data is obtained.

[0197] At block 1008, the methods comprise training the machine learning model using the trained data obtained at block 1002, the positive genomic windows selected at 1004 and the negative genomic windows selected at 1006. Block 1008 comprises training a machine 55MF-364940674Docket No.: 165362002340learning mode to predict a chromatin accessibility measurement for a sequence comprise training the machine learning model to predict a measurement of chromatin accessibility from a sequence wherein the training is based on the ATAC-seq read coverage at the plurality of positive genomic windows, ATAC-seq read coverage at the plurality negative genomic windows, the sequence in the plant reference genome corresponding to each of the positive genomic windows, the sequence in plant reference genome corresponding to each of the negative genomic windows. In some embodiments, the plant reference genome is Wm82v4. The methods at block 1008, comprise training the machine learning model by optimizing parameters of the machine learning model

[0198] In some embodiments, the machine model described herein comprises one or more deep machine learning models. In some embodiments, the deep machine learning model is a neural network, such as an artificial neural network. In some embodiments the neural network is convolutional neutral network (CNN).

[0199] A CNN comprises an input layer with input neurons, an output later with at least one output layer, and multiple hidden layers between the input layer and output layer. In some embodiments, the hidden layers of a convolutional layers, ReLU (Rectified Linear Units) layers (i.e., activation function layers), pooling layers, fully connected layers, and normalization layers.

[0200] In some embodiments, the multiple hidden layers may be a sequence of convolutional and pooling layers. Each convolution layer may comprise a plurality of parameters used for performing the convolution operations. Each convolution layer may also comprise one or more filters, which in turn may comprise one or more weighting factors or other adjustable parameters. In some instances, the parameters may include biases (i.e., parameters that permit the activation function to be shifted). In some cases,the convolutional layers are followed by a layer of ReLU activation function. Other activation functions can also be used, for example the saturating hyperbolic tangent, identity, binary step, logistic, arcTan, softsign, parameteric rectified linear unit, exponential linear unit, softPlus, bent identity, softExponential, Sinusoid, Sine, Gaussian, the sigmoid function and various others. The convolutional, pooling and ReLU layers may function as learnable features extractors.

[0201] In some embodiments, the dilated CNN. A dilated CNN is a CNN adapted to increase the receptive field for a network without increasing the number of parameters. In a56MF-364940674Docket No.: 165362002340dilated CNN gaps may be inserted between filters that are used to perform convolution operations on the data imputed into the input layer. The size of the gaps may be adjusted during hyperparameter finetuning. Use of a dilated CNN may improve computational efficiency for the machine learning model as described herein because more complex patterns in the input data can be encoded in the model without increasing the number of parameters that need to be adjusted during training of the machine learning model.

[0202] In some embodiments, the machine learning model is BPNet, Avsec et.al., Base-resoultuion model of transcription-factor binding refeal soft motif syntax, 53 Nature Genetics (2021), or a machine learning model derived therefrom. In some embodiments, the machine learning model is ChromBPNet., Brennan et al., Chromatin accessibility is a two-tier process regulated by transcription factor pioneering and enhancer activation, Bioarxiv (2022).ChromBPNet is a modification to BPNet wherein the training process accounts for ATAC-seq data parameters such as Tn5 sequence bias that may influence ATAC-seq read data. In some embodiments, training a ChromBPNet model comprises use of a frozen pre-trained model that has been trained to account for Tn5 sequence bias and an unfrozen, randomly initialized model that learns the unbiased associations between ATAC-seq read coverage and a DNA sequence. After training ChromBPNet, the frozen Tn5 model can be removed from the architecture and the unfrozen model can be used as the machine learning model described herein.

[0203] In some embodiments, training the machine learning models as described here comprise optimization of parameters and hyperparameters. In some embodiments, training is performed as described in Brennan et al., Chromatin accessibility is a two-tier process regulated by transcription factor pioneering and enhancer activation, Bioarxiv (2022), using training data for plant tissues as described here. In some embodiments, training the machine learning model comprises k-fold or cross validation.

[0204] For training, a cross-validation method can be employed to split the training data into a training data set and a validation data set. The training data set is used in the backpropagation training of the network weights. The validation data set is used to verify that the trained network generalizes to make good predictions. The best network weight set can be taken as the one that best predicts the outputs of the training data. Similarly, varying the number of network hidden nodes and determining the network that performs best with the data sets optimizes the number of hidden nodes.57MF-364940674Docket No.: 165362002340

[0205] As described herein, in some embodiments, the training methods can be applied to multiple plant tissues and the two or more machine learning models can be configured in an ensemble model format, wherein the predicted chromatin accessibility measurements from each independently trained machine learning model can be aggregated.

[0206] In some embodiments, software known in the art, such as but not limited to Keras and TensorFlow, may be used for training and applying the machine learning models described herein.G. Systems

[0207] FIG. 7 illustrates a flow diagram for identifying an enhancer element in plant, in accordance with the presently disclosed embodiments. The flow diagram may be performed utilizing one or more processing devices that may include hardware (e.g., a general purpose processor, a graphic processing unit (GPU), an application- specific integrated circuit (ASIC), a system-on-chip (SoC), a microcontroller, a field-programmable gate array (FPGA), a central processing unit (CPU), an application processor (AP), a visual processing unit (VPU), a neural processing unit (NPU), a neural decision processor (NDP), a deep learning processor (DLP), tensor processing unit (TPU), or any other processing device(s) that may be suitable for processing genomics data or other omics data), software (e.g., instructions running / executing on one or more processors), firmware (e.g., microcode), or some combination thereof.

[0208] Provided herein are systems that can be used to identify an enhancer element in a plant, the systems comprising: one or more processors; a user input device and a memory communicatively coupled to the one or more processors and configured to store instructions that, when executed by the one or more processors, cause the system to: (a) receive a first candidate sequence from the user input device; (b) generate a plurality of candidate sequences from the first candidate sequence using single base pair saturation mutagenesis; (c) for each promoter sequence in a plurality of promoter sequences, generate a plurality of input sequences, each comprising a candidate sequence from the plurality of candidate sequences and the promoter sequence; (d) input the plurality of input sequences into two or more machine learning models trained to predict a measurement of chromatin accessibility in one or more plant tissues from a sequence; (e) receive, from the two or more machine learning models, a predicted chromatin accessibility measurement for each of the plurality of input sequences; (f) synthesize a composite measurement for each candidate sequence related to the58MF-364940674Docket No.: 165362002340predicted chromatin accessibility measurements for each input sequence in the plurality of input sequences comprising the candidate sequence from each of the two or more machine learning models; (g) select a top performing sequence from the plurality of candidate sequences based on the composite measurement; (h) if the top performing sequence matches the first candidate sequence, identify the top performing sequence as the enhancer element in the plant, and if the top performing sequence does not match the first candidate sequence, update the first candidate sequence to be top performing sequence and repeating steps (b) -(g). In some embodiments, the systems comprise one or more computing systems as described herein.

[0209] FIG. 11 illustrates an example computing system for identifying an enhancer elements in a plant 1100 that may be utilized for provisioning a platform account and associated sub-account and servicing transactions utilizing the provisioned platform account and associated sub-account, in accordance with the presently disclosed embodiments. In certain embodiments, one or more genome editing computing system 1100 perform one or more steps of one or more methods described or illustrated herein. Herein, reference to a computer system may encompass a computing device, and vice versa, where appropriate. Moreover, reference to a computer system may encompass one or more computer systems, where appropriate.

[0210] This disclosure contemplates any suitable number of systems 1100. This disclosure contemplates computing system 1100 taking any suitable physical form. As example and not by way of limitation, computing system 1100 may be an embedded computer system, a system-on-chip (SOC), a single-board computer system (SBC) (e.g., a computer-on-module (COM) or system-on-module (SOM)), a desktop computer system, a laptop or notebook computer system, an interactive kiosk, a mainframe, a mesh of computer systems, a mobile telephone, a personal digital assistant (PDA), a server, a tablet computer system, an augmented / virtual reality device, or a combination of two or more of these. Where appropriate, computing system 1000 may include one or more computing systems 1100; be unitary or distributed; span multiple locations; span multiple machines; span multiple data centers; or reside in a cloud, which may include one or more cloud components in one or more networks.

[0211] Where appropriate, one or more computing systems 1100 may perform without substantial spatial or temporal limitation one or more steps of one or more methods described or illustrated herein. As an example, and not by way of limitation, one or more editing 59MF-364940674Docket No.: 165362002340computing systems 1100 may perform in real time or in batch mode one or more steps of one or more methods described or illustrated herein. One or more computing system 1100 may perform at different times or at different locations one or more steps of one or more methods described or illustrated herein, where appropriate.

[0212] In certain embodiments, the computing system 1100 includes a processor 1102, memory 1104, database 1106, an input / output (I / O) interface 1108, a communication interface 1110, and a bus 1112. Although this disclosure describes and illustrates a particular computer system having a particular number of particular components in a particular arrangement, this disclosure contemplates any suitable computer system having any suitable number of any suitable components in any suitable arrangement. In certain embodiments, processor 1102 includes hardware for executing instructions, such as those making up a computer program. As an example, and not by way of limitation, to execute instructions, processor 1102 may retrieve (or fetch) the instructions from an internal register, an internal cache, memory 1104, or database 1106; decode and execute them; and then write one or more results to an internal register, an internal cache, memory 1104, or database 1106. In certain embodiments, processor 1102 may include one or more internal caches for data, instructions, or addresses. This disclosure contemplates processor 1102 including any suitable number of any suitable internal caches, where appropriate. As an example, and not by way of limitation, processor 1102 may include one or more instruction caches, one or more data caches, and one or more translation lookaside buffers (TLBs). Instructions in the instruction caches may be copies of instructions in memory 1104 or database 1106, and the instruction caches may speed up retrieval of those instructions by processor 1102.

[0213] Data in the data caches may be copies of data in memory 1104 or database 1106 for instructions executing at processor 1102 to operate on; the results of previous instructions executed at processor 1102 for access by subsequent instructions executing at processor 1102 or for writing to memory 1104 or database 1106; or other suitable data. The data caches may speed up read or write operations by processor 1102. The TLBs may speed up virtual-address translation for processor 1102. In certain embodiments, processor 1102 may include one or more internal registers for data, instructions, or addresses. This disclosure contemplates processor 1102 including any suitable number of any suitable internal registers, where appropriate. Where appropriate, processor 1102 may include one or more arithmetic logic units (ALUs); be a multi-core processor; or include one or more processors 1102. Although60MF-364940674Docket No.: 165362002340this disclosure describes and illustrates a particular processor, this disclosure contemplates any suitable processor.

[0214] In certain embodiments, memory 1104 includes main memory for storing instructions for processor 1102 to execute or data for processor 1102 to operate on. As an example, and not by way of limitation, computing system 1100 may load instructions from database 1106 or another source (such as, for example, another computing system 1100) to memory 1104. Processor 1102 may then load the instructions from memory 1104 to an internal register or internal cache. To execute the instructions, processor 1102 may retrieve the instructions from the internal register or internal cache and decode them. During or after execution of the instructions, processor 1102 may write one or more results (which may be intermediate or final results) to the internal register or internal cache. Processor 1102 may then write one or more of those results to memory 1104. In certain embodiments, processor 1102 executes only instructions in one or more internal registers or internal caches or in memory 1104 (as opposed to database 1106 or elsewhere) and operates only on data in one or more internal registers or internal caches or in memory 1104 (as opposed to database 1106 or elsewhere).

[0215] One or more memory buses (which may each include an address bus and a data bus) may couple processor 1102 to memory 1104. Bus 1112 may include one or more memory buses, as described below. In certain embodiments, one or more memory management units (MMUs) reside between processor 1102 and memory 1104 and facilitate accesses to memory 1104 requested by processor 1102. In certain embodiments, memory 1104 includes random access memory (RAM). This RAM may be volatile memory, where appropriate. Where appropriate, this RAM may be dynamic RAM (DRAM) or static RAM (SRAM). Moreover, where appropriate, this RAM may be single-ported or multi-ported RAM. This disclosure contemplates any suitable RAM. Memory 1104 may include one or more memory devices 111104, where appropriate. Although this disclosure describes and illustrates particular memory, this disclosure contemplates any suitable memory.

[0216] In certain embodiments, database 1106 includes mass storage for data or instructions. As an example, and not by way of limitation, database 1106 may include a hard disk drive (HDD), a floppy disk drive, flash memory, an optical disc, a magneto-optical disc, magnetic tape, or a Universal Serial Bus (USB) drive or a combination of two or more of these. Database 1106 may include removable or non-removable (or fixed) media, where appropriate. Database 1106 may be internal or external to the computing system 1100, where 61MF-364940674Docket No.: 165362002340appropriate. In certain embodiments, database 1106 is non-volatile, solid-state memory. In certain embodiments, database 1106 includes read-only memory (ROM). Where appropriate, this ROM may be mask-programmed ROM, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), electrically alterable ROM (EAROM), or flash memory or a combination of two or more of these. This disclosure contemplates mass database 1106 taking any suitable physical form. Database 1106 may include one or more storage control units facilitating communication between processor 1102 and database 1106, where appropriate. Where appropriate, database 1106 may include one or more storages 1106. Although this disclosure describes and illustrates particular storage, this disclosure contemplates any suitable storage.

[0217] In certain embodiments, VO interface 1108 includes hardware, software, or both, providing one or more interfaces for communication between genome editing computing system 1100 and one or more VO devices. Computing system 1100 may include one or more of these VO devices, where appropriate. One or more of these VO devices may enable communication between a person and genome editing computing system 1100. In some embodiments, the VO interface may comprise a user input device. As an example, and not by way of limitation, an VO device may include a keyboard, keypad, microphone, monitor, mouse, printer, scanner, speaker, still camera, stylus, tablet, touch screen, trackball, video camera, another suitable VO device or a combination of two or more of these. An VO device may include one or more sensors. This disclosure contemplates any suitable VO devices and any suitable VO interfaces 1106 for them. Where appropriate, VO interface 1108 may include one or more device or software drivers enabling processor 1102 to drive one or more of these VO devices. VO interface 1108 may include one or more VO interfaces 1106, where appropriate. Although this disclosure describes and illustrates a particular I / O interface, this disclosure contemplates any suitable VO interface.

[0218] In certain embodiments, communication interface 1110 includes hardware, software, or both providing one or more interfaces for communication (such as, for example, packetbased communication) between genome editing computing system 1100 and one or more other computer systems 1100 or one or more networks. As an example, and not by way of limitation, communication interface 1110 may include a network interface controller (NIC) or network adapter for communicating with an Ethernet or other wire-based network or a wireless NIC (WNIC) or wireless adapter for communicating with a wireless network, such62MF-364940674Docket No.: 165362002340as a WI-FI network. This disclosure contemplates any suitable network and any suitable communication interface 810 for it.

[0219] As an example, and not by way of limitation, computing system 1100 may communicate with an ad hoc network, a personal area network (PAN), a local area network (LAN), a wide area network (WAN), a metropolitan area network (MAN), or one or more portions of the Internet or a combination of two or more of these. One or more portions of one or more of these networks may be wired or wireless. As an example, computing system 1100 may communicate with a wireless PAN (WPAN) (such as, for example, a BLUETOOTH WPAN), a WI-FI network, a WLMAX network, a cellular telephone network (such as, for example, a Global System for Mobile Communications (GSM) network), or other suitable wireless network or a combination of two or more of these. Genome editing computing system 1100 may include any suitable communication interface 1110 for any of these networks, where appropriate. Communication interface 1110 may include one or more communication interfaces 1110, where appropriate. Although this disclosure describes and illustrates a particular communication interface, this disclosure contemplates any suitable communication interface.

[0220] In certain embodiments, bus 1112 includes hardware, software, or both coupling components of genome editing computing system 1100 to each other. As an example, and not by way of limitation, bus 1012 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a front-side bus (FSB), a HYPERTRANSPORT (HT) interconnect, an Industry Standard Architecture (ISA) bus, an INFINIBAND interconnect, a low-pin-count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCIe) bus, a serial advanced technology attachment (SATA) bus, a Video Electronics Standards Association local (VLB) bus, or another suitable bus or a combination of two or more of these. Bus 1112 may include one or more buses 1112, where appropriate. Although this disclosure describes and illustrates a particular bus, this disclosure contemplates any suitable bus or interconnect.

[0221] Herein, a computer-readable non-transitory storage medium or media may include one or more semiconductor-based or other integrated circuits (ICs) (such, as for example, field-programmable gate arrays (FPGAs) or application-specific ICs (ASICs)), hard disk drives (HDDs), hybrid hard drives (HHDs), optical discs, optical disc drives (ODDs), magneto-optical discs, magneto-optical drives, floppy diskettes, floppy disk drives (FDDs),63MF-364940674Docket No.: 165362002340magnetic tapes, solid-state drives (SSDs), RAM-drives, SECURE DIGITAL cards or drives, any other suitable computer-readable non-transitory storage media, or any suitable combination of two or more of these, where appropriate. A computer-readable non-transitory storage medium may be volatile, non-volatile, or a combination of volatile and non-volatile, where appropriate.

[0222] The embodiments disclosed herein are only examples, and the scope of this disclosure is not limited to them. Embodiments according to this disclosure are in particular disclosed in the attached claims directed to a method, a storage medium, a system and a computer program product, wherein any feature mentioned in one claim category, e.g. method, can be claimed in another claim category, e.g. system, as well. The dependencies or references back in the attached claims are chosen for formal reasons only. However, any subject matter resulting from a deliberate reference back to any previous claims (in particular multiple dependencies) can be claimed as well, so that any combination of claims and the features thereof are disclosed and can be claimed regardless of the dependencies chosen in the attached claims. The subject-matter which can be claimed comprises not only the combinations of features as set out in the attached claims but also any other combination of features in the claims, wherein each feature mentioned in the claims can be combined with any other feature or combination of other features in the claims. Furthermore, any of the embodiments and features described or depicted herein can be claimed in a separate claim and / or in any combination with any embodiment or feature described or depicted herein or with any of the features of the attached claims.

[0223] The scope of this disclosure encompasses all changes, substitutions, variations, alterations, and modifications to the example embodiments described or illustrated herein that a person having ordinary skill in the art would comprehend. The scope of this disclosure is not limited to the example embodiments described or illustrated herein. Moreover, although this disclosure describes and illustrates respective embodiments herein as including particular components, elements, feature, functions, operations, or steps, any of these embodiments may include any combination or permutation of any of the components, elements, features, functions, operations, or steps described or illustrated anywhere herein that a person having ordinary skill in the art would comprehend. Furthermore, reference in the appended claims to an apparatus or system or a component of an apparatus or system being adapted to, arranged to, capable of, configured to, enabled to, operable to, or operative to perform a particular function encompasses that apparatus, system, component, whether or not it or that particular64MF-364940674Docket No.: 165362002340function is activated, turned on, or unlocked, as long as that apparatus, system, or component is so adapted, arranged, capable, configured, enabled, operable, or operative. Additionally, although this disclosure describes or illustrates certain embodiments as providing particular advantages, certain embodiments may provide none, some, or all of these advantages.EXEMPLARY IMPLEMENTATIONSAmong the provided embodiments are:1. A computer implemented method for identifying an enhancer element in a plant, the method comprising:by one or more computing devices comprising one or more processors and memory:(a) receiving a first candidate sequence from the memory;(b) generating a plurality of candidate sequences from the first candidate sequence using single base pair saturation mutagenesis;(c) for each promoter sequence in a plurality of promoter sequences, generating a plurality of input sequences, each comprising a candidate sequence from the plurality of candidate sequences and the promoter sequence;(d) inputting the plurality of input sequences into two or more machine learning models trained to predict a measurement of chromatin accessibility in one or more different plant tissues from a sequence;(e) receiving, from the two or more machine learning models, a predicted chromatin accessibility measurement for each of the plurality of input sequences;(f) synthesizing a composite measurement for each candidate sequence related to the predicted chromatin accessibility measurement for each input sequence in the plurality of input sequences comprising the candidate sequence from each of the two or more machine learning models;(g) selecting a top performing sequence from the plurality of candidate sequences based on the composite measurement; and(h) if the top performing sequence matches the first candidate sequence, identifying the top performing sequence as the enhancer element in the plant, and if the top performing sequence does not match the first candidate sequence, updating the first candidate sequence to be top performing sequence and repeating steps (b) - (g).65MF-364940674Docket No.: 1653620023402. The computer implemented method of embodiment 1, further comprising experimentally validating the enhancer element increases expression of one or more genes in the plant.3. The computer implemented method of embodiment 1 or embodiment 2, wherein the first candidate sequence comprises a polynucleotide sequence.4. The computer implemented method of any one of embodiments 1-3, wherein the first candidate sequence comprises a randomly generated polynucleotide sequence comprising a uniform distribution of adenine, guanine, cytosine, and thymine nucleotides.5. The computer implemented method of any one of embodiments 1-4, wherein the first candidate sequence has a length of N nucleotides.6. The computer implemented method of embodiment 5, wherein N is between 10 and 60.7. The computer implemented method of embodiment 5 or 6, wherein the plurality of candidate sequences comprises 3 x IV + 1 candidate sequences.8. The computer implemented method of any one of embodiments 1-3, wherein the first candidate sequence comprises a polynucleotide sequence comprising a known enhancer motif flanked by one or more variable nucleotides.9. The computer implemented method of any one of embodiments 1-3 and 8, wherein the first candidate sequence comprises a polynucleotide sequence comprising a known enhancer motif repeated one or more times.10. The computer implemented method of embodiment 8 or 9, wherein the known enhancer motif is a k-mer associated with a promoter of a plurality of highly expressed genes.11. The computer implemented method of embodiment 10, wherein the k-mer comprises a polynucleotide sequence identified by:66MF-364940674Docket No.: 165362002340identifying a plurality of promoters associated with highly expressed genes in the plant; andidentifying a polynucleotide sequence enriched in sequences for the plurality of promoters.12. The computer implemented method of embodiment 10 or 11, wherein the plurality of promoters have been identified as accessible in a plurality of cell types using snATAC-seq.13. The computer implemented method of any one of embodiments 10-12, wherein the k-mer comprises a polynucleotide sequence less than about 12 nucleotides.14. The computer implemented method of any one of embodiments 8-13, wherein the known enhancer motif is held constant during the single base pair saturation mutagenesis.15. The computer implemented method of any one of embodiments 8-14, wherein the plurality of candidate sequences comprises 3 x M + 1 candidate sequences, wherein M is the number of variable nucleotides in the first candidate sequence.16. The computer implemented method of any one of embodiments 1-15, wherein single base pair saturation mutagenesis comprises separately substituting the nucleotide at each position in the first candidate sequence with an adenine, a thymine, a guanine, and a cytosine nucleotide.17. The computer implemented method of any one of embodiments 1-16, wherein a promoter sequence in the plurality of promoter sequences is a known promoter sequence element in the genome of the plant.18. The computer implemented method of any one of embodiments 1-17, wherein each input sequence comprises the candidate sequence at the 5’ end of the promoter sequence.19. The computer implemented method of any one of embodiments 1-18, wherein each input sequence comprises the candidate sequence repeated about 1, about 2, or about 3 times.67MF-364940674Docket No.: 16536200234020. The computer implemented method of any one of embodiments 1-19, wherein the two or more machine learning models have each been trained to predict a measurement of chromatin accessibility a plant tissue based on ATAC-seq data collected from the plant tissue.21. The computer implemented method of any one of embodiments 1-20, wherein the two or more machine learning models have been trained to predict chromatin accessibility in different plant tissues.22. The computer implemented methods of any one of embodiments 1-21, wherein the one or more plant tissues comprise plant tissues from a dicot.23. The computer implemented methods of any one of embodiments 1-22, wherein the one or more plant tissues comprise plant tissues from a dicot and the plant is a dicot.24. The computer implemented methods of any one of embodiments 1-23, wherein the one or more plant tissues comprise plant tissues from the same species as the plant.25. The computer implemented method of any one of embodiments 1-24, wherein a machine learning model of the two or more machine learning models comprises a convolutional neural network (CNN).26. The computer implemented method of embodiment 25, wherein the CNN is a dilated CNN.27. The computer implemented method of any one embodiments 1-26, wherein the one or more plant tissues are selected from a group consisting of bud, cotyledon, flower, flower bud, hypocotyl, leaf, pod, and root.28. The computer implemented method of any one of embodiments 1-27, wherein synthesizing the composite measurement comprises averaging the predicted chromatin accessibility measurements for each input comprising the candidate sequence from one of the two or more machine learning models and averaging the predicted chromatin accessibility measurements for each candidate sequence across the two or more machine learning models.68MF-364940674Docket No.: 16536200234029. The computer implemented method of any one of embodiments 1-28, wherein the composite measurement is based on the chromatin accessibility of the candidate sequence when the candidate sequence is linked to the plurality of promoters in two or more tissues.30. The computer implemented method of any one of embodiments 1-29, wherein the selecting the top performing sequence comprises selecting the candidate sequence with the highest composite measurement among the composite measurements for the plurality of candidate sequences.31. The computer implemented method of any one of embodiments 1-30, wherein the repeating steps (b)-(g) increases a probability the top performing sequence matches the first candidate sequence in a subsequent repetition of the steps.32. The computer implemented method of any one of embodiments 1-31, comprising repeating steps (b)-(g) about 10 times.33. The computer implemented method of any one of embodiments 16-28, wherein the selecting the top performing sequence comprises selecting two or more candidate sequences with the highest composite measurement among the composite measurements for the plurality of candidate sequences and performing steps (b)-(g) with each of the two or more candidate sequences as the first candidate sequence.34. The computer implemented method of any one of embodiments 1-33, wherein the top performing sequence matches the candidate sequence if the nucleotide sequence of the top performing sequence is the same as the nucleotide sequence of the candidate sequence.35. The computer implemented method of any one of embodiments 1-34, wherein the plant is soybean.36. The computer implemented method of any one of embodiment 35, wherein the ATAC-seq data comprises soybean data collected from Williams 82 soybeans.37. The computer implemented method of any one of embodiments 2-36, wherein experimentally validating the enhancer element comprises a reporter assay.69MF-364940674Docket No.: 16536200234038. The computer implemented method of embodiment 37, wherein the reporter assay comprises transforming a polynucleotide vector comprising, the enhancer element operably linked to an endogenous promoter, a transcription start site, and a reporter gene into a plant or part thereof, wherein the reporter gene is a luciferase gene, and testing for expression of the luciferase gene.39. The computer implemented method of embodiment 38, wherein the polynucleotide vector comprises the enhancer element repeated three times.40. The computer implemented method of embodiment 37, wherein the reporter assay comprises:transforming a first polynucleotide vector comprising, the enhancer element operably linked to an endogenous promoter a transcription start site, and a reporter gene into a first plant or part thereof from the plant;transforming a second polynucleotide vector comprising, a control sequence element operably linked to the endogenous promoter, the transcription start site, and the reporter gene into a second plant or part thereof from the plant; andmeasuring the expression of the reporter gene in the first plant or part thereof and in the second plant or part thereof, wherein increased expression of the reporter gene in the first plant or part thereof compared to the second plant or part thereof indicates validation of the enhancer element.41. The computer implemented method of embodiment 40, wherein the first polynucleotide vector comprises the enhancer element repeated three times.42. The computer implemented method of embodiment 41, wherein the control sequence element comprises an octopine synthase (OCS) enhancer element.43. The computer implemented method of embodiment 41, wherein the control sequence element comprises a G-box element.44. The computer implemented method of any one of embodiments 38-43 wherein the endogenous promoter is promoter for a target gene.70MF-364940674Docket No.: 16536200234045. The computer implemented method of embodiment 44, wherein the target gene is a gene selected from a group consisting of AlPlOa, AlPlOb, AML4, CRN, HB-1, RIC1, RIC2, RPF1, JAG1, JAG2, KHZ1, PP2C, TCP5-L, FTla, BS1, BS2, TFLlb, CYP76C-1, CYP76C-2, and NF-YC4.46. The computer implemented method of any one of embodiments 38-45, wherein the plant or part thereof is a protoplast.47. The computer implemented method of any one of embodiments 1-46, comprising synthesizing a plurality of polynucleotide vectors, wherein each polynucleotide vector in the plurality of polynucleotides comprises the enhancer element, a different promoter, and a reporter gene;transforming each polynucleotide vectors into two or more plant tissues; measuring expression of the reporter gene in the two or more plant tissues, wherein expression of the reporter gene in the two or more plant tissues and with different promoters indicates validation of the enhancer element.48. The computer implemented method of any one of embodiments 38-47, wherein the reporter gene is luciferase.49. The computer implemented method of any one of embodiments 2-36, wherein experimentally validating the enhancer element comprises inserting the enhancer element within about 1 kb of a gene in a plant, or part thereof and measuring expression of the gene, wherein increased expression of the gene compared to endogenous expression of the gene indicates validation of the enhancer element.50. The computer implemented method of embodiment 49, wherein the plant or part thereof is a protoplast.51. The computer implemented method of any one of embodiments 2-36, wherein experimentally validating the enhancer element comprises inserting the enhancer element within about 1 kb of a gene in two or more plant tissues and measuring expression of the gene in the two or more plant tissues, wherein increased expression of the gene compared to 71MF-364940674Docket No.: 165362002340endogenous expression of the gene in at least two or the two or more plant tissues indicates validation of the enhancer element.52. The computer implemented method of any one of embodiments 49-51, wherein the enhancer element is inserted into a noncoding genomic region upstream of the gene.53. The computer implemented method of any one of embodiments 49-51, wherein the enhancer element is inserted into a noncoding genomic region downstream of the gene.54. The computer implemented method of any one of embodiments 49-51, wherein the enhancer element is inserted into a noncoding genomic region within the gene.55. The computer implemented method of any one of embodiments 49-54, comprising measuring endogenous expression of the gene without inserting the enhancer element.56. The computer implemented method of any one of embodiments 49-55, wherein measuring expression comprises performing qPCR or RNA-sequencing.57. A method for identifying an enhancer element in a plant, the method comprising:(a) obtaining a first candidate sequence;(b) generating a plurality of candidate sequences from the first candidate sequence using single base pair saturation mutagenesis;(c) for each promoter sequence in a plurality of promoter sequences, generating a plurality of input sequences, each comprising a candidate sequence from the plurality of candidate sequences and the promoter sequence;(d) inputting the plurality of input sequences into two or more machine learning models trained to predict a measurement of chromatin accessibility in one or more plant tissues from a sequence;(e) receiving, from the two or more machine learning models, a predicted chromatin accessibility measurement for each of the plurality of input sequences;(f) synthesizing a composite measurement for each candidate sequence related to the predicted chromatin accessibility measurements for each input sequence in the plurality of input sequences comprising the candidate sequence from each of the two or more machine learning models;72MF-364940674Docket No.: 165362002340(g) selecting a top performing sequence from the plurality of candidate sequences based on the composite measurement;(h) if the top performing sequence matches the first candidate sequence, identifying the top performing sequence as the enhancer element in a plant, and if the top performing sequence does not match the first candidate sequence, updating the first candidate sequence to be top performing sequence and repeating steps (b) - (g);(i) synthesizing a polynucleotide vector comprising the candidate enhancer sequence operably linked to an endogenous promoter, a transcription start site, and a reporter gene;(j) transforming a plant or part thereof with the polynucleotide vector;(k) transforming a second polynucleotide vector comprising, a control element operably linked to the endogenous promoter, the transcription start site, and the reporter gene into a second plant or part thereof; and(l) measuring the expression of the reporter gene in the first plant or part thereof and in the second plant or part thereof, wherein increased expression of the reporter gene in the first plant or part therefor compared to the second plant or part thereof indicates that the candidate sequence comprises an enhancer element.58. A system for identifying an enhancer element in a plant, the system comprising: one or more processors;a user input device anda memory communicatively coupled to the one or more processors and configured to store instructions that, when executed by the one or more processors, cause the system to:(a) receive a first candidate sequence from the user input device;(b) generate a plurality of candidate sequences from the first candidate sequence using single base pair saturation mutagenesis;(c) for each promoter sequence in a plurality of promoter sequences, generate a plurality of input sequences, each comprising a candidate sequence from the plurality of candidate sequences and the promoter sequence;(d) input the plurality of input sequences into two or more machine learning models trained to predict a measurement of chromatin accessibility in one or more plant tissues from a sequence;(e) receive, from the two or more machine learning models, a predicted chromatin accessibility measurement for each of the plurality of input sequences;73MF-364940674Docket No.: 165362002340(f) synthesize a composite measurement for each candidate sequence related to the predicted chromatin accessibility measurements for each input sequence in the plurality of input sequences comprising the candidate sequence from each of the two or more machine learning models;(g) select a top performing sequence from the plurality of candidate sequences based on the composite measurement;(h) if the top performing sequence matches the first candidate sequence, identify the top performing sequence as the enhancer element in the plant, and if the top performing sequence does not match the first candidate sequence, update the first candidate sequence to be top performing sequence and repeating steps (b) - (g).59. The system of embodiment 58, wherein the first candidate sequence comprises a polynucleotide sequence.60. The system of embodiment 58 or 59, wherein the first candidate sequence comprises a randomly generated polynucleotide sequence comprising a uniform distribution of adenine, guanine, cytosine, and thymine nucleotides.61. The system of any one of embodiments 58-60, wherein the first candidate sequence has a length of N nucleotides.62. The system of embodiment 61, wherein N is between 10 and 60.63. The system of embodiment 61 or 62, wherein the plurality of candidate sequences comprises 3 x IV + 1 candidate sequences.64. The system of any one of embodiments 58-63, wherein the first candidate sequence comprises a polynucleotide sequence comprising a known enhancer motif flanked by one or more variable nucleotides.65. The system of any one of embodiments 58-60 and 64, wherein the first candidate sequence comprises a polynucleotide sequence comprising a known enhancer motif repeated one or more times.74MF-364940674Docket No.: 16536200234066. The system of embodiment 65, wherein the known enhancer motif is a k-mer associated with a promoter of a plurality of highly expressed genes.67. The system of embodiment 66, wherein the k-mer comprises a polynucleotide sequence identified by:identifying a plurality of promoters associated with highly expressed genes in the plant; andidentifying a polynucleotide sequence enriched in sequences for the plurality of promoters.68. The system of embodiments 66 or 67, wherein the plurality of promoters have been identified as accessible in a plurality of cell types using snATAC-seq.69. The system of any one of embodiments 66-68, wherein the k-mer comprises a polynucleotide sequence less than about 12 nucleotides.70. The system of any one of embodiment 64- 68, wherein the known enhancer motif is held constant during the single base pair saturation mutagenesis.71. The system of any one of embodiments 64-70, wherein the plurality of candidate sequences comprises 3 x M + 1 candidate sequences, wherein M is the number of variable nucleotides in the first candidate sequence.72. The system of any one of embodiments 58-71, wherein single base pair saturation mutagenesis comprises separately substituting the nucleotide at each position in the first candidate sequence with an adenine, a thymine, a guanine, and a cytosine nucleotide.73. The system of any one of embodiments 58-72, wherein a promoter sequence in the plurality of promoter sequences is a known promoter sequence element in the genome of the plant.74. The system of any one of embodiments 58-73 , wherein each input sequence comprises the candidate sequence at the 5’ end of the promoter sequence.75MF-364940674Docket No.: 16536200234075. The system of any one of embodiments 58-74, wherein each input sequence comprises the candidate sequence repeated about 1, about 2, or about 3 times.76. The system of any one of embodiments 58-75, wherein the two or more machine learning models have each been trained to predict a measurement of chromatin accessibility a plant tissue based on ATAC-seq data collected from the plant tissue.77. The system of any one of embodiments 58-76, wherein the two or more machine learning models have been trained to predict chromatin accessibility in different plant tissues.78. The system of any one of embodiments 58-77, wherein the one or more plant tissues comprise plant tissues from a dicot.79. The system of any one of embodiment 58-78, wherein the one or more plant tissues comprise plant tissues from a dicot and the plant is a dicot.80. The system of any one of embodiments 58-79, wherein the one or more plant tissues comprise plant tissues from the same species as the plant.81. The system of any one of embodiments 58-80, wherein a machine learning model of the two or more machine learning models comprises a convolutional neural network (CNN).82. The system of embodiment 81, wherein the CNN is a dilated CNN.83. The system of any one of embodiments 58-82, wherein the one or more plant tissues are selected from a group consisting of bud, cotyledon, flower, flower bud, hypocotyl, leaf, pod, and root.84. The system of any one of embodiments 58-83, wherein synthesizing the composite measurement comprises averaging the predicted chromatin accessibility measurements for each input comprising the candidate sequence from one of the two or more machine learning models and averaging the predicted chromatin accessibility measurements for each candidate sequence across the two or more machine learning models.76MF-364940674Docket No.: 16536200234085. The system of any one of embodiments 58-84, wherein the composite measurement is based on the chromatin accessibility of the candidate sequence when the candidate sequence is linked to the plurality of promoters in two or more tissues.86. The system of any one of embodiments 58-85, wherein the selecting the top performing sequence comprises selecting the candidate sequence with the highest composite measurement among the composite measurements for the plurality of candidate sequences.87. The system of any one of embodiments 58-86, wherein the repeating steps (b)-(g) increases a probability the top performing sequence matches the first candidate sequence in a subsequent repetition of the steps.88. The system of any one of embodiments 58-87, comprising repeating steps (b)-(g) about 10 times.89. The system of embodiment 88, wherein the selecting the top performing sequence comprises selecting two or more candidate sequences with the highest composite measurement among the composite measurements for the plurality of candidate sequences and performing steps (b)-(g) with each of the two or more candidate sequences as the first candidate sequence.90. The system of embodiment 89, wherein the top performing sequence matches the candidate sequence if the nucleotide sequence of the top performing sequence is the same as the nucleotide sequence of the candidate sequence.91. The system of any of embodiments 58-90 , wherein the plant is soybean.92. The system of embodiment 91, wherein the ATAC-seq data comprises soybean data collected from Williams 82 soybeans.93. A method of training a machine learning model to predict a measurement of chromatin accessibility in a dicot tissue from a sequence, the method comprising:77MF-364940674Docket No.: 165362002340obtaining training data comprising ATAC-seq read coverage for the dicot tissue, wherein the ATAC-seq read coverage represents a measurement of chromatin accessibility in the tissue from the dicot;selecting a plurality of positive genomic windows, wherein a positive genomic window is a region of Wm82v4 with ATAC-seq read coverage in the training data;selecting a plurality of negative genomic windows, wherein a negative genomic window is region of Wm82v4 without ATAC-seq read coverage in the training data; and training the machine learning model to predict a measurement of chromatin accessibility from a sequence wherein the training is based on the ATAC-seq read coverage at the plurality of positive genomic windows, ATAC-seq read coverage at the plurality negative genomic windows, the sequence in Wm82v4 corresponding to each of the positive genomic windows, the sequence in Wm82v4 corresponding to each of the negative genomic windows.94. The method of embodiments 93, wherein the dicot tissue is a soybean tissue.95. The method of embodiments 93 or 94, wherein training the machine learning model comprises optimizing parameters of the machine learning model.96. A method of enhancing the expression level of a gene in a plant or part thereof, the method comprising:(a) identifying an enhancer element according to the computer implemented method of any one of embodiments 1-56; and(b) introducing the identified enhancer element into a plant or part thereof, wherein the enhancer element is operably linked to the gene, and wherein the enhancer element is inserted upstream of a transcription start site associated with the gene,wherein the expression level of the gene in the plant or part thereof comprising the enhancer element is enhanced as compared to the expression level of the gene in a control plant or part thereof lacking the enhancer element.97. A method of enhancing the expression level of a gene in a plant or part thereof, the method comprising:(a) obtaining a first candidate sequence;78MF-364940674Docket No.: 165362002340(b) generating a plurality of candidate sequences from the first candidate sequence using single base pair saturation mutagenesis;(c) for each promoter sequence in a plurality of promoter sequences, generating a plurality of input sequences, each comprising a candidate sequence from the plurality of candidate sequences and the promoter sequence;(d) inputting the plurality of input sequences into two or more machine learning models trained to predict a measurement of chromatin accessibility in one or more plant tissues from a sequence;(e) receiving, from the two or more machine learning models, a predicted chromatin accessibility measurement for each of the plurality of input sequences;(f) synthesizing a composite measurement for each candidate sequence related to the predicted chromatin accessibility measurements for each input sequence in the plurality of input sequences comprising the candidate sequence from each of the two or more machine learning models;(g) selecting a top performing sequence from the plurality of candidate sequences based on the composite measurement;(h) if the top performing sequence matches the first candidate sequence, selecting the top performing sequence as an enhancer element, and if the top performing sequence does not match the first candidate sequence, updating the first candidate sequence to be top performing sequence and repeating steps (b) - (g); and(i) introducing the enhancer element into a plant or part thereof, wherein the enhancer element is operably linked to the gene, and wherein the enhancer element is inserted upstream of a transcription start site associated with the gene,wherein the expression level of the gene in the plant or part thereof comprising the enhancer element is enhanced as compared to the expression level of the gene in a control plant or part thereof lacking the enhancer element.EXAMPLESExample 1: Identifying a candidate enhancer sequence using a predictive model based on a random motif input

[0224] In this Example, a top candidate enhancer sequence was identified using a randomly generated motif as an initial input sequence for a predictive model.79MF-364940674Docket No.: 165362002340

[0225] To investigate the activity of candidate enhancer sequences that might upregulate gene expression, candidate enhancer sequences (“Enhancer element”) were computationally inserted three times in tandem in a promoter region upstream of multiple genes (FIG. 1A).The insertion of the enhancer element was predicted to correspond to an increase in expression of the gene(s). Using chromatin accessibility data generated from ATAC-seq collected from eight soybean tissues and mapped to the Wm82v4 genome, predictive models were trained to predict chromatin accessibility for novel sequences (FIG. IB). As described herein, the models were used to investigate candidate enhancer sequences and their activity across promoter regions in a plant genome.

[0226] As an initial candidate enhancer sequence, a 12 base pair (bp) long DNA motif was randomly generated with a uniform distribution of adenine, cytosine, guanine, and thymine nucleotides (FIG. 2A, Enhancer 98, Sequence Table). Then, the 12 bp motif was subjected to single base pair saturation mutagenesis, generating 37 candidate enhancer sequences. Single base pair saturation mutagenesis generated a total of 3 x N + 1 sequences based on an initial candidate sequence (where N is the number of base pairs in the initial candidate sequence) by substituting the nucleotide at each position in the initial candidate sequence with an adenine, cytosine, guanine, and thymine nucleotide.

[0227] Next, multiple promoter contexts were tested for each candidate enhancer sequence by inserting the sequences into different promoters to assess their chromatin accessibility (and thus, activity) in different gene expression environments. To generate a series of input sequences for use in a predictive model of chromatin accessibility, each of the 37 candidate enhancer sequences was inserted three times in tandem into 5,438 promoter regions in the soybean Wm82v4 genome (FIG. 2A). Each three-copy set of enhancer sequences was inserted 150 bp upstream of the transcription start sites across the soybean genome corresponding to the 5,438 promoter regions. This generated a set of 201,206 input sequences, each containing three copies of a candidate enhancer sequence and a promoter region.

[0228] The trained machine learning models were then used to generate chromatin accessibility predictions for each of the 201,206 input sequences (FIG. 2A). The models were dilated convolutional neural networks (CNNs) trained on eight soybean tissues to predict measurements of chromatin accessibility within regions of the soybean genome, based on ATAC-seq data collected from soybean Wm82v4 tissues. Based on a consensus of the80MF-364940674Docket No.: 165362002340predictions from the models, for each of the 201,206 input sequences, a unique score was generated to represent a predicted chromatin accessibility measurement.

[0229] For each candidate enhancer sequence, a composite measurement was calculated based on the average of the predicted chromatin accessibility measurements for all of the input sequences associated with that candidate enhancer sequence (i.e., for all 5,438 promoter regions combined with that candidate enhancer sequence) (FIG. 2A, right panel). This composite measurement per candidate enhancer sequence represents the candidate’s predicted chromatin accessibility across multiple soybean promoters.

[0230] These composite measurements were then used for greedy selection, wherein the candidate enhancer sequence with the highest composite measurement was selected as the top candidate enhancer sequence of the iteration (FIG. 2B). The top candidate enhancer sequence was then used to restart a second iteration of the steps above. The top sequence was used as the initial candidate enhancer sequence subject to single base pair saturation mutagenesis, and the steps of FIG. 2A were repeated. This iterative process was repeated until no candidate enhancer sequence could be identified that outperformed the initial candidate enhancer sequence used for steps of FIG. 2A in that iteration. The process was ended at the point of convergence, where no single base pair substitution was able to improve the score of a candidate enhancer sequence under investigation. One top candidate enhancer sequence (Enhancer 100, SEQ ID NO: 100) was identified to converge after 3 iterations and was selected for subsequent validation experiments (FIG. 2B, marked with a star).Example 2: Generating a set of candidate enhancer sequences using a predictive model based on the G-box motif

[0231] In this Example, 17 candidate enhancer sequences were identified using a G-box motif anchor as an initial input sequence for a predictive model.

[0232] The G-box motif is a regulatory element that enhances gene expression in many plant genomes (see Ishige et al., A G-box motif (GCCACGTGCC) tetramer confers high-level constitutive expression in dicot and monocot plants, 18 The Plant Journal 443-448 (1999)). The motif comprises the sequence CACGTG and may be flanked by 2 base pair (bp) long flanking regions on either side (FIG. 3A). Beginning with one reference G-box enhancer sequence, single base pair saturation mutagenesis was used to generate 256 variations of the G-box motif and 2 bp flanking regions, for a total of 257 candidate enhancer sequences (FIG.81MF-364940674Docket No.: 1653620023403B). To generate the 256 variations, the G-box motif was held constant and every combination of adenine, cytosine, guanine, and thymine was substituted into the 2 bp flanking regions.

[0233] Next, multiple promoter contexts were tested for each candidate enhancer sequence by inserting the sequences into different promoters to assess their chromatin accessibility (and thus, activity) in different gene expression environments. To generate a series of input sequences for use in a predictive model of chromatin accessibility, each of the 257 candidate enhancer sequences was inserted four times in tandem into 5,438 promoter regions in the soybean Wm82v4 genome (FIG. 3C). This generated a set of 1,397,566 input sequences, each containing four copies of a candidate enhancer sequence and a promoter region.

[0234] The trained machine learning models were then used to generate chromatin accessibility predictions for each of the 1,397,566 input sequences (FIG. 3C). The models were dilated CNNs trained on eight soybean tissues to predict measurements of chromatin accessibility within regions of the soybean genome, based on ATAC-seq data collected from soybean Wm82v4 tissues. Based on a consensus of the predictions from the models, for each of the 1,397,566 input sequences, a unique score was generated to represent a predicted chromatin accessibility measurement.

[0235] The predicted chromatin accessibility measurements were analyzed by rank distribution for each candidate enhancer sequence, considering that the input sequences, in groups of 5,438, represented all 257 candidate enhancer sequences. The predicted chromatin accessibility measurement for each of the 5,438 promoters was examined per candidate enhancer sequence (FIGS. 3D-3F). This data was rank ordered based on the average predicted chromatin accessibility measurement across all 5,438 promoters per candidate enhancer sequence. The candidate enhancer sequences with the highest predicted chromatin accessibility measurements across all 5,438 promoters are towards the right side of the plot in FIG. 3D and FIG. 3F.Example 3: Validating the activity of candidate enhancer sequences in polynucleotide vectors using dual luciferase assays

[0236] To validate the activity of candidate enhancer sequences, a series of polynucleotide vectors were generated for use in dual luciferase assays.82MF-364940674Docket No.: 165362002340

[0237] All dual luciferase vectors contained two cassettes: firefly luciferase (FLUC) for testing experimental conditions and Renilla luciferase (RLUC) for transfection efficiency control (FIG. 4A). Known or candidate enhancer sequences were evaluated in dual luciferase assays by measuring the increase in gene expression of FLUC, measured by the ratio of luminescence signal from FLUC over RLUC, The dual luciferase vectors were used to test known enhancer sequences (“3x OCS”, “35S enhancer”) and candidate enhancer sequences (“Experimental enhancer”) (FIG. 4B).

[0238] A low expressing promoter, TFLlb, was amplified from the soybean genome and cloned upstream of the FLUC gene, followed by the CaMV 35S terminator. The RLUC cassette was cloned with the CaMV 35S promoter, RLUC gene, and CaMV 35S terminator. The dual luciferase vector was then transfected into soy epicotyl protoplasts and the experiment was run using the Dual-Glo kit. The assay readout was FLUC / RLUC luminescence, which measures changes in FLUC expression based on the experimental condition while normalizing to RLUC expression to account for sample-to-sample variation.

[0239] To screen enhancer candidates, a LacZ cassette was cloned via Gibson assembly into the TFLlb promoter sequence of the dual luciferase vector 75 bp upstream of the FLUC transcription start site. The LacZ cassette enables blue-white screening to identify vectors with correctly inserted enhancers. Known enhancers and candidate enhancers were synthesized with 4 base pair overhangs and ligated into the screening vector digested with Esp3I. The vectors were transfected into soy epicotyl protoplasts to read the FLUC / RLUC luminescence. Candidate enhancers were validated when the FLUC / RLUC ratio was increased compared to the TFLlb control (wildtype) vector. The following vectors were tested in an initial experiment to validate the dual luciferase assays:83MF-364940674Docket No.: 165362002340

[0240] The insertion of the OCS enhancer three times in tandem (FIG. 5, “3x OCS”) into the promoter region upstream of FLUC produced a 28-fold increase in luminescence, indicating a high degree of enhancer activity. The insertion of the CaMV 35S enhancer (FIG.5, “35S Enhancer”) into the promoter region upstream of FLUC produced a 59-fold increase in luminescence, indicating a high degree of enhancer activity.

[0241] Candidate enhancers were also tested in dual luciferase vectors. A vector (SEQ ID NO: 124, comprising SEQ ID NOs: 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, FIG. 12A) containing three copies of Enhancer 100 (single copy SEQ ID NO: 100; three copies SEQ ID NO: 138) increased FLUC / RLUC luminescence over a control vector containing the wild-type TFLlb promoter, and increased FLUC / RLUC luminescence over a control vector containing three copies of the OCS element (FIG. 6). A vector (SEQ ID NO: 126, comprising SEQ ID NOs: 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, FIG.12B) containing three copies of Enhancer 115 (single copy SEQ ID NO: 115; three copies SEQ ID NO: 140) increased FLUC / RLUC luminescence over a control vector containing the wild-type TFLlb promoter and increased FLUC / RLUC luminescence over a control vector containing three copies of the OCS element (FIGS. 7A-7B). A vector (SEQ ID NO: 125, comprising SEQ ID NOs: 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, FIG. 12C) containing three copies of Enhancer 107 (single copy SEQ ID NO: 107; three copies SEQ ID NO: 139) increased FLUC / RLUC luminescence over a control vector containing the wildtype TFLlb promoter and increased FLUC / RLUC luminescence over a control vector containing three copies of the OCS element (FIGS. 7A-7B).

[0242] Enhancers were also tested in dual luciferase vectors containing promoters from control genes that were randomly selected based on their expression levels in the protoplast assay system and specifically, expression as determined by RNA sequencing data taken at various timepoints during the protoplast lifespan (FIGS. 8A-8B). The vectors were constructed following the steps described above, but with a promoter from a mediumexpression control gene (a sample gene known to exhibit medium expression levels within the protoplast assay system) and a promoter from a high-expression control gene (a sample gene known to exhibit high expression levels within the protoplast assay system).84MF-364940674Docket No.: 165362002340

[0243] Dual luciferase assays, as described above, were conducted for the following constructs: a vector containing the wild-type TFLlb promoter (“TFLlb - WT”), a vector containing three copies of the OCS enhancer and the TFLlb promoter (“TFLlb - 3x OCS”, SEQ ID NO: 163), a vector containing the wild-type promoter of the medium-expression control gene (“MID - WT”, SEQ ID NO: 165), a vector containing three copies of the OCS enhancer and the promoter of the medium-expression control gene (“MID - 3x OCS”, SEQ ID NO: 166), a vector containing the wild-type promoter of the high-expression control gene (“HIGH - WT”, SEQ ID NO: 167), and a vector containing three copies of the OCS enhancer and the promoter of the high-expression control gene (“HIGH - 3x OCS”, SEQ ID NO: 168).

[0244] The dual luciferase assays were performed with vectors at 55 ng / pL (FIG. 8A) or 110 ng / pL (FIG. 8B). Irrespective of the promoter (TFLlb, medium-expression control, high-expression control) or vector concentration (55 ng / pL, 110 ng / pL), the vectors containing three copies of the OCS enhancer consistently increased FLUC / RLUC luminescence over control vectors containing only the wild- type promoters.

[0245] Next, the ability of Enhancer 100 (SEQ ID NO: 100) to improve expression of different promoters was tested. Enhancer 100 was tested in three copies with no linker between each copy (SEQ ID NO: 138, “3x_12bp_2666”). Dual luciferase assays, as described above, were conducted for the following constructs: a vector containing the wildtype TFLlb promoter (“TFLlb - WT”), a vector containing three copies of the OCS enhancer and the TFLlb promoter (“TFLlb - 3x OCS”), a vector containing 3x_12bp_2666 and the TFLlb promoter (“TFLlb - 3x_12bp_2666”, SEQ ID NO: 164), a vector containing the wild-type promoter of the high-expression control gene (“HIGH - WT”), a vector containing three copies of the OCS enhancer and the promoter of the high-expression control gene (“HIGH - 3x OCS”), and a vector containing 3x_12bp_2666 and the promoter of the high-expression control gene (“HIGH - 3x_12bp_2666”).

[0246] The dual luciferase assays were performed with these vectors at 110 ng / pL (FIG. 13). The results of these experiments surprisingly showed that irrespective of the promoter (TFLlb or high expression promoter control) the vectors containing three copies of Enhancer 100 (SEQ ID NO: 138) consistently increased FLUC / RLUC luminescence over control vectors containing only the wild- type promoters.

[0247] To investigate the copy number effect of enhancer elements, Enhancer 100 (SEQ ID NO: 100) was repeated in different enhancer constructs from 2x to 6x (i.e., in two copies,85MF-364940674Docket No.: 165362002340three copies, four copies, five copies, or six copies) without linker or spacer sequencers present between each copy (2x_12bp_2666 (SEQ ID NO: 149), 3x_12bp_2666 (SEQ ID NO: 138), 4x_12bp_2666 (SEQ ID NO: 150), 5x_12bp_2666 (SEQ ID NO: 151), 6x_12bp_2666 (SEQ ID NO: 152)). The ability of the different enhancer constructs to enhance the TFL1B promoter was tested in the dual luciferase assay as described above.

[0248] The dual luciferase assays were performed with the vectors comprising Enhancer 100 in different copy numbers and the TFL1B promoter at 110 ng / pL and results are displayed in FIG. 14. The results illustrated a dependency on copy number of Enhancer 100 (i.e., increased copy number was associated with an increase in expression), with a plateau reached between five and six copies of the enhancer.

[0249] Next, the distance of the enhancers to the transcription start site (TSS) was investigated further. The construct with three copies of Enhancer 100 without spacers (3x_12bp_2666; SEQ ID NO: 138) and 3xOCS were cloned at various distances from the TSS of the luciferase reporter gene of the dual luciferase assay as described herein.

[0250] The enhancers were cloned at 75 base pairs, 100 base pairs, 150 base pairs, 225 base pairs and 350 base pairs from the TSS, using the TFL1B promoter. The dual luciferase assay was run as described above with 110 ng / pL of each vector. FIG. 15 illustrates the dependence on distance from the TSS of the enhancer element (i.e., expression was most enhanced at 75 base pairs from the TSS). This was observed for both the OCS element and Enhancer 100.Example 4: Genomic insertion of enhancer elements

[0251] The efficiency of the various enhancer elements as described herein is tested using the soybean transient transgenic hairy root system as is performed essentially as described in the literature (Toth et al., Curr Protoc Plant Biol 2016 May; 1(1): 1-13. Doi:10.1002 / cppb.20017; Song et al., Curr Protoc 2021 July; l(7):el95. doi: 10.1002 / cpzl.l95). Specifically, the ability of Enhancer 100 (SEQ ID NO: 100) to improve expression of different promoters is tested using the hairy root system. Therefore, the Enhancer 100 is tested in three copies with no linker between each copy (SEQ ID NO: 138, “3x_12bp_2666”) and is inserted approximately 50 base pairs upstream of a gene of interest and its effect on gene expression of the gene of interest is evaluated. In a separate experiment the enhancer element according to SEQ ID NO: 144 is inserted approximately 50 base pairs upstream of a86MF-364940674Docket No.: 165362002340gene of interest and its effect on gene expression of the gene of interest is evaluated in the soybean transgenic hairy root system.

[0252] Various other methods exist in the art for the insertion of DNA elements such as enhancer elements into the genome of a plant such as a soy plant, maize plant or wheat plant. Examples of obtaining a stably transformed plant by precision editing an insertion of a enhancer element may be found in PCT / US2020 / 039410 “Improved homology dependent repair genome editing”, US 17 / 456465 “Genetic regulatory element”, Chen et al., An update on precision genome editing by homology-directed repair in plants. Plant Physiol. 2022 Mar 28;188(4):1780-1794, Dong et al., Targeted DNA insertion in plants, Proc. Natl. Acad. Sci. U.S.A. 118 (22) e2004834117 (2021), Zhongsen et al., Cas9-Guide RNA Directed Genome Editing in Soybean, Plant Physiology, 169(2): 960-970 (2015), and Claeys et al., Coordinated gene regulation in maize though CRISPR / Cas-mediated enhancer insertion, Plant Biotechnol J. 22(1):16-18 (2023).SEQUENCES87MF-364940674Docket No.: 16536200234088 MF-364940674Docket No.: 16536200234089MF-364940674Docket No.: 16536200234090MF-364940674Docket No.: 16536200234091MF-364940674Docket No.: 16536200234092MF-364940674Docket No.: 16536200234093MF-364940674Docket No.: 16536200234094MF-364940674Docket No.: 16536200234095MF-364940674Docket No.: 16536200234096MF-364940674Docket No.: 16536200234097MF-364940674Docket No.: 16536200234098MF-364940674

Claims

Docket No.: 165362002340CLAIMS1. A computer implemented method for identifying an enhancer element in a plant, the method comprising:by one or more computing devices comprising one or more processors and memory:(a) receiving a first candidate sequence from the memory;(b) generating a plurality of candidate sequences from the first candidate sequence using single base pair saturation mutagenesis;(c) for each promoter sequence in a plurality of promoter sequences, generating a plurality of input sequences, each comprising a candidate sequence from the plurality of candidate sequences and the promoter sequence;(d) inputting the plurality of input sequences into two or more machine learning models trained to predict a measurement of chromatin accessibility in one or more different plant tissues from a sequence;(e) receiving, from the two or more machine learning models, a predicted chromatin accessibility measurement for each of the plurality of input sequences;(f) synthesizing a composite measurement for each candidate sequence related to the predicted chromatin accessibility measurement for each input sequence in the plurality of input sequences comprising the candidate sequence from each of the two or more machine learning models;(g) selecting a top performing sequence from the plurality of candidate sequences based on the composite measurement; and(h) if the top performing sequence matches the first candidate sequence, identifying the top performing sequence as the enhancer element in the plant, and if the top performing sequence does not match the first candidate sequence, updating the first candidate sequence to be top performing sequence and repeating steps (b) - (g).

2. The computer implemented method of claim 1, further comprising experimentally validating the enhancer element increases expression of one or more genes in the plant.

3. The computer implemented method of claim 1, wherein the first candidate sequence comprises a polynucleotide sequence.99MF-364940674Docket No.: 1653620023404. The computer implemented method of claim 1, wherein the first candidate sequence comprises a randomly generated polynucleotide sequence comprising a uniform distribution of adenine, guanine, cytosine, and thymine nucleotides.

5. The computer implemented method of claim 1, wherein the first candidate sequence has a length of N nucleotides.

6. The computer implemented method of claim 5, wherein N is between 10 and 60.

7. The computer implemented method of claim 5, wherein the plurality of candidate sequences comprises 3 x IV + 1 candidate sequences.

8. The computer implemented method of claim 1, wherein the first candidate sequence comprises a polynucleotide sequence comprising a known enhancer motif flanked by one or more variable nucleotides.

9. The computer implemented method of claim 1, wherein the first candidate sequence comprises a polynucleotide sequence comprising a known enhancer motif repeated one or more times.

10. The computer implemented method of claim 8, wherein the known enhancer motif is a k-mer associated with a promoter of a plurality of highly expressed genes.

11. The computer implemented method of claim 10, wherein the k-mer comprises a polynucleotide sequence identified by:identifying a plurality of promoters associated with highly expressed genes in the plant; andidentifying a polynucleotide sequence enriched in sequences for the plurality of promoters.

12. The computer implemented method of claim 10, wherein the plurality of promoters have been identified as accessible in a plurality of cell types using snATAC-seq.100MF-364940674Docket No.: 16536200234013. The computer implemented method of claim 10, wherein the k-mer comprises a polynucleotide sequence less than about 12 nucleotides.

14. The computer implemented method of claim 8, wherein the known enhancer motif is held constant during the single base pair saturation mutagenesis.

15. The computer implemented method of claim 8, wherein the plurality of candidate sequences comprises 3 x M + 1 candidate sequences, wherein M is the number of variable nucleotides in the first candidate sequence.

16. The computer implemented method of claim 1, wherein single base pair saturation mutagenesis comprises separately substituting the nucleotide at each position in the first candidate sequence with an adenine, a thymine, a guanine, and a cytosine nucleotide.

17. The computer implemented method of claim 1, wherein a promoter sequence in the plurality of promoter sequences is a known promoter sequence element in the genome of the plant.

18. The computer implemented method of claim 1, wherein each input sequence comprises the candidate sequence at the 5’ end of the promoter sequence.

19. The computer implemented method of claim 1, wherein each input sequence comprises the candidate sequence repeated about 1, about 2, or about 3 times.

20. The computer implemented method of claim 1, wherein the two or more machine learning models have each been trained to predict a measurement of chromatin accessibility a plant tissue based on ATAC-seq data collected from the plant tissue.

21. The computer implemented method of claim 1, wherein the two or more machine learning models have been trained to predict chromatin accessibility in different plant tissues.

22. The computer implemented methods of claim 1, wherein the one or more plant tissues comprise plant tissues from a dicot.101MF-364940674Docket No.: 16536200234023. The computer implemented methods of claim 1, wherein the one or more plant tissues comprise plant tissues from a dicot and the plant is a dicot.

24. The computer implemented methods of claim 1, wherein the one or more plant tissues comprise plant tissues from the same species as the plant.

25. The computer implemented method of claim 1, wherein a machine learning model of the two or more machine learning models comprises a convolutional neural network (CNN).

26. The computer implemented method of claim 25, wherein the CNN is a dilated CNN.

27. The computer implemented method of claim 1, wherein the one or more plant tissues are selected from a group consisting of bud, cotyledon, flower, flower bud, hypocotyl, leaf, pod, and root.

28. The computer implemented method of claim 1, wherein synthesizing the composite measurement comprises averaging the predicted chromatin accessibility measurements for each input comprising the candidate sequence from one of the two or more machine learning models and averaging the predicted chromatin accessibility measurements for each candidate sequence across the two or more machine learning models.

29. The computer implemented method of claim 1, wherein the composite measurement is based on the chromatin accessibility of the candidate sequence when the candidate sequence is linked to the plurality of promoters in two or more tissues.

30. The computer implemented method of claim 1, wherein the selecting the top performing sequence comprises selecting the candidate sequence with the highest composite measurement among the composite measurements for the plurality of candidate sequences.

31. The computer implemented method of claim 1, wherein the repeating steps (b)-(g) increases a probability the top performing sequence matches the first candidate sequence in a subsequent repetition of the steps.102MF-364940674Docket No.: 16536200234032. The computer implemented method of claim 1, comprising repeating steps (b)-(g) about 10 times.

33. The computer implemented method of claim 16, wherein the selecting the top performing sequence comprises selecting two or more candidate sequences with the highest composite measurement among the composite measurements for the plurality of candidate sequences and performing steps (b)-(g) with each of the two or more candidate sequences as the first candidate sequence.

34. The computer implemented method of claim 1, wherein the top performing sequence matches the candidate sequence if the nucleotide sequence of the top performing sequence is the same as the nucleotide sequence of the candidate sequence.

35. The computer implemented method of claim 1, wherein the plant is soybean.

36. The computer implemented method of claim 35, wherein the ATAC-seq data comprises soybean data collected from Williams 82 soybeans.

37. The computer implemented method of claim 2, wherein experimentally validating the enhancer element comprises a reporter assay.

38. The computer implemented method of claim 37, wherein the reporter assay comprises transforming a polynucleotide vector comprising, the enhancer element operably linked to an endogenous promoter, a transcription start site, and a reporter gene into a plant or part thereof, wherein the reporter gene is a luciferase gene, and testing for expression of the luciferase gene.

39. The computer implemented method of claim 38, wherein the polynucleotide vector comprises the enhancer element repeated three times.

40. The computer implemented method of claim 37, wherein the reporter assay comprises:103MF-364940674Docket No.: 165362002340transforming a first polynucleotide vector comprising, the enhancer element operably linked to an endogenous promoter a transcription start site, and a reporter gene into a first plant or part thereof from the plant;transforming a second polynucleotide vector comprising, a control sequence element operably linked to the endogenous promoter, the transcription start site, and the reporter gene into a second plant or part thereof from the plant; andmeasuring the expression of the reporter gene in the first plant or part thereof and in the second plant or part thereof, wherein increased expression of the reporter gene in the first plant or part thereof compared to the second plant or part thereof indicates validation of the enhancer element.

41. The computer implemented method of claim 40, wherein the first polynucleotide vector comprises the enhancer element repeated three times.

42. The computer implemented method of claim 41, wherein the control sequence element comprises an octopine synthase (OCS) enhancer element.

43. The computer implemented method of claim 41, wherein the control sequence element comprises a G-box element.

44. The computer implemented method of claim 38, wherein the endogenous promoter is promoter for a target gene.

45. The computer implemented method of claim 44, wherein the target gene is a gene selected from a group consisting of AlPlOa, AlPlOb, AML4, CRN, HB-1, RIC1, RIC2, RPF1, JAG1, JAG2, KHZ1, PP2C, TCP5-L, FTla, BS1, BS2, TFLlb, CYP76C-1, CYP76C-2, and NF-YC4.

46. The computer implemented method of claim 38, wherein the plant or part thereof is a protoplast.

47. The computer implemented method of claim 1, comprising104MF-364940674Docket No.: 165362002340synthesizing a plurality of polynucleotide vectors, wherein each polynucleotide vector in the plurality of polynucleotides comprises the enhancer element, a different promoter, and a reporter gene;transforming each polynucleotide vectors into two or more plant tissues; measuring expression of the reporter gene in the two or more plant tissues, wherein expression of the reporter gene in the two or more plant tissues and with different promoters indicates validation of the enhancer element.

48. The computer implemented method of claim 38, wherein the reporter gene is luciferase.

49. The computer implemented method of claim 2, wherein experimentally validating the enhancer element comprises inserting the enhancer element within about 1 kb of a gene in a plant, or part thereof and measuring expression of the gene, wherein increased expression of the gene compared to endogenous expression of the gene indicates validation of the enhancer element.

50. The computer implemented method of claim 49, wherein the plant or part thereof is a protoplast.

51. The computer implemented method of claim 2, wherein experimentally validating the enhancer element comprises inserting the enhancer element within about 1 kb of a gene in two or more plant tissues and measuring expression of the gene in the two or more plant tissues, wherein increased expression of the gene compared to endogenous expression of the gene in at least two or the two or more plant tissues indicates validation of the enhancer element.

52. The computer implemented method of claim 49, wherein the enhancer element is inserted into a noncoding genomic region upstream of the gene.

53. The computer implemented method of claim 49, wherein the enhancer element is inserted into a noncoding genomic region downstream of the gene.105MF-364940674Docket No.: 16536200234054. The computer implemented method of claim 49, wherein the enhancer element is inserted into a noncoding genomic region within the gene.

55. The computer implemented method of claim 49, comprising measuring endogenous expression of the gene without inserting the enhancer element.

56. The computer implemented method of claim 49, wherein measuring expression comprises performing qPCR or RN A- sequencing.

57. A method for identifying an enhancer element in a plant, the method comprising:(a) obtaining a first candidate sequence;(b) generating a plurality of candidate sequences from the first candidate sequence using single base pair saturation mutagenesis;(c) for each promoter sequence in a plurality of promoter sequences, generating a plurality of input sequences, each comprising a candidate sequence from the plurality of candidate sequences and the promoter sequence;(d) inputting the plurality of input sequences into two or more machine learning models trained to predict a measurement of chromatin accessibility in one or more plant tissues from a sequence;(e) receiving, from the two or more machine learning models, a predicted chromatin accessibility measurement for each of the plurality of input sequences;(f) synthesizing a composite measurement for each candidate sequence related to the predicted chromatin accessibility measurements for each input sequence in the plurality of input sequences comprising the candidate sequence from each of the two or more machine learning models;(g) selecting a top performing sequence from the plurality of candidate sequences based on the composite measurement;(h) if the top performing sequence matches the first candidate sequence, identifying the top performing sequence as the enhancer element in a plant, and if the top performing sequence does not match the first candidate sequence, updating the first candidate sequence to be top performing sequence and repeating steps (b) - (g);(i) synthesizing a polynucleotide vector comprising the candidate enhancer sequence operably linked to an endogenous promoter, a transcription start site, and a reporter gene;106MF-364940674Docket No.: 165362002340(j) transforming a plant or part thereof with the polynucleotide vector;(k) transforming a second polynucleotide vector comprising, a control element operably linked to the endogenous promoter, the transcription start site, and the reporter gene into a second plant or part thereof; and(l) measuring the expression of the reporter gene in the first plant or part thereof and in the second plant or part thereof, wherein increased expression of the reporter gene in the first plant or part therefor compared to the second plant or part thereof indicates that the candidate sequence comprises an enhancer element.

58. A system for identifying an enhancer element in a plant, the system comprising: one or more processors;a user input device anda memory communicatively coupled to the one or more processors and configured to store instructions that, when executed by the one or more processors, cause the system to:(a) receive a first candidate sequence from the user input device;(b) generate a plurality of candidate sequences from the first candidate sequence using single base pair saturation mutagenesis;(c) for each promoter sequence in a plurality of promoter sequences, generate a plurality of input sequences, each comprising a candidate sequence from the plurality of candidate sequences and the promoter sequence;(d) input the plurality of input sequences into two or more machine learning models trained to predict a measurement of chromatin accessibility in one or more plant tissues from a sequence;(e) receive, from the two or more machine learning models, a predicted chromatin accessibility measurement for each of the plurality of input sequences;(f) synthesize a composite measurement for each candidate sequence related to the predicted chromatin accessibility measurements for each input sequence in the plurality of input sequences comprising the candidate sequence from each of the two or more machine learning models;(g) select a top performing sequence from the plurality of candidate sequences based on the composite measurement;(h) if the top performing sequence matches the first candidate sequence, identify the top performing sequence as the enhancer element in the plant, and if the107MF-364940674Docket No.: 165362002340top performing sequence does not match the first candidate sequence, update the first candidate sequence to be top performing sequence and repeating steps (b) - (g).

59. A method of training a machine learning model to predict a measurement of chromatin accessibility in a dicot tissue from a sequence, the method comprising:obtaining training data comprising ATAC-seq read coverage for the dicot tissue, wherein the ATAC-seq read coverage represents a measurement of chromatin accessibility in the tissue from the dicot;selecting a plurality of positive genomic windows, wherein a positive genomic window is a region of Wm82v4 with ATAC-seq read coverage in the training data;selecting a plurality of negative genomic windows, wherein a negative genomic window is region of Wm82v4 without ATAC-seq read coverage in the training data; and training the machine learning model to predict a measurement of chromatin accessibility from a sequence wherein the training is based on the ATAC-seq read coverage at the plurality of positive genomic windows, ATAC-seq read coverage at the plurality negative genomic windows, the sequence in Wm82v4 corresponding to each of the positive genomic windows, the sequence in Wm82v4 corresponding to each of the negative genomic windows.

60. The method of claim 59, wherein the dicot tissue is a soybean tissue.

61. The method of claim 59 or 60, wherein training the machine learning model comprises optimizing parameters of the machine learning model.

62. A method of enhancing the expression level of a gene in a plant or part thereof, the method comprising:(a) identifying an enhancer element according to the computer implemented method of claim 1 ; and(b) introducing the identified enhancer element into a plant or part thereof, wherein the enhancer element is operably linked to the gene, and wherein the enhancer element is inserted upstream of a transcription start site associated with the gene,wherein the expression level of the gene in the plant or part thereof comprising the enhancer element is enhanced as compared to the expression level of the gene in a control plant or part thereof lacking the enhancer element.108MF-364940674Docket No.: 16536200234063. A method of enhancing the expression level of a gene in a plant or part thereof, the method comprising:(a) obtaining a first candidate sequence;(b) generating a plurality of candidate sequences from the first candidate sequence using single base pair saturation mutagenesis;(c) for each promoter sequence in a plurality of promoter sequences, generating a plurality of input sequences, each comprising a candidate sequence from the plurality of candidate sequences and the promoter sequence;(d) inputting the plurality of input sequences into two or more machine learning models trained to predict a measurement of chromatin accessibility in one or more plant tissues from a sequence;(e) receiving, from the two or more machine learning models, a predicted chromatin accessibility measurement for each of the plurality of input sequences;(f) synthesizing a composite measurement for each candidate sequence related to the predicted chromatin accessibility measurements for each input sequence in the plurality of input sequences comprising the candidate sequence from each of the two or more machine learning models;(g) selecting a top performing sequence from the plurality of candidate sequences based on the composite measurement;(h) if the top performing sequence matches the first candidate sequence, selecting the top performing sequence as an enhancer element, and if the top performing sequence does not match the first candidate sequence, updating the first candidate sequence to be top performing sequence and repeating steps (b) - (g); and(i) introducing the enhancer element into a plant or part thereof, wherein the enhancer element is operably linked to the gene, and wherein the enhancer element is inserted upstream of a transcription start site associated with the gene,wherein the expression level of the gene in the plant or part thereof comprising the enhancer element is enhanced as compared to the expression level of the gene in a control plant or part thereof lacking the enhancer element.109MF-364940674

Images