Model-based prokaryotic transcriptional regulation method and system

Abstract

The present invention relates to the field of synthetic biology, and in particular to a model-based prokaryotic transcriptional regulation method and system. The method comprises: acquiring DNA element information required by each transcriptional regulation system in a corresponding test strain, and constructing transcriptional regulation systems; establishing biophysical quantitative models describing the transcriptional regulation systems; acquiring valid test data used for fitting the biophysical quantitative models; and using the valid test data to train the biophysical quantitative models. The present invention characterizes the plurality of transcriptional regulation systems in the test strain, identifies key factors affecting the performance of these regulation systems under different evaluation criteria, and performs system optimization accordingly, thus improving system gene expression and induction fold change, and solving the problems of difficult engineering, long time consumption and limited room for optimization in prokaryotic transcriptional regulation systems.

Description

A model-based prokaryotic transcriptional regulation method and system Technical Field

[0001] This invention relates to the field of synthetic biology, and more specifically, to a model-based method and system for prokaryotic transcriptional regulation. Background Technology

[0002] In synthetic biology, the bottom-up design paradigm has been widely used to develop complex and diverse gene circuits. Its basic approach involves assembling modular gene elements into gene devices, which are then further integrated into higher-level systems. To rationally implement this assembly process, researchers must fully characterize various components at each level of abstraction (element, device, system) to ensure they can perform their intended functions in a controllable manner in the final product. At the gene element level, researchers need to design the required gene expression regulatory elements, such as promoters, transcription factors (TFs), ribosome binding sites (RBSs), and terminators, according to specific "design specifications"—a set of requirements for element function and performance indicators. These requirements are typically quantified and can be expressed as gene expression level functions. Because the interactions between promoters, transcription factors (TFs), and transcription machinery (RNA polymerase (RNAP) and its cofactors in the host cell) can influence gene expression levels early in the gene expression process (i.e., the transcription initiation stage), these three elements naturally become crucial for controlling local gene expression levels. Consequently, they have attracted the attention of numerous gene element modification methods, such as generating new promoter sequences through base mutations and regulating the spatial structure of protein elements (TFs, RNAPs, and cofactors) through amino acid residue mutations. However, despite the vast variable space of DNA and polypeptide sequences, functional promoters and protein variants are relatively scarce. Therefore, after obtaining mutant libraries, extensive screening or directed evolution is required, significantly increasing the time and material costs of the design process. Furthermore, in directed evolution experiments targeting mutants, the selection of mutation-sensitive regions mainly relies on manual review and evaluation, making this process inefficient and difficult to integrate into automated analysis and testing workflows.

[0003] Among existing approaches to engineering prokaryotic transcriptional regulatory systems, model-based design methods for these systems offer promise in alleviating the burden of constructing and characterizing large-scale libraries. Specifically, the model-based design process includes:

[0004] (1) Establish a quantitative model to describe gene transcription intensity (IT). TX Dependence of TF on promoter sequence, TF properties (e.g., the binding specificity of TF to its operating sequence), external signals (e.g., inducer concentration in induced expression) and other key variables in the system;

[0005] (2) Training the model, i.e., determining the I of the target system on the probe library through experiments. TX And use this data to acquire or calibrate model parameters;

[0006] (3) Based on the design specifications, new and higher-performing promoter-TF combinations are generated using the model. By establishing a computational model that can capture sequence-function relationships, researchers can simultaneously conduct virtual mutant library construction and other operations with traditional experiments, and systematically screen and evolve "hotspot" regions identified by the model, thereby improving researchers' understanding of mutant I. TX This improves the predictive ability of the system, thereby increasing screening efficiency and reducing trial-and-error costs.

[0007] Furthermore, steps (2) and (3) in this process form a closed loop, so this design can be integrated into an automated, iterative gene circuit optimization process, further shortening the "delivery cycle" of the final product and realizing the rapid construction and optimization of the prokaryotic transcriptional regulation system.

[0008] In summary, the shortcomings of existing technologies are as follows:

[0009] (1) Existing models, such as convolutional neural networks (CNN), recurrent neural networks (RNN), and transformers (attention models), often lack mechanistic explanations of relevant component levels (i.e., different molecular categories), making it difficult to understand and construct effective transcriptional regulation systems "from the bottom up" based on the information in the model.

[0010] (2) For models that require transfer learning, when there are few samples available for training in the new system, the model’s prediction accuracy may be low, which leads to the need to spend more costs to characterize the new system, further weakening the advantages of modeling. Summary of the Invention

[0011] This invention provides a model-based method and system for prokaryotic transcriptional regulation, which addresses the problems of high engineering difficulty, long time consumption, and limited optimization space in prokaryotic transcriptional regulation systems.

[0012] According to an embodiment of the present invention, a model-based prokaryotic transcriptional regulation method is provided, comprising the following steps:

[0013] Obtain the DNA element information required for each group of transcriptional regulatory systems in the corresponding test strains, and construct the transcriptional regulatory system;

[0014] Establish a biophysical quantitative model describing the transcriptional regulatory system;

[0015] Obtain valid test data for fitting quantitative biophysical models;

[0016] Use valid test data to train the biophysical quantitative model.

[0017] Furthermore, DNA element information includes transcription factor sequences and inducible promoter operator sequences.

[0018] Furthermore, valid test data include promoter background transcription intensity, transcription factor expression quantification, and exogenous inducer concentration.

[0019] Furthermore, this also includes: biophysical quantitative models providing optimal transcription factor expression levels; or

[0020] Biophysical quantitative models offer new, higher-performance promoter-TF combinations for constructing transcriptional regulatory systems; or

[0021] Multiple iterations of the biophysical quantitative model were performed to generate multiple promoter-TF combinations to optimize the performance of the transcriptional regulatory system.

[0022] Furthermore, methods for obtaining DNA element information include direct synthesis or amplification of a specific DNA template.

[0023] Furthermore, promoters include constitutive promoters, activating promoters, and repressive promoters.

[0024] Furthermore, in promoter library designs of varying intensities, rational design can be σ 70 The core promoter library identified by family members, containing orthogonal combinations of -10 and -35 variants.

[0025] Furthermore, the correct vector was obtained through plasmid extraction, DNA sequencing, and enzyme digestion verification, and then transferred into the corresponding chassis cells to obtain the corresponding test strain.

[0026] Furthermore, a transcriptional regulation system is constructed, the transcriptional intensity within the system is calculated, and the system is tested.

[0027] According to another embodiment of the present invention, a model-based prokaryotic transcriptional regulation system is provided, comprising:

[0028] The system construction unit is used to obtain the DNA element information required for each group of transcriptional regulatory systems in the corresponding test strain and to construct the transcriptional regulatory system.

[0029] The model building unit is used to build a biophysical quantitative model describing the transcriptional regulatory system.

[0030] The data acquisition unit is used to acquire valid test data for fitting biophysical quantitative models.

[0031] The model training unit is used to train a biophysical quantitative model using valid test data.

[0032] A storage medium storing program files capable of implementing any of the above-described model-based prokaryotic transcriptional regulation methods.

[0033] A processor for running a program, wherein the program, when running, executes any of the above-mentioned model-based prokaryotic transcriptional regulation methods.

[0034] The model-based prokaryotic transcriptional regulation method and system in this invention characterizes multiple transcriptional regulatory systems in test strains and identifies key factors affecting the performance of these regulatory systems under different evaluation criteria. Based on this, the system is optimized to improve gene expression and induction fold, solving the problems of high engineering difficulty, long time consumption, and limited optimization space in prokaryotic transcriptional regulation systems. Attached Figure Description

[0035] Figure 1 shows the structure of the three regulatory promoters in this invention;

[0036] Figure 2 is a diagram of the software architecture optimization scheme for transcriptional regulation used in this invention;

[0037] Figure 3 illustrates the recruitment of RNAP and TF to σ during the transcription initiation phase in different scenarios of this invention. 70 A diagram of a biophysical model of a specific promoter;

[0038] Figure 4 shows the biophysical models describing the spontaneous and induced allosteric changes of TF in different scenarios in this invention;

[0039] Figure 5 shows the verification of the linear energy model E in this invention. P picture;

[0040] Figure 6 shows the fitting and verification of the Ep linear energy model in this invention on additional data;

[0041] Figure 7 shows the validation diagram of the parameterized transcriptional regulation model in this invention;

[0042] Figure 8 shows the fitting diagram of the single-component transcription factor model in this invention to uninduced and induced transcriptional activities;

[0043] Figure 9 shows the validation of the mixed regulation model in B. subtilis for uninduced and induced transcriptional activity.

[0044] Figure 10 is a diagram illustrating the inherent characteristics of transcription factors in this invention;

[0045] Figure 11 shows the model-based performance optimization of transcription factors in this invention;

[0046] Figures 12a-12c are schematic diagrams of the optimized regulatory scheme in this invention; wherein, Figure 12a is a schematic diagram of the optimized transcription factor path in E. coli; Figure 12b is a schematic diagram of the optimized transcription factor path in B. subtilis; and Figure 12c is a schematic diagram of the optimized transcription factor path in C. glutamicum. Detailed Implementation

[0047] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0048] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0049] To address the shortcomings of existing technologies, this invention provides an optimized scheme for modeling and optimizing the design of prokaryotic transcriptional regulatory systems, solving the problems of high engineering difficulty, long time consumption, and limited optimization space in prokaryotic transcriptional regulatory systems. Furthermore, to further enhance the practicality of this scheme in guiding experimental design, this invention also designs a software architecture (hereinafter referred to as "software architecture") centered on this optimized scheme.

[0050] The software architecture comprises three independent, replaceable components:

[0051] (1) Constitutive expression model for predicting the binding ability of RNAP to promoter in the absence of TF;

[0052] (2) Regulatory expression model, used to predict gene expression intensity under TF regulation (I TX );

[0053] (3) Target optimizer, which can optimize user-specified target functions for promoter sequences and other controllable variables (such as TF expression level).

[0054] In this invention, components (1) and (2) are implemented as biophysical models at different levels of abstraction. Then, under a unified theoretical framework, multiple transcriptional regulatory systems in Escherichia coli (E. coli), Bacillus subtilis (B. subtilis), and Corynebacterium glutamicum (C. glutamicum) are characterized. Key factors affecting the performance of these regulatory systems are identified under different evaluation criteria. Based on this, the system is optimized to improve gene expression and fold induction.

[0055] The basic content of the technical solution of this invention is as follows:

[0056] 1) Through literature review, obtain information on the DNA (deoxynucleic acid) elements required for each transcriptional regulatory system, including transcription factor sequences, inducible promoter operator sequences, promoters used for systemic expression, ribosome binding sites, and terminator sequences.

[0057] 2) Establish a biophysical quantitative model describing the prokaryotic transcriptional regulatory system, which includes variables for characterizing promoters, transcription factors, and inducer molecules.

[0058] 3) Design experiments to obtain effective data for model fitting. Test data include promoter background transcription intensity, transcription factor expression quantification, and concentration of exogenous inducers.

[0059] 4) Model training: Train the model using the test data from 3) to obtain or calibrate the model parameters.

[0060] 5) Through multiple iterations, the model provides new and better-performing "promoter-TF" combinations for system construction.

[0061] The technical solution of this invention is described in detail from the following four aspects:

[0062] I. Construction of Transcriptional Regulatory System and Test Strains

[0063] 1) DNA element acquisition

[0064] Component sequence files are obtained through literature review or generated through rational design.

[0065] Primers and gene synthesis were performed at Sangon Biotech (Shanghai) Co., Ltd. DNA element sequences are shown in Table 1.

[0066] Table 1 shows the DNA element sequences involved in this invention.

[0067] 2) Transcriptional regulatory promoter structure

[0068] Based on the prokaryotic transcriptional regulation mechanism, there are three main types of promoters: constitutive promoters, activating promoters, and repressive promoters. In constitutive promoters, RNAP binds directly to the promoter to initiate transcription. In activating promoters, transcription activators and RNAP interact, working together to initiate transcription. In repressive promoters, the promoter is occupied by transcription repressors, preventing RNAP binding and thus inhibiting transcription. Both activating and repressive promoters have at least one transcription factor binding site, called the operator sequence. When designing promoters, the operator sequence of activating promoters is generally located upstream of promoter -35; the operator sequence of repressive promoters is generally located between -35 and -10, downstream of -10, or both. This invention mainly designs three types of promoters. The first type is the activating promoter, with the operation sequence located in the upstream region of -35°C (Figure 1a); the second type is the repression promoter, with the operation sequence located in the downstream region of -10°C (Figure 1b); and the third type is the hybrid promoter, which has both an activating operation sequence and a repression operation sequence, located in the upstream region of -35°C and the downstream region of -10°C, respectively (Figure 1c).

[0069] Figure 1: Schematic diagrams of three regulatory promoter structures. Activating promoter structure (a); Inhibiting promoter structure (b); Hybrid promoter structure (c). A : Activator manipulator sequence, O R : Repressor operator sequence, +1: Promoter transcription start site, -35: Promoter -35 region, -10: Promoter -10 region.

[0070] 3) Design of promoter libraries with different strengths

[0071] Using software architecture to rationally design what can be σ 70This invention utilizes a core promoter library of orthogonal combinations of variants from the -10 and -35 regions identified by family members. First, it collects constitutive B. subtilis expression promoters (SigA specific) from the DBTBS database, and constitutive C. glutamicum expression promoters obtained by Albersmeier et al. through RNA-Seq screening. Next, MEME (a motif-finding tool based on maximum likelihood estimation) is used to perform global and local alignment of these promoter sequences to estimate σ. 70 Base preference upon promoter binding (represented by PWM in the -10 and -35 regions). Subsequently, all possible variants in the -10 and -35 regions are sorted by their respective weights W:

[0072] Where w ij For the i-th row (i = 1, 2, ..., 6, corresponding to the i-th base in each region) and j-th column (i = 1, 2, 3, or 4, corresponding to bases A, C, G, and T respectively) in the PWM, δ ij =0 or 1 represents a valid bit in the encoding matrix of a certain variant that is related to w. ij The corresponding elements. Based on the sorting results, a 5×4 orthogonal subset was selected from these variants, involving two steps:

[0073] 1. Select 5 variants from region -10 and 4 variants from region -35, ensuring that the order spacing between the selected variants is equal, provided that the preset minimum order is met.

[0074] 2. Replace some variants mentioned above but not mentioned in step 1. This step is optional and is used to introduce specific σ into the promoter library. 70 Prior information about factor preference sequences; this prior information can be the experimenter's own experience or promoter sequences reported in the literature.

[0075] Finally, these 20 motifs are combined and embedded into a fixed context (i.e., the same upstream sequence (UP), downstream sequence, and spacer sequence) to obtain a complete promoter library. Here, we assume the ΔW and corresponding ΔE for any two -10 (or -35) region variants. P There is a positive correlation between them; therefore, if the linear energy model holds, then the Ei of each promoter sequence is... P (and the corresponding I) TX The order of ) should be the same as W. -10 (or W) -35 This invention employs a hierarchical design approach for key promoter elements, which helps to cover a relatively broad and measurable range of transcriptional intensity while keeping the library size within a reasonable and manually characterizable range.

[0076] 4) Carrier construction

[0077] All plasmids used in this invention were assembled using Golden Gate technology and cloned in E. coli DH10B-Alt (a derivative strain of E. coli DH10B, whose genome contains two repressive transcription factors (LacI and TetR) and one activating transcription factor (AraC)) or E. coli “Marionette” strain. The backbone plasmid sequences used in this invention are shown in Table 2.

[0078] Table 2 shows the backbone plasmid sequences involved in this invention.

[0079] bs-RT: Bacillus subtilis reporter gene expression plasmid; bs-TF-T1: Bacillus subtilis TF expression control plasmid (LacI control system); bs-TF-T2: Bacillus subtilis TF expression control plasmid (cymR control system); cg-T: Corynebacterium glutamicum system test plasmid; ec-RT: Escherichia coli reporter gene expression plasmid; ec-TF-T: Escherichia coli TF expression control plasmid.

[0080] 5) DNA introduction via chemical transformation and electroconversion methods

[0081] The correct vector was obtained through plasmid extraction, DNA sequencing, and enzyme digestion verification. This vector was then transformed into the corresponding chassis cells to obtain the corresponding test strains. For *E. coli*, the chemical transformation method (TSS method) was used. For *B. subtilis*, competent cells were prepared using the MD culture transformation method for DNA transformation. For *C. glutamicum*, electroporation was used. To prepare competent cells for electroporation, 0.5 mL of overnight culture was inoculated into 100 mL of EPO medium and cultured at 30°C on a shaker at 220 rpm until the OD of the culture medium reached a certain level. 600 The concentration was 0.8–1.0. Next, the culture medium was centrifuged at 4000g and 4°C for 10 minutes, and the precipitate was resuspended in 20 mL of 10% glycerol solution. This centrifugation and resuspension process was repeated twice to wash away any residual culture medium in the bacterial culture. Finally, the precipitate was resuspended in 2.0 mL of 10% glycerol solution. The suspension could be aliquoted into 1.5 mL centrifuge tubes and stored at -80°C or used immediately for electroporation. Electroporation was performed using a Bio-Rad Micropulser, with each electroporator containing 50 μL of cell suspension and a pulse voltage set to 1.8 kV.

[0082] 6) Construction of test strains

[0083] For *E. coli*, wild-type strain MG1655 was selected as the base strain for prokaryotic transcriptional regulation system testing. To effectively control the expression level of TFs in *E. coli*, the endogenous lacI gene (MG1655ΔlacI) was first knocked out of the genome; then, two repressive TFs were combined, vanR–cymR (strain ECint034) or vanR–lacI (strain ECint035), and integrated into the λ attachment site (attB), respectively; in addition, the endogenous araC gene in ECint035 was knocked out using the lambda Red recombination method, resulting in the araC-deficient strain ECint036. For *B. subtilis*, strain 168 (trpC2, ICEBs1) was selected. + The basic strain was selected, and the corresponding TF expression module was integrated into the amyE site in the genome, while the reporter gene module was integrated into the ICEBs1 site. For C. glutamicum, strain ATCC 13032 was selected as the basic strain. All strains used in this invention are shown in Table 3.

[0084] Table 3 Information on the test strains of this invention

[0085] 7) Inducer molecules

[0086] The inducing molecule used in this invention is shown below:

[0087] N-(β-ketohexanoyl)-L-homoserine lactone (3-O-C6-HSL; Sigma-Aldrich, K3007); N-(3-oxododecanoyl)-L-homoserine lactone (3-O-C12-HSL; Sigma-Aldrich, O9139); N-(p-coumaryl)-L-homoserine lactone (pC-HSL; Sigma-Aldrich, 07077); N-(3-hydroxytetradecanoyl)-DL-homoserine lactone (3-OH-C14-HSL; Sigma-Aldrich, 51481); N-(isovaleroyl)-L-homoserine lactone (IV-HSL; chemically synthesized, Shenzhen Qianyan Pharmaceutical R&D Technology Co., Ltd.); cumic acid (CA; Sigma-Aldrich, 268402); vanillic acid (VA; Sigma-Aldrich, 94770); isopropyl-β-D-1-thiogalactoside (IPTG; Macklin, I6148); L-(+)-arabinose (BBI, A610071); D-(+)-xylose (BBI, A600998); L-(+)-rhamnose (Solarbio, G8560).

[0088] II. Transcriptional Regulatory System Testing

[0089] 1) Transcription intensity calculation

[0090] The test intensity of all reporter genes (sfYFP) is expressed as relative gene expression intensity (relative promoter unit, RPU):

[0091] Among them, OD, OD ref OD0 and OD10 represent the OD values ​​of the strain containing the target reporter gene, the strain containing the reference reporter gene, and the strain containing the basic strain, respectively. 600 ;I and I ref The fluorescence intensities of strains containing the target reporter gene and strains containing the reference reporter gene are respectively; OD 600 Both fluorescence intensity and fluorescence intensity were measured during the logarithmic phase.

[0092] 2) Transcriptional Regulation System Test

[0093] For E. coli, the transformed cells were first placed in LB medium and cultured overnight at 37°C and 800 rpm in 96-well microplates. Then, 1 μL of the bacterial culture was inoculated into 100 μL of fresh M9 medium (containing the appropriate antibiotic and inducer for TF expression) in each well, and cultured at 37°C and 800 rpm for 2 hours (this is the first induction stage). Next, another 100 μL of M9 medium (containing the appropriate antibiotic and inducer for reporter gene expression) was added to each well, and cultured at 37°C and 800 rpm for 3 hours (this is the second induction stage). Finally, the OD of the liquid in each well was measured using a microplate reader. 600 And fluorescence intensity. For B. subtilis and C. glutamicum, the strains were activated from glycerol stock solution and incubated overnight before testing. For B. subtilis, the second induction phase was 2 hours; other conditions were the same as for E. coli. For C. glutamicum, BHI medium was used for overnight incubation at 30°C; EZ-rich medium was used for inoculation into microplates; other conditions were the same as for B. subtilis.

[0094] III. Model Construction and Training

[0095] In training constitutive promoter transcription strength model M P At that time, the transcription intensity I containing N promoters was first... TX A subset Θ of the model parameters fitted on the dataset P =(E -10 E -35 ,I max These promoters are typically derived from a core promoter library (fixed upstream (UP), downstream, and spacer regions). The fitting method is about Θ. P Minimize model predictions Compared with the measured value I TX Mean squared logarithmic error (MSLE) between:

[0096] Where i (or j) is the variant number for zone -10 (or zone -35), For I TX The geometric mean, E UP It is fixed at 0. Afterwards, parameter E... UP,k Minimize the transcription strength I of the promoter containing the k-th UP region variant TX (k) and the corresponding predicted value The MSLE between -10 and -35 regions is obtained, at which point the sequences of -10 and -35 regions remain fixed.

[0097] In training a transcription strength model M regulated by a single TF promoter TF At that time, the transcription intensity dataset (I) was obtained under N sets of experimental variables. TX (c TF ,c I A subset of the fitted model parameters Θ TF The fitting method is about Θ TF Minimize model predictions Compared with the measured value I TX MSLE between:

[0098] in Θ TF =(E A E AP ,k1,k2,n)(activating TF) or (E R (k1,k2,n)(suppressive TF), c TF and c I These represent the TF expression level and inducer concentration under the control of the experimental protocol, respectively, E P and I max From the above M P Provided.

[0099] When constructing a transcriptional strength model of a promoter that can be simultaneously regulated by activating and repressive TFs, M AR The parameter subset Θ of the activation-type TF model TF =(E A E AP The parameter subset Θ of the suppressed TF model (k1,k2,n) TF '=(E R Combining k1', k2', n', we get M. AR Complete parameters: Θ AR =(E A E AP E R ,k1,k2,n,k1',k2',n').

[0100] The above fitting process was conducted under the joint guidance of the L-BFGS-B algorithm and the basin-hopping strategy. For M... P The initial parameters are given by W. -10 and W -35 Given; for M TF The initial parameters are given by c TF and c I The simplified relationship between them was estimated.

[0101] IV. Model Optimization

[0102] When optimizing the objective F defined by equation 6, from a single TF model M TF To begin; the parameters of the model have been fitted to the initial promoter S0 (Table 5) at different c TF The transcription intensity was determined under the following conditions, where c TF,0 To maximize F, c TF Subsequently, regarding E P,i and c TF,i (i = 1, 2, ... represents the current iteration number) Maximize F:

[0103] Where j is the index of a key motif assembly in a core promoter library (of size L), and Comb(j) returns the candidate promoter sequence obtained by embedding motif assembly j and the binding site (operator sequence) of the TF into a fixed context, M P The constitutive expression promoter model trained in the previous section (used to predict E from Comb(j)) P ), c TF,max This represents the upper limit of TF expression. Here, a hybrid optimization strategy is employed: by enumerating combinations of -10 and -35 region variants in the promoter sublibrary, and then applying c to each combination j... TF Perform local optimization (L-BFGS-B) to obtain the local minimum F. i,j =F(c TF M P (Comb(j)),M TF Then compare all F's. i,j The global optimum F is obtained. i and the corresponding combination of variables (E) P,i *,c TF,i *), and used it as the recommended control scheme for testing. Subsequently, in a new round of characterization experiments, F(E) was determined. P,i *,c TF,i *;M TF ), that is, let E P,i+1 =E P,i *、S i+1 =Comb(j*) and c TF,i+1 ≈c TF,i *; and the corresponding transcription intensity data I TX (S i+1 ,c TF,i+1 ,c I The subset of parameters used to update the model after being merged with existing data is specifically described using the method concerning Θ. TF Minimize all N sets of model predictions Compared with the measured value I TX The MSLEs between (involving promoters S0, ..., S)i and S i+1 ):

[0104] Where Seq(j) is the promoter sequence used in the j-th data set. Θ TF,i =(E A E AP ,k1,k2,n)(activating TF) or (E R (k1,k2,n)(suppressive TF), w j The weights for the j-th data group:

[0105] Then, increment the iteration number i by 1 and continue executing this "optimization-representation-update" loop until E... P,i * =E P,i And F(E) P,i * ,c TF,i * M TF )≤2F(E P,i ,c TF,i M TF ), or i≥5 (maximum number of iterations). Finally, E P,i S i and c TF,i The final optimization plan will be reported to the user.

[0106] The key points and areas to be protected in this invention are:

[0107] 1. This invention provides a model algorithm for model training and objective function optimization.

[0108] 2. The promoter architecture design involved in this invention includes, but is not limited to, the promoters and promoter library sequences involved in this invention (Table 1).

[0109] 3. The prokaryotic transcription system test backbone plasmid sequences involved in this invention (Table 2).

[0110] 4. The method for optimizing the prokaryotic transcription system based on promoter strength and transcription factors proposed in this invention.

[0111] Compared with the prior art, the advantages of the present invention are:

[0112] 1. In this invention, a large number of high-performance components can be rapidly obtained based on the optimized scheme. Modular system design and standardized testing significantly reduce the time and manpower costs of engineering prokaryotic transcription systems.

[0113] 2. This invention constructs a biophysical model based on the prokaryotic transcriptional regulatory machine. This method offers fast iteration speed and high prediction accuracy. Only a small number of construction and testing samples are needed to fit the parameters, significantly reducing the number of models required and improving optimization speed.

[0114] The invention has been proven feasible through experiments, simulations, and use, as shown in the following results:

[0115] I. Software Architecture Optimization Platform Setup

[0116] The software architecture accepts user-input design specifications (goals), including the desired TF type and expected transcriptional intensity; it then outputs an optimized regulatory scheme, including externally controllable variables such as promoter sequences and TF expression levels (Figure 2a). To achieve this functionality, users need to complete three steps using the software architecture:

[0117] (1) Select a model that can quantitatively describe the effect of the binding strength of transcription factors to promoters on transcription strength. This model can also include the response to external variables (such as TF expression level and inducer concentration).

[0118] (2) Obtain the parameters of the selected model, which can be fitted from publicly available datasets (such as Regulon DB), or designed with the guidance of the initial model to collect experimental data and use it for fitting;

[0119] (3) Use an optimizer to optimize the promoter sequence and control variables.

[0120] Steps (2) and (3) correspond to the “test-learning” and “design-build” stages in the traditional “design-build-test-learn (DBTL)” cycle, respectively. The entire process can be iterated multiple times until a control scheme that meets the design specifications is finally obtained, that is, the optimal parameter combination under the prediction of the selected model.

[0121] The built-in biophysical model in the software architecture is the core of the entire design process and the key to predicting the performance of candidate solutions. In order to facilitate the acquisition and verification of model parameters, the binding process of RNAP / σ factor complex and TF to promoter is decoupled from the (spontaneous or induced) allosteric process of TF. For the former, a general statistical thermodynamic model needs to be constructed, the key parameters of which include: (1) the energy change (E) when RNAP / σ factor complex binds to promoter alone. P ), and the maximum transcription intensity (I) in a specified transcription system. max (2) Energy changes (E) when the RNAP / σ factor complex binds to activating TF and promoter in concert. A and E AP (3) Energy change when the repressive TF binds exclusively to the promoter (E)R (Figure 2b; Figure 3a–e). For the latter, the effective TF concentration c needs to be... TF,eff The concentration of TF molecules that can activate or inhibit transcription initiation is approximately expressed as a partition function containing multiple microstates, the weights of which depend on the reaction balance between TF monomers and oligomers in the absence of an inducer (k1) and with an inducer (k2) (Figure 2b; Figure 4). For each combination of RNAP / σ, TF, and operator sequence, the above model parameters can be obtained by fitting the transcriptional intensity data (I) obtained from the overall experiment. TX It can be obtained from other relevant experimental data, or estimated from other relevant experimental data.

[0122] The optimization process of the software architecture involves searching and iterating model parameters, with the goal of maximizing a set of user-specified objective functions. By altering the sequence of the promoter's core region (i.e., the key region identified by the σ factor) and adjusting TF expression levels (controlled by another inducible expression system), the effects of input variable E on the optimization of the software architecture were systematically explored. P and c TF to I TX The influence (Figure 2c). In this process, it is assumed that other parameters in the model remain unchanged. This assumption holds that the relevant parameters are only affected by the following factors: (1) specific protein-DNA interaction interfaces, such as E A and E R It mainly depends on the contact surface between the TF and the operator sequence; (2) specific protein-protein interaction interfaces, such as E AP (3) The polypeptide sequence itself, such as k1 and k2, mainly depends on the spatial structure of the TF; (4) Other controllable factors (such as gene copy number, culture conditions, etc.) max Given the model parameters and input variables, the value of each objective function can be calculated, and then they are weighted and summed to obtain a scalar function F, which is then used for single-objective optimization. The optimized input variable (E) P * and c TF * This is then converted into the corresponding transcriptional regulation protocol (promoter sequence S). * and c TF * And report it to the user.

[0123] Finally, to facilitate researchers' use of the software architecture, a web-based graphical user interface (GUI) was also developed. This interface works in conjunction with the corresponding backend modules to provide users with a simple and easy-to-use workflow (d in Figure 2): (1) First, the user needs to upload a list, which includes known promoter sequences recognized by a certain σ factor; based on this list, the software architecture will generate an initial transcription intensity prediction model M. P (2) Based on M P The software architecture will provide the I that users need to determine. TX The promoter sequence; subsequently, the software architecture will use this experimental data to calibrate M. P (3) Users obtain a series of transcription intensity data by testing specific combinations of “TF-operator sequences” on one or more promoters. TX Then, using this data, a transcriptional regulation model M is created. TF (4) The software architecture will automatically optimize M TF The input variables are used to maximize the user-specified objective function. Steps (3) and (4) can be iterated repeatedly until the final objective function value meets the user's requirements.

[0124] Figure 2: Optimization of transcriptional regulation schemes using software architecture. a) The user inputs one or more targets to be maximized (e.g., transcriptional intensity, fold change) (left column) and selects the desired TF model from the database. Based on the model type (activation, repression, or heterozygous regulation) and its parameters, the software architecture provides the optimized transcriptional regulation scheme (right column), including the reporter gene promoter sequence (S*) and its RNAP binding energy (E). P *), required TF expression level (c TF *) and the corresponding objective function value. When the existing model cannot meet the requirements, users can create a new model by uploading experimental data, which includes characterizations measured on one or more “promoter-TF” combinations, and can be iteratively updated via DBTL loops (middle column) until the model meets the user's expectations for the optimized objective. b, Overview of the biophysical models used in this study; and transcriptional strength (I TX The main factors related to this are presented in the form of model parameters. This invention divides the model parameters into two groups: inter-component interactions (promoter-RNAP, promoter-TF, etc.; left column) and intra-component interactions (TF-ligand, TF monomer interactions, etc.).

[0125] II. Modular Design of Promoter Core Region and Operator Sequence In prokaryotic cells, the promoter core region and operator sequence are designed by σ 70 (σ AThe promoters recognized by the σ factor of the σ family have specific regions of base preference (i.e., motifs). These motifs interact with specific structural domains of the σ factor, directly affecting their binding efficiency. This invention uses an energy-related term (E... P This interaction is described using the formula (Figure 3a). Studies have shown that although introducing interaction effects into the motifs (-10, -35, and UP regions) can further improve the model's predictive accuracy, previous experimental evidence in *E. coli* and *B. subtilis* indicates that mutations in the promoter's -10, -35, and UP regions have a significant impact on the predictive accuracy of the model. P The contributions can be considered relatively independent (Figure 5a). Based on this, the present invention proposes a linear energy model, the formula of which is as follows: E P =E0+E -10 +E -35 +E UP (1)

[0126] Where E0 is the negative Gibbs free energy change (i.e., -ΔG) of a reference promoter binding to the RNAP / σ complex; E -10 E -35 and E UP These are the differences between -ΔG and E0 (i.e., -ΔG) that bind to the RNAP / σ complex after base changes in the -10, -35, and upstream regions of the reference promoter, respectively.

[0127] Similar to previous work reported in E. coli, based on σ 70 Based on the base preference (represented by the position weight matrix (PWM)) (b–d in Figure 5), this invention starts from the conserved sequences (TTGACA–TATAAT) in the -10 and -35 regions and constructs a 4×5 promoter orthogonal library in B. subtilis and C. glutamicum respectively using a software architecture (Table 4-5).

[0128] Table 4 shows the -35 and -10 startup sub-libraries generated by the software architecture.

[0129] *Mutants are sorted in descending order of their intensity; bases that do not differ from the conserved sequence are underlined.

[0130] Table 5 shows the promoters and related parameters used to check convergence when optimizing the B. subtilis LuxR system.

[0131] a, energy-related parameters in terms of k B T is expressed in units.

[0132] b. Transcriptional activity is expressed in RPUs and is the average of three independent measurements.

[0133] c, the optimization objective is calculated using formula (8) in the paper, and relative to the promoter activity model (M). P Estimated I max (5.955 RPU) was normalized.

[0134] The measured background strength of these promoters (I TX The range is wide (B. subtilis: 928-fold; C. glutamicum: 570-fold), and it is accurately predicted by linear energy models on a logarithmic scale (Figure 5e, Figure 5f). Each model includes 11 free parameters (4 Ei, 4 Ej, ... -35 5 E -10 1 E0 and 1 I max (Table 4); The performance is comparable to that of the previous 6×8 orthogonal promoter library in E.coli (g in Figure 5 and a in Figure 6; Table 4), which used 16 free parameters. In contrast, only PWM is used to estimate E P The initial model (hj in Figure 5) and previously reported models for the I of the E. coli promoter library TX The prediction accuracy is low (Figure 7), and the prediction bias is more pronounced for medium-strength promoters.

[0135] Next, this invention clarifies E UP For E P The additive effect of RNAP, specifically the C-terminal domain (CTD), also means that the C-terminal domain of RNAP binds independently to the upstream region of the core promoter. This implies that when designing activation-inducible promoters, it is only necessary to insert the corresponding operator sequence upstream of the promoter. This invention replaces the UP region of the original promoter with different TF operator sequences and estimates the corresponding E... UP The obtained E UP In B. subtlis (Fig. 5 b, left column; Fig. 6 e and Fig. 6 f), C. glutamicum (Fig. 5 c, left column; Fig. 6 g and Fig. 6 h) and E. coli (Fig. 5 d, left column; Fig. 6 c and Fig. 6 d), different E -35 +E -10 The results remain consistent under all conditions. These results collectively demonstrate the applicability of the linear energy model to promoters with deterministic structures. This provides a basis for designing promoters with specific Ei values. P The promoters provide valuable resources and are more cost-effective and accurate compared to screening-based methods.

[0136] Figure 5: Validation of the linear energy model E PFigure 5a shows a schematic diagram of the interaction between the RNAP / σ complex and the key promoter region; this study explored the interaction pair E. P The contribution. In Figure 5, for b–d, B. subtilis (b), C. glutamicum (c), and E. coli (d), σ 70 (σ A The base preference of the E factor for the -35 region (middle column) and -10 region (right column) of the promoter is shown in logo form. This base preference is derived from motif alignment and is used to guide the construction of orthogonal promoter libraries in various species. Different upstream (UP) sequences were also selected to verify E. UP For E P Additivity (left column). Each regional variant pairs with E. P The relative contributions were derived from model fitting and are indicated by color gradients. In the -10 region of each promoter library, only the last six bases were changed. Figure 5 shows e–g, representing I values ​​obtained based on orthogonal promoter libraries. TX The figure shows the fitting results of the linear energy model. In Figure 5, h–j is similar to e–g, but an uncalibrated model is used to predict I. TX E -35 and E -10 The value is estimated from the base preference in b–d.

[0137] Figure 6: E p Fitting and validation of the linear energy model on additional data. a) Fitting a linear energy model to predict transcriptional activity (in terms of relative promoter activity p) for an E. coli promoter library containing the upstream (UP) region of LasO. bound (Represented). b, similar to a, the only difference being that two libraries, LasR and AraC, were used for fitting. c–h, for E. coli (c and d), B. subtilis (e and f), and C. glutamicum (g and h), the energy contribution (E) to the UP region. UP Cross-validation is performed to estimate E. UP The subsets of data used to validate the fit were labeled in the title of each subplot.

[0138] III. Transcription Factor Modeling within a Unified Thermodynamics Framework

[0139] Transcription factors exhibit various interactions (such as affinity or repulsion) with the RNAP / σ factor complex, and their allosteric responses to external signals (such as inducers) are also diverse. Without loss of generality, this invention also employs a modular strategy similar to that used in promoter design to establish a parameterized transcription factor model, which includes an energy-related term (E...). A / E R and EAP Figure 3) and allosteric equilibrium terms (k1, k2, and n); Figure 4, are used to describe the interaction between TF and the promoter and RNAP / σ (Figure 3), and the effect of TF on its expression level (c TF ) and inducer concentration (c I The response of ) (Figure 4). For activating TF, the model includes five free parameters (E). A E AP k1, k2, and n; in Figure 3b), this model, in three chassis (E. coli, B. subtilis, and C. glutamicum), can be used not only to describe the I in the LuxR control system. TX The responses to LuxR expression levels and corresponding inducer (C6-HSL) concentrations (a–c in Figure 7) can also be used to describe the regulatory characteristics of seven other activating TFs (Figure 8); in most models, n is fixed at 2. This invention uses the relative transcriptional activity of the promoter (p...) bound Z is defined as the probability ratio between the promoter microstates that are bound by RNAP / σ and the promoter microstates that are not bound, i.e., (Z P +Z AP ) / (1+Z A It was observed that in different c TF and c I Under the condition, p bound The model converged to the predicted curve in all three chassis (Figure 7b). Similar results were obtained for repressive transcription factors like LacI (Figure 8), where each model contains four free parameters (E). R k1, k2 and n; in Figure 3c), and observe p bound Compared to the "bound / unbound" ratio (i.e., Z), P / (1+Z R The convergence of the prediction curve (Figure 7).

[0140] Figure 3: Under different scenarios, RNAP and TF are recruited to σ during the transcription initiation phase. 70 Biophysical models of specific promoters. The left column shows schematic diagrams of promoter binding states under various scenarios; the right column shows models used to quantitatively describe transcription intensity (IT). TX For each microstate (Z0, Z...) P Z A The partition function of the (etc.) dependency relationship represents the probability (p) of observing a microstate related to transcription initiation in a macrocanonical ensemble. boundIn Figure 3a, transcription initiation of a constitutively expressed (unregulated) promoter requires binding to RNAP (containing the σ factor). In Figure 3b, transcription initiation of a promoter regulated by a single activating TF (“Activator”) requires binding to both RNAP and / or the Activator. In Figure 3c, transcription initiation of a promoter regulated by a single repressor TF (“Repressor”) requires binding to RNAP but not to the Repressor. In Figure 3d, transcription initiation of a promoter regulated by both activating and repressor TFs requires binding to both RNAP and / or the Activator, but not to the Repressor. In Figure 3e, an extension of b and d, additionally considers RNAP binding to the promoter (Z). AP ) Rather than escaping from the starter subregion (Z TX The competitive process of ). E P E A and E R E represents the negative of the Gibbs free energy change of the promoter binding to RNAP, the activating TF, and the repressive TF, respectively; AP E is the negative of the difference in Gibbs free energy change when the activating TF binds to the promoter alone or in conjunction with RNAP and the promoter; P ' is the negative of the Gibbs free energy change when RNAP binds to the transcription initiation region (i.e., successfully escapes from the promoter region). max : Maximum transcriptional intensity given the expression system and promoter structure; k B Boltzmann constant; T: thermodynamic temperature (unit: Kelvin).

[0141] Figure 4: Biophysical models describing spontaneous and induced allosteric changes of TF under different scenarios. Figure 4a shows the microstates associated with the effective TF form in each scenario. Figure 4b shows a quantitative description of the effective TF concentration (c...). TF,eff ) on total TF expression level (c TF ) and inducer concentration (c I The partition functions of the dependencies are: (i) TFs function as oligomers that are unbound or bound to the inducer (e.g., CinR bound to C14-HSL); (ii) TFs function only as oligomers that are unbound to the inducer (e.g., LacI to IPTG); (iii) TFs function only as oligomers that are bound to the inducer; (i) and (ii) are the default transcriptional regulation models for activating and repressive TFs implemented in the software architecture, respectively.

[0142] Figures 3 and 4 show the specific expressions of the model under specific regulation types. Figure 2c illustrates the objective function path optimization process in a typical iterative optimization process; this optimization process is guided by the software architecture and includes three iterations. A user-selected TF model is updated in each round of optimization, subsequently providing a predicted value of F within the feasible region; the optimizer then continuously searches for F with respect to E. P and c TF The global maximum value. In the third iteration, the optimizer suggests the optimal combination of variables (E0). P *,c TF * (marked with an asterisk) are sufficiently close to the actual regulatory scheme tested; therefore, the optimization process terminates after this iteration according to the termination criteria. Figure 2d shows a schematic diagram of the software architecture used via a web front-end in a typical gene element design workflow. Constitutive expression promoter transcription strength model (M... P ) and transcriptional regulation model (M TF These are not available before the process begins, but can be created and iteratively updated based on experimental data.

[0143] Figure 7: Validation of the parameterized transcriptional regulation model. In Figure 7a, the fit has four free parameters (E). A E AP A model of activating transcription factors (k1 and k2; n fixed at 2) was applied to the transcriptional activation systems of E. coli, B. subtilis, and C. glutamicum; the biophysical model used for fitting is shown again at the top (Figure 3b). Figure 7b shows the relative promoter activities (pk1 and k2; n fixed at 2) of activating transcription factors in the three chassis. bound ) and (Z P +Z AP ) / Z A The relationship between them. In Figure 7c, p of LacI in the three chassis. bound With Z P / (1+Z R The relationship between ) is shown in Figure 7. In E. coli, under the mixed regulation of LuxR+LacI, p bound With (Z) P +Z AP ) / (1+Z A +Z R The relationship between the two models is shown in Figure 7e. The simple fitting model (red line) and the extended model (blue line) are applied to the LasR-activated transcription system in E. coli. The biophysical models used for fitting are shown on the left.

[0144] After successfully constructing the single transcription factor model, this invention further incorporates the regulatory effects of multiple transcription factors (referred to as "mixed regulation") into its framework. Based on the fitting model of the single-component LuxR and LacI (Figure 3d), this invention creates a two-component mixed regulation model of LuxR+LacI and tests its predictive ability on promoters containing LuxO and LacO operator sequences in E. coli. Similar to the single-component transcription system, although the two-component mixed model I... TX In different c TF (LuxR), c TF (LacI), c I (C6-HSL) and c I The values ​​vary under (IPTG) conditions, but are generally captured well by the model (Figure 8), and when p bound Rewritten as the "bound / unbound" ratio (Z) P +Z AP ) / (1+Z A +Z R When the function is expressed as ), it can also converge well to the model prediction curve (d in Figure 7). In other two-component mixed regulation systems, such as XylR in B. subtilis A +XylR R (Both are induced by xylose, but the latter is a naturally occurring repressive TF in the genome) heterozygous system, in a single component (XylR) A and XylR R Even without parameter calibration, this method exhibits good predictive ability (Figure 9). These results not only confirm the effectiveness of this parameterization method in describing TF-DNA interactions, TF-RNAP / σ interactions, and the allosteric effects of TF on I... TX The effectiveness in terms of impact also verified the composability of the model, that is, when individual TF modules are combined into a hybrid regulatory module, there is no need to additionally characterize the properties of the combined system.

[0145] Figure 8: Fitting of single-component transcription factor models to uninduced and induced transcriptional activities. In Figure 8a, (from left to right, top to bottom) I of LuxR, RpaR, CinR, BjaR, LasR, AraC, and LacI in E. coli. TX With TF expression level (c TF The dependencies between them are shown in Figure 8b (from left to right, top to bottom). The dependencies between LuxR, RpaR, CinR, BjaR, LasR, AraC, XylR, RhaS, and LacI in B. subtilis are shown. TX With TF expression level (c TFThe dependencies between them are shown in Figure 8. (From left to right, top to bottom) The dependencies of LuxR, RpaR, CinR, BjaR, AraC, RhaS, and LacI in C. glutamicum are shown. TX With TF expression level (c TF The dependency relationship between ) and ). Experimental value I TX I is represented by points, while model predictions are represented by lines. It for each transcription factor in the presence or absence of the corresponding inducer. TX They are represented by dark blue and light blue, respectively.

[0146] Figure 9: Validation of uninduced and induced transcriptional activity in B. subtilis using a mixed regulatory model. XylR and XylR R In the promoter co-regulated by (local repressor factors in the B. subtilis genome), I TX With c TF Dependence between (XylR); I of each transcription factor in the presence or absence of a corresponding inducer TX They are represented by dark blue and light blue, respectively.

[0147] IV. TF Performance Evaluation and Optimization

[0148] In the third point, the present invention systematically alters the external variables (c) of the transcription system. TF and c I Furthermore, it was observed that TFs in E. coli, B. subtilis, and C. glutamicum produced different response characteristics. Specifically, some TFs (such as LuxR) showed generally lower transcriptional intensity (i.e., "leakiness") in the absence of an inducer, and were less responsive to c TF Insensitive (a–c in Figure 7), while another group of TFs (such as CinR in B. subtilis and C. glutamicum) are in c TF Higher levels of expression resulted in higher leakage expression (Figure 8). Similarly, when this invention focuses on the overall activating capacity of TFs, i.e., the maximum difference in transcriptional intensity when both TFs and their inducers are present versus absent, it can be seen that most activating TFs in E. coli can induce I at higher expression levels. TX Upgraded to I max Approximately; however, in C. glutamicum, all TFs at their activation I TX All (relative to I) maxThe performance of TFs is relatively low (Figure 8). According to the aforementioned model, these performance differences actually reflect the differences in interactions between TF monomers and between them and inducers and RNAP / σ complexes, and directly affect the performance of higher-level gene circuits constructed based on these TFs. Therefore, this invention will explore the key factors affecting TF performance; then, this invention will propose a general workflow for optimizing single or multiple TF performance metrics under the guidance of parameterized TF models.

[0149] First, this invention defines "intensity multiple difference (FC)". s ")," which is the ratio of transcription intensity in the presence or absence of TF and its corresponding inducer:

[0150] This definition is similar to the traditional "fold-change (FC)" because both reflect c I to I TX The impact. However, FC s Also followed c TF This factor, when the TF's background activation ability is strong, c TF It has a significant impact on the leakage expression level; CinR in E. coli and C. glutamicum is a good example (Figure 8). For simple activating TF models, if the background transcriptional strength of the promoter is low (Z... P <<1), and the proportion of the manipulate sequence occupied by TF is relatively high when there is an inducer (when c TF >>0 and c I >>0, Z A >>1), FC s It can be approximated as Z P (dependent only on E) P ) and E AP Functions:

[0151] Clearly, for a given TF, its E AP If fixed, then FC s With Z P (and E) P A negative correlation was observed. Conversely, for baseline transcriptional strengths close to (i.e., E...), a negative correlation was observed. P For promoters that are similar in appearance but regulated by different TFs, FC s With E AP A positive correlation was observed (Figure 10a). It is noteworthy that the maximum achievable FC varies for different species. s The values ​​differ; however, within the same species, E AP Assuming accurate estimation is possible, the FC of each TFs s Sorting and its EAP They are positively correlated.

[0152] Therefore, to maximize the activation effect of TF, it is necessary to reduce the background transcriptional strength of the promoter (i.e., reduce E). P ), and / or increase the interaction strength between TF and the RNAP / σ complex (i.e., increase E AP Based on this theory, the present invention utilizes BjaR (a species of E. coli with relatively low E. coli content). AP (approximately 4k) B The ligand-binding domain (LBD) of the transcription factor T) is associated with RpaR and LasR (two transcription factors with high E2 levels). AP (approximately 7k) B By fusing the DNA-binding domain (DBD) of a transcription factor (T) into its DNA-binding domain, two hybrid transcription factors were designed, named BjaR-RpaR and BjaR-LasR, respectively. Subsequently, the performance of RpaR and LasR-specific promoters in the presence and absence of BjaR-specific inducers was characterized. Results showed that the Et of these fused transcription factors… AP There has indeed been an improvement (approximately 6k). B T), compared to BjaR (Figure 10, left column), the corresponding FC s It was also improved (by 6 to 7 times, corresponding to exp(ΔE)). AP )≈e 6-4 ≈7). This result not only supports the applicability of Equation 3 in estimating the strength of TF-RNAP interactions, but also demonstrates the dominant role of TF's DBD in stabilizing the TF-RNAP complex to promote transcriptional activation. In summary, FC s This method more directly characterizes the activation capability of TFs than traditional FCs and may be more valuable for applications in automated genetic circuit design. In this design, the expression level of TFs is precisely controlled to achieve specific logical flows, thereby not only guiding the selection of promoters but also optimizing the selection of TFs.

[0153] Figure 10: Description of intrinsic properties of transcription factors. In Figure 10a, promoter strength fold induction (FC) in E. coli, B. subtilis, and C. glutamicum. s The interaction between ) and activating transcription factors, RNAP, and the promoter process (E) AP The functional relationship between ) is shown in Figure 10b, where TF inducement multiple (FC) is the most significant factor among the three chassis types. TF The relationship between the TF characteristic function and the negative logarithm (k1) of the equilibrium constant between the TF monomer and n-mer is shown at the top of the figure. FC s and FC TF By c TF c and ITX The calculations show that the simplified expression (reference model) is plotted as a solid curve, while the measured values ​​are represented as points. This is because, at the highest possible c... TF In B. subtilis, the activation of LasR did not reach saturation (Figure 8), therefore its FCs could not be reliably determined from the experimental data and were excluded from the analysis. Similarly, the parameter n could not be fixed during fitting for AraC in E. coli and RpaR and RhaS in B. subtilis, which is inconsistent with other TFs, thus their FCs were also excluded. TF It was also excluded. In Figure (b), in order to ensure that the experimental setup could reliably quantify the maximum c. TF The predicted values ​​for B. subtilis are truncated above 600, and the predicted values ​​for C. glutamicum are truncated above approximately 100.

[0154] Subsequently, this invention defines the performance metric "TF induction factor" (FC). TF This refers to the ratio of TF expression levels required to achieve a specific promoter activity level I0, in the presence and absence of an inducer.

[0155] Where I0∈(0,max{I(c TF ,c I ):c I =0) represents the predetermined maximum leakage expression; in the simple model of activating TF (Figure 3b), c TF,1 and c TF,2 It can be uniquely determined by I0. Intuitively, FC TF It is reflected in "I" TX ~c TF "The horizontal distance between two points with the same ordinate I0 located on two different induced curves in the logarithmic relationship graph. More importantly, FC..." TF A c is given TF The scope of work, i.e., c TF The maximum range of variation; beyond this range, no activity (i.e., I) is induced. TX (c I =0)) will exceed the inducible activity (i.e. I) TX (c I >>0)), this is due to the increase in the number of effective dimers caused by TF basal dimerization. To illustrate this more clearly, this invention assumes that the promoter has low basal transcriptional strength (i.e., Z). P <<1), and in the absence of an inducer, the proportion of the promoter occupied by TF alone is much lower than the proportion occupied by TF and the RNAP / σ complex (i.e., when c I When Z = 0, A < <Z AP), then FC TF This can be approximated as k1 and c TF,1 Functions: FC TF ≈1+(k1c TF,1 ) 1-n (5)

[0156] This invention notes that when n>1 and c TF,1 When fixed, FC TF It is negatively correlated with k1 (i.e., the tendency of TF monomers to spontaneously polymerize into n-mers), as shown in Figure 10b (represented as FC in the figure). TF It is positively correlated with log(1 / k1). Under this condition, FC in the same species TF The order is mainly determined by k1. For example, in B. subtilis, LasR(k1~10) -1 RPU -1 CinR(k1~10) in C. glutamicum -1 RPU -1 The k1 of TF 1 was significantly higher than that of other TFs (<10). -5 RPU -1 This is directly reflected in the unusually high leakage expression of both (Figure 8). It is evident that FC... TF This directly links the intrinsic property of TF (spontaneous aggregation tendency) to its ability to tolerate adverse external conditions (TF overexpression). Therefore, FC TF It can be used not only as a performance indicator during TF screening, but also to indicate its spatial structural characteristics during TF modification, thereby guiding the structural optimization process.

[0157] Finally, this invention proposes a general TF performance optimization problem. This optimization is based on a parameterized TF model and revolves around two externally controllable variables, namely E... P and c TF And then expand. For each fixed parameter (activation type: E) A E AP k1, k2, and n; Suppression type: E R Given a TF model M (f1, f2, ..., k1), this invention defines a function F as an objective function containing multiple TF performance indicators (f1, f2, ...): F = f(f1(E) P ,c TF ;M),f2(E P ,c TF ;M),...) (6)

[0158] Therefore, F is also E P and c TFThe function is given by M. This invention aims to find the combination of variables (E) that globally maximizes F. P * ,c TF * ):

[0159] Where ε contains multiple E P A discrete set of values, corresponding to a sequence of promoters formed by a specific motif (such as the -10 and -35 regions) combined with a specific context; c TF,max This represents the maximum possible expression level of TF in the experimental system. Ideally, considering I... TX For c TF and c I The monotonicity and nonlinearity of c T and c I Orthogonal transformation, E P When n is known and fixed, the model parameters (E) A / E R E AP The values ​​of k1 and k2 should be uniquely determined from a set of promoter strength data (i.e., the "initial guess"). However, real-world transcriptional regulatory systems contain noise, I TX Random errors also exist during the measurement process, which makes I-based... TX The fitted model parameters have uncertainties. Therefore, it is necessary to construct a new control scheme (E) using Equation 6. P * and c TF * After acquiring new transcription intensity data, data accumulation and model updates (i.e., the "learning" phase) are performed, thus completing the DBTL closed loop (Figure 11a). After refitting the model using the original and new data, this invention again adjusts the model parameters based on the new parameters regarding E. P and c TF Optimize F; this invention repeats this iterative process until E is reached. P * Convergent and F with respect to c TF * No significant improvement (<2x) or the number of iterations exceeds the predetermined limit (3 times). This invention uses LuxR in B. subtilis as an example to illustrate this iterative optimization process (b–e in Figure 11). Here, this invention considers the optimization objective F. Max*FC That is, the product of maximum transcription intensity and conventional induction fold:

[0160] Where c I,max This is the maximum usable inducing agent concentration in the experimental system. The reason this invention chose F... Max*FCThis optimization objective is because it balances two often conflicting goals in induced expression systems: increasing c I The maximum transcription strength is >>0, while c decreases. I The leakage expression when F = 0. Therefore, F Max*FC This comprehensively reflects TF in a given E P c TF and c I The invention begins with the promoter S0, which contains the motif combination TTTACA–TATAGT(E). P,0 = -3.24k B T), this combination was used in the previous section for the analysis of LuxR (Figure 7a; corresponding to the red dots in Figure 11c). Based on the model M0 fitted to the first test data, and the available motif combinations in the B. subtilis promoter library (Figure 5b), this invention obtained an optimized promoter sequence S1 (Table 5), which contains the motif combination TTGACC–GATAAT(E P,0 * = -5.56k B T), and optimized TF expression level (c TF,0 * =2.2 RPU), this level has reached the preset upper limit (c TF,0 * =2.2RPU) Therefore, the present invention constructed promoter S1 and determined its c TF Approximately 2 RPU, c I 0 and c respectively I,max The transcription intensity was measured, thus obtaining a new set of promoter strength data (Figure 11b and Figure 11c, blue dots). This data was then merged with the original dataset (Figure 11b and Figure 11c, red dots) and the model was retrained to obtain model M1 (Figure 11b and Figure 11c, red and blue dots). Subsequently, this invention used M1 for a new round of optimization; however, this optimization resulted in E... P (E P,1 * ) and E P,0 * Same, optimized C TF (c TF,1 * ) and c TF,0 * This is consistent with the results, indicating that the optimization process has converged. To further verify the rationality of this optimization strategy and the optimized objective function F1 = F(E... P,1 * ,c TF,1 *It is indeed the global maximum value. This invention additionally constructs promoters S2, S3, and S4 (Table 5), and makes E P,2 <E P,1 <E P,3 <E P,0 <E P,4 And the experiment confirmed that F2 <f1>F3>F0>F4 (Figure 11b).

[0161] Figure 11. Model-based optimization of transcription factor performance. In Figure 11a, for E... P and c TF The general workflow for iteratively optimizing TF performance. Figure 11b shows the induction curves of the LuxR system under different C6-HSL inducer gradients, measured in the first round (red dots) and the second round (blue dots), respectively. The original promoter derived from Vibrio fischeri was also measured (grey dots); the model's prediction results are represented by solid lines. Figure 11c–e show the optimization paths of the LuxR model in *B. subtilis*, mapped to c respectively. TF Dimension (c) or E P In dimension (d), F Max*FC This is shown in heatmap (e). In graph (d), I TX The minimum detectable threshold was set to 0.001 RPU. This was done to reliably quantify the minimum Ig of the B. subtilis promoter through experimental setup. TX In Figure (d), the model's predicted values ​​are truncated below 0.01 RPU. The vertical and horizontal dashed lines in the figure represent the optimal E values ​​at the end of the second round of optimization. P and c TF The arrow indicates the optimal path (e).

[0162] Finally, this invention displays the variable trajectory during the optimization process in E. P ×c TF In space (e in Figure 11), where the objective function F Max*FC The value is calculated using M1 as the reference model. From F... Max*FC The color gradient is clearly visible; to optimize the target, the initial guess value (E) should be... P,0 Reduce E on the basis of ) P (i.e., using a weaker promoter), while in E P At lower (<-2kBT) values, c TF It should be kept at its maximum value. This conclusion not only aligns with the intuitive judgment of the present invention in the previous section, but also provides more precise guidance for promoter design in a quantitative manner. By comparing each type of transcription factor in the three chassis, the present invention notes that among activating transcription factors, LuxR, as a representative quorum sensing transcription factor, compared to before optimization, F Max*FC The improvement was 1.8–3.2-fold; while the common activating transcription factor AraC showed only a slight improvement, approximately 0.7–1.1-fold (Figure 12). For repressive transcription factors, optimizing LacI in E. coli to make its F Max*FC It increased by approximately 60-fold, while in B. subtilis and C. glutamicum, the LacI system was optimal even in the first round of promoter optimization (Figure 12). Other activating transcription factors in F Max*FC Different degrees of improvement were also observed, with the most significant improvement observed in the RpaR system of B. subtilis, which showed an improvement of approximately 20 times (Figure 12).

[0163] Figures 12a-12c are schematic diagrams of the optimized regulatory pathways. Figure 12a is a schematic diagram of the transcription factor optimization pathway in E. coli. Figure 12b is a schematic diagram of the transcription factor optimization pathway in B. subtilis. Figure 12c is a schematic diagram of the transcription factor optimization pathway in C. glutamicum. For each transcription factor, the optimization target (F) is shown on the left. Max*FC Two-dimensional heat map (E) P ×c TF The middle section shows the induction curves (I) of the corresponding promoter before optimization (blue) and after optimization (red). TX Compared to c I The right side shows the optimized maximum F-value (i.e., Frelationship), and the "original" promoter (gray); the right side shows the optimized maximum F-value (i.e., Frelationship). Max*FC ), and the corresponding optimal promoter binding energy (E P ) and optimal transcription factor expression levels (c TF The bottom shows the model parameters at the end of the optimization, including the energy-related term E. P E A E R and E P , with k B T is the unit; k1 is in RPU (Round-PU) 1-n Units: k2 (RPU) 1-n ·mol -n ·L n As a unit, F Max*FC Using the prediction of maximum transcriptional activity (I) max Normalize.

[0164] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0165] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0166] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The system embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of units or modules may be electrical or other forms.

[0167] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0168] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0169] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0170] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for model-based regulation of prokaryotic transcription, characterized in that Includes the following steps: Obtain the DNA element information required for each group of transcriptional regulatory systems in the corresponding test strains, and construct the transcriptional regulatory system; Establish a biophysical quantitative model describing the transcriptional regulatory system; Obtain valid test data for fitting quantitative biophysical models; Use valid test data to train the biophysical quantitative model.

2. The model-based prokaryotic transcription regulation method according to claim 1, wherein, DNA element information includes transcription factor sequences and inducible promoter operator sequences.

3. The model-based prokaryotic transcription regulation method of claim 1, wherein, Valid test data include promoter background transcription intensity, transcription factor expression quantification, and concentration of exogenous inducers.

4. The model-based prokaryotic transcription regulation method of claim 1, wherein, Also includes: Biophysical quantitative models provide the optimal expression levels of transcription factors; or Biophysical quantitative models offer new, higher-performance promoter-TF combinations for constructing transcriptional regulatory systems; or Multiple iterations of the biophysical quantitative model were performed to generate multiple promoter-TF combinations to optimize the performance of the transcriptional regulatory system.

5. The model-based prokaryotic transcription regulation method of claim 1, wherein, Information about DNA elements can be obtained through direct synthesis or amplification of a specific DNA template.

6. The model-based prokaryotic transcription regulation method of claim 2, wherein, Promoters include constitutive promoters, activating promoters, and repressive promoters.

7. The model-based prokaryotic transcription regulation method of claim 6, wherein, In the design of libraries of promoters of different strengths, rational design can be used to σ 70 Core promoter libraries comprising orthogonal combinations of -10 region and -35 region variants recognized by members of the σ family.

8. The model-based prokaryotic transcription regulation method of claim 1, wherein, By extracting plasmids, sequencing DNA, and verifying enzyme digestion, the correct vector was obtained and then transformed into the corresponding chassis cells to obtain the corresponding test strain.

9. The model-based prokaryotic transcription regulation method of claim 1, wherein, The transcriptional regulatory system was constructed, and the transcriptional intensity of the system was calculated and tested.

10. A model-based prokaryotic transcription regulation system, characterized by, include: The system construction unit is used to obtain the DNA element information required for each group of transcriptional regulatory systems in the corresponding test strain and to construct the transcriptional regulatory system. The model building unit is used to build a biophysical quantitative model describing the transcriptional regulatory system. The data acquisition unit is used to acquire valid test data for fitting biophysical quantitative models. The model training unit is used to train a biophysical quantitative model using valid test data.

Images