Method and device for directed evolution of enzyme

Abstract

The present invention provides a method and device for directed evolution of an enzyme, and a computer-readable storage medium. The directed evolution method comprises the following steps: determining multiple single-point mutation sites that improve the performance of a starting enzyme; generating multiple combination mutations based on each of the single-point mutation sites; constructing a corresponding combination mutation description tensor based on the multiple combination mutations; constructing a performance description matrix based on the catalytic activity, specificity and tolerance of the starting enzyme and each of the combination mutations; and performing mathematical fitting based on the combination mutation description tensor and the performance description matrix to determine a solution that conforms to the evolutionary direction. By performing these steps, the directed evolution method can efficiently, reliably and objectively construct and screen mutants through a digital method, thereby determining an evolved enzyme with significantly improved performance in a shorter cycle and at a lower cost.

Description

Technical Field

[0001] The present invention relates to enzyme directed evolution technology, and in particular to an enzyme directed evolution method, an enzyme directed evolution device, and a computer-readable storage medium. Background Art

[0002] Enzymes are macromolecules with catalytic effects in organisms, and their chemical composition is usually proteins, ribonucleic acids, or complexes of them with small organic molecules and metal ions. Enzyme-catalyzed reactions can often reduce steps and achieve higher atomic economy and yield than pure organic chemical synthesis, and the enzymes themselves are degradable and can be obtained from the biological world, and are renewable resources. Therefore, the industry generally regards enzyme catalysis technology as a green chemical technology. However, natural enzymes often cannot adapt to the structures of non-natural substances such as drug intermediates and active pharmaceutical ingredients (APIs), and enzyme-catalyzed reactions are generally only carried out under normal temperature, normal pressure, aqueous solution, and neutral pH conditions. Once there are more than two industrial production reaction conditions such as temperature higher than room temperature, acidic or alkaline pH, high substrate concentration, high proportion of organic solvents, and heterogeneous systems, enzymes often find it difficult to show their superior performance under natural conditions.

[0003] In order to improve the performance of enzymes under the above-mentioned industrial production reaction conditions, a series of directed evolution technologies for enzymes have been proposed in the art, using molecular biological techniques to introduce one or more differences in the molecular structure of the enzyme, and replacing one or more of the amino acid residues constituting the protein with other different amino acids to perform mutations. The enzymes with one or more changes in the molecular structure thus obtained are called mutants. Afterwards, technicians perform different mutations on a previous generation enzyme, and then screen the combination of a large number of mutants obtained under specific reaction conditions to find out the offspring enzymes with improved performance than the previous generation enzymes. By constantly repeating the above-mentioned mutation and screening process, and after a limited number of iterations, performance improvements can be continuously accumulated to obtain offspring enzymes with significantly improved performance than the earliest previous generation enzyme (i.e., ancestral enzyme) and suitable for non-natural substrates and non-natural reaction conditions.

[0004] However, the above directed evolution scheme requires the construction and screening of a large number of mutants. The larger the number, the greater the probability of obtaining mutants with greater performance improvement. Therefore, the existing directed evolution technology generally requires repeated and large-scale repetition of the mutation and screening process, which has the disadvantages of long evolution cycle and high evolution cost.

[0005] In order to overcome the above-mentioned defects of the prior art, there is an urgent need in the art for an enzyme directed evolution technology for efficiently, reliably and objectively constructing and screening mutants, so as to determine evolved enzymes with significantly improved performance in a shorter cycle and at a lower cost. Summary of the invention

[0006] A brief summary of one or more aspects is given below to provide a basic understanding of these aspects. This summary is not an exhaustive overview of all conceived aspects, and is neither intended to identify the key or decisive elements of all aspects nor to define the scope of any or all aspects. Its only purpose is to give some concepts of one or more aspects in a simplified form as a prelude to a more detailed description that will be given later.

[0007] In order to overcome the above-mentioned defects in the prior art, the present invention provides a method for directed evolution of an enzyme, a device for directed evolution of an enzyme, and a computer-readable storage medium.

[0008] Specifically, the directed evolution method of the above enzyme provided according to the first aspect of the present invention includes the following steps: determining multiple single-point mutation sites that improve the performance of the starting enzyme; generating multiple combination mutations based on each of the single-point mutation sites; constructing a corresponding combination mutation description tensor S based on the multiple combination mutations; constructing a performance description matrix P based on the catalytic activity, specificity and tolerance of the starting enzyme and each of the combination mutations; and performing mathematical fitting based on the combination mutation description tensor S and the performance description matrix P to determine a solution that conforms to the evolutionary direction. By performing these steps, the directed evolution method can efficiently, reliably and objectively construct and screen mutants through digital methods, thereby determining an evolved enzyme with significantly improved performance in a shorter cycle and at a lower cost.

[0009] In addition, the directed evolution device of the above-mentioned enzyme provided according to the second aspect of the present invention comprises a memory and a processor. The processor is connected to the memory and is configured to implement the directed evolution method of the above-mentioned enzyme provided by the first aspect of the present invention. By implementing the directed evolution method, the directed evolution device can efficiently, reliably and objectively construct and screen mutants by digital methods, thereby determining the evolved enzyme with significantly improved performance with shorter cycles and lower costs.

[0010] In addition, the above-mentioned computer-readable storage medium provided according to the third aspect of the present invention has computer instructions stored thereon. When the computer instructions are executed by the processor, the directed evolution method of the above-mentioned enzyme provided by the first aspect of the present invention is implemented. By implementing the directed evolution method, the computer-readable storage medium can efficiently, reliably and objectively construct and screen mutants by digital methods, thereby determining the evolved enzyme with significantly improved performance with shorter cycles and lower costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above features and advantages of the present invention can be better understood after reading the detailed description of the embodiments of the present disclosure in conjunction with the following drawings. In the drawings, the components are not necessarily drawn to scale, and components with similar related properties or features may have the same or similar reference numerals.

[0012] Figure 1 A schematic flow chart of a directed evolution method of an enzyme provided according to some embodiments of the present invention is shown.

[0013] Figure 2 A schematic diagram of a process for determining a single point mutation site according to some embodiments of the present invention is shown.

[0014] Figure 3 A schematic diagram of a process for determining a target primer sequence according to some embodiments of the present invention is shown.

[0015] Figure 4 A schematic diagram of a positive determination provided according to some embodiments of the present invention is shown.

[0016] Figure 5 A schematic flow chart of a directed evolution method of an enzyme provided according to some embodiments of the present invention is shown.

[0017] Figure 6 A schematic diagram of a process for determining combined mutations that conform to an evolutionary direction according to some embodiments of the present invention is shown.

[0018] Figure 7 A schematic diagram of a process for determining combined mutations that conform to an evolutionary direction according to some embodiments of the present invention is shown.

[0019] Figure 8 A comparative schematic diagram of the directed evolution process of enzymes provided according to some embodiments of the present invention is shown.

[0020] Fig. 9 A schematic flow chart of a directed evolution method of an enzyme provided according to some embodiments of the present invention is shown.

[0021] Fig.10 A schematic flow chart of a directed evolution method of an enzyme provided according to some embodiments of the present invention is shown. DETAILED DESCRIPTION

[0022] The following specific embodiments illustrate the implementation of the present invention, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Although the description of the present invention will be introduced in conjunction with the preferred embodiment, this does not mean that the features of this invention are limited to this implementation. On the contrary, the purpose of introducing the invention in conjunction with the implementation is to cover other options or modifications that may be extended based on the claims of the present invention. In order to provide a deep understanding of the present invention, the following description will include many specific details. The present invention can also be implemented without using these details. In addition, in order to avoid confusion or blurring the focus of the present invention, some specific details will be omitted in the description.

[0023] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0024] It is understood that although the terms "first", "second", "third", etc. may be used herein to describe various components, regions, layers and / or parts, these components, regions, layers and / or parts should not be limited by these terms, and these terms are only used to distinguish different components, regions, layers and / or parts. Therefore, the first component, region, layer and / or part discussed below may be referred to as the second component, region, layer and / or part without departing from some embodiments of the present invention.

[0025] As mentioned above, the directed evolution of enzymes requires the construction and screening of a large number of mutants. The larger the number, the greater the probability of obtaining mutants with greater performance improvement. Therefore, existing directed evolution techniques generally require repeated and large-scale repetition of the mutation and screening process, which has the disadvantages of long evolution cycle and high evolution cost.

[0026] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method for directed evolution of an enzyme, a device for directed evolution of an enzyme, and a computer-readable storage medium, which can efficiently, reliably and objectively construct and screen mutants through digital methods, thereby determining evolved enzymes with significantly improved performance in a shorter cycle and at a lower cost.

[0027] In some non-limiting embodiments, the directed evolution method of the above-mentioned enzyme provided by the first aspect of the present invention can be implemented by the directed evolution device of the above-mentioned enzyme provided by the second aspect of the present invention. Specifically, a memory and a processor can be configured in the directed evolution device. The memory includes but is not limited to the above-mentioned computer-readable storage medium provided by the third aspect of the present invention, on which computer instructions are stored. The processor is connected to the memory and is configured to execute the computer instructions stored on the memory to implement the directed evolution method of the above-mentioned enzyme provided by the first aspect of the present invention.

[0028] The working principle of the directed evolution device of the above enzyme will be described below in conjunction with some embodiments of the directed evolution method of enzymes. It will be appreciated by those skilled in the art that these directed evolution methods are only some non-restrictive embodiments provided by the present invention, which are intended to clearly demonstrate the main concept of the present invention and provide some specific schemes that are convenient for the public to implement, rather than being used to limit the full functions or all working modes of the directed evolution device. Similarly, the directed evolution device is also only a non-restrictive embodiment provided by the present invention, and does not constitute a limitation on the implementation subject of each step in these directed evolution methods.

[0029] Please refer to Figure 1 , Figure 1 A schematic flow chart of a directed evolution method of an enzyme provided according to some embodiments of the present invention is shown.

[0030] like Figure 1 As shown in the figure, the directed evolution of enzymes can generally be divided into two stages: single-point mutation and combined mutation. The single-point mutation stage is mainly used to identify and confirm the single-point mutation sites that can improve the performance of the starting enzyme, while the combined mutation stage is to combine these single-point mutation sites and find the combination with the highest comprehensive performance improvement to determine the evolved enzyme that meets the evolutionary direction.

[0031] In the above single-point mutation stage, the directed evolution device can first perform step [A1] to determine multiple single-point mutation sites that can improve the performance of the starting enzyme as a basis for the subsequent combined mutation stage. In some embodiments, the above step [A1] can be implemented via artificial intelligence (AI) based on a pre-built relationship model.

[0032] Please refer to Figure 2In some embodiments, the process of mutation site selection (step [A1]) can be implemented in two stages: offline training and online prediction. In the offline training stage, the trainer can first collect a large amount of enzyme evolution case data to construct a training set of the relationship model M. Afterwards, the trainer can obtain multiple starting enzyme samples from the training set and construct a model that approximates the real reaction based on its amino acid sequence, substrate or product structure, and target conditions of the enzyme-catalyzed reaction (step [A1a]).

[0033] The main components of the above-mentioned model may include ligands, receptors and solvent boxes. The ligands are usually small molecules that bind to enzyme molecules, including but not limited to substrates, products, coenzymes, metal ions, bound water, bound ions, surfactants, glycosyls, activators, inhibitors, etc. The receptors are usually enzyme proteins themselves, and the types of enzymes involved may be single protein molecules, or they may be polymers composed of two or more identical or different protein molecules through van der Waals forces, hydrophobic interactions, disulfide bonds or coordination bonds. The solvent box is usually a three-dimensional model composed of the solvent used under the conditions where the enzyme catalyzed reaction occurs, including all other molecules and ions in the solution except ligands and receptors, and their stacking methods.

[0034] In the process of generating a model based on the amino acid sequence of the starting enzyme sample, the trainer can first perform a receptor modeling operation. Please refer to Table 1. Currently, there are a variety of experimental methods, commercial software, and open source software that can generate receptor structure models that meet the requirements, all of which are applicable to the present invention and will not be described in detail here.

[0035] Table 1

[0036]

[0037] Afterwards, the technician can also perform complex modeling operations, that is, generate a model of the receptor molecule model and the ligand molecule combined according to the transition state of the catalytic reaction. Currently, there are a variety of commercial software and open source software such as GOLD, DS LigandFit, DS LibDock, DS CDocker, DS Flexible Docking, SYBYL Surflex-Dock, AutoDock, etc. that can generate receptor structure models that meet the requirements, which are all applicable to the present invention and will not be described in detail here.

[0038] After that, technicians can also perform solvation operations, that is, according to the actual reaction conditions, a large number of solvent molecules (water molecules, organic co-solvent molecules), ions (buffer salt ions, hydronium ions, hydroxide ions, metal ions, etc.), and other required molecules are stacked into a cube or other shape according to a certain proportion and rule, and the aforementioned complex model is placed in this solvent box. Similarly, there are currently a variety of commercial software and open source software such as AMBER LEaP, PackMol, DS Solvation, etc. that can complete this solvation process, which will not be described in detail here.

[0039] like Figure 2 As shown, after completing step [A1a] and obtaining the structural model based on the amino acid sequence of the starting enzyme sample, the technician can simulate each starting enzyme sample separately and determine the structural description matrix S' of each starting enzyme sample according to the simulation results.

[0040] The main theories of the above-mentioned simulation can include molecular mechanics (MM) and quantum mechanics (QM). The former mainly sets molecular force fields (such as AMBER, CHARMm, Opls, GMX, etc.), and calculates the motion of atoms and ions in these molecular force fields. The latter mainly calculates density functional theory (DFT). In addition, the calculation method of the simulation can include Monte Carlo simulation (MC), molecular dynamics simulation (MD), local quantum mechanics / local molecular mechanics-dynamics simulation (QM / MM-MD), etc. At present, there are various commercial software and open source software such as CHARMm, AMBER, GROMACS, NAMD, GAUSSIAN, DS Dynamics, DS QM-MD to realize the above-mentioned calculation method, which will not be repeated here.

[0041] In the process of simulating each starting enzyme sample, the trainer can first obtain trajectory data from the simulation data. The trajectory data includes but is not limited to the position, velocity, acceleration and other data of each atom or each amino acid residue, ligand, solvent and other components in the simulation process. By processing these data, a series of important results can be derived, among which the root mean square deviation (RMSD) can describe the deviation of each structural unit (such as amino acid residue) in the simulation system relative to the reference position (usually the center of mass of the complex), and the root mean square fluctuation (RMSF) and B-factor describe the jitter amplitude of each structural unit (such as amino acid residue) relative to the equilibrium position in the simulation system.

[0042] Specifically, RMSD is calculated as the square root of the average of the sum of the squares of the deviations of the positions of all structural units constituting the system (such as all amino acid residues of the receptor protein molecule) relative to the reference position (such as the center of mass of the complex) at any time, that is,

[0043]

[0044] In the formula, R t is the RMSD value of the receptor at time t, and the vector r i,t is the position coordinate of amino acid i at time t, vector is the reference position.

[0045] It can be seen that R t It changes with time t, that is, different frames of simulation have different RMSD. Therefore, the present invention can use the average RMSD corresponding to all frames of simulation, that is,

[0046]

[0047] In addition, RMSF is calculated as the square root of the average value of the square of the position offset of a particle within time T and over time. For the i-th amino acid residue in a protein molecule, the calculation of its RMSF can be written as the integral of the interval (0, T) in the time domain, that is,

[0048]

[0049] In the formula, R i is the RMSF value of amino acid i, vector r i,t is the position coordinate of amino acid i at time t, vector is the reference position of this amino acid.

[0050] When processing simulation data, the above integral can be replaced by the following numerical calculation formula, namely:

[0051]

[0052] Among them, R i is the RMSF value of amino acid i, usually in units of (i.e. 10 -10 m), vector r i,f is the position coordinate of amino acid i in the simulation frame f, and vector r i,0 are the position coordinates of amino acid i in the simulation frame 0.

[0053] In addition, the B-factor, also known as the temperature factor, is used to quantify the uncertainty of each atom in the protein. The B-factor has the following mathematical relationship with the RMSF:

[0054] B i =(8 / 3)·(πR i ) 2 [Formula 1]

[0055] In addition, trainers can also obtain energy data from simulation data, that is, the binding free energy of the ligand (mainly substrate or product) molecule and the receptor (enzyme) molecule to form a transition state complex in a simulated solvent environment, that is,

[0056]

[0057] Among them, the superscript vac indicates the state in vacuum, the superscript sol indicates the solvated state, the subscript cmp indicates the complex, the subscript lig indicates the ligand, and the subscript rec indicates the receptor.

[0058] Furthermore, based on different force fields, there are many different binding free energy calculation methods for each amino acid residue i, such as molecular mechanics / Poisson-Boltzmann surface area implicit solvation (MM / PBSA), molecular mechanics / generalized Born surface area implicit solvation (MM / GBSA), thermodynamic integration (TI), free energy perturbation (FEP), etc. At present, FEP is recognized as a method with high accuracy in free energy calculation, but the demand for computer computing power is also relatively high. MM / PBSA and MM / GBSA are methods with a relatively compromise in computing power and accuracy.

[0059] With the help of these algorithms, the binding free energy of the receptor-ligand can usually be decomposed into each amino acid residue, that is, the binding free energy of the amino acid residue in each receptor molecule to the ligand can be calculated, that is,

[0060]

[0061] Among them, ΔG and ΔG i The unit is kJ / mol.

[0062] Furthermore, in some embodiments of the present invention, the simulation of each starting enzyme sample can be divided into two rounds, namely, control simulation (step [A1b]) and target simulation (step [A1b']). The control simulation can be performed based on the optimal starting conditions before the directed evolution of the starting enzyme sample, and the target simulation can be performed based on the target conditions of the directed evolution of the starting enzyme sample, and the condition parameters are shown in Table 2.

[0063] Table 2

[0064]

[0065] The trainer can calculate the key trajectory data B of each amino acid residue i in each starting enzyme sample in the control simulation. i and critical energy ΔG i , and the key trajectory data B of each amino acid residue i of each starting enzyme sample in the target simulation i ′ and critical energy ΔG i′, and then by comparing the corresponding data of the corresponding amino acids in the two rounds of simulation, the influence of the amino acid residue i at the corresponding position on a certain performance of the enzyme can be calculated.

[0066] Specifically, the trainer can first use the trajectory data B i And trajectory data B i ', determine the trajectory descriptor β i ,Right now

[0067]

[0068] In addition, the trainer can use the energy data ΔG i and energy data ΔG i ', determine the energy descriptor ε i ,Right now

[0069] ε i =(ΔG′ i -ΔG i ) / RT [Formula 5]

[0070] Wherein, molar gas constant R = 8.314 J / mol / K, T is thermodynamic temperature, unit K.

[0071] Afterwards, the trainer can use the trajectory descriptor β i and the energy descriptor ε i , determine the structural descriptor of the ith amino acid residue, that is,

[0072] s i =[β i ,ε i ] [Formula 6]

[0073] Afterwards, the trainer can form a structural description matrix with all the structural descriptors from the first amino acid residue to the last amino acid residue of the protein constituting the starting enzyme sample, that is,

[0074] S=[s1,s2,…,s i ,…,s n-1 ,s n ] [Formula 7]

[0075] In addition, trainers can also follow the following enzyme catalyzed reactions

[0076]

[0077] And related experimental data, construct the performance description matrix P of the starting enzyme sample.

[0078] Specifically, in the above enzyme-catalyzed reaction, the differential equation for the change of substrate concentration with time is

[0079]

[0080] In the formula, K is the reaction rate constant, and the negative sign indicates that the substrate is continuously reduced during the reaction. It is proportional to the concentration of active enzyme in the reaction and the catalytic activity of the unit enzyme, that is,

[0081] K=k·[Active Enzyme] [Formula 8b]

[0082] Where [Active Enzyme] is the concentration of active enzyme, k is the specific catalytic rate constant, and the unit reflects the catalytic activity of the enzyme.

[0083] Considering the situation where the enzyme is continuously inactivated during the reaction,

[0084]

[0085] Where D is the enzyme inactivation rate constant, the change of the remaining active enzyme concentration over time can be expressed as the following differential equation:

[0086]

[0087] In the formula, the negative sign indicates that the active enzyme decreases continuously during the reaction.

[0088] After that, by integrating the above formula, we can get

[0089]

[0090] Then, substitute [Formula 8b] and [Formula 8c] into [Formula 7a] and perform definite integral to obtain

[0091]

[0092] Next, consider the following enzyme-catalyzed reaction:

[0093]

[0094]

[0095]

[0096] …

[0097]

[0098] The specificity of the enzyme can be expressed as

[0099]

[0100] For each enzyme mutant, assuming that its ith amino acid residue mutates, then among all 20 possible mutations (there are 20 natural amino acids that make up proteins), the optimal single-point mutation can be experimentally measured, for example, the one with the largest specific catalytic rate constant k i (the bigger the better), maximum specificity E i (larger is better) and / or minimum inactivation rate constant D i (The smaller the better). The catalytic rate constant k i Indicates the catalytic activity after the amino acid residue i is mutated. i Indicates the specificity of amino acid residue i after mutation. The inactivation rate constant D i Indicates the tolerance of amino acid residue i after mutation. In this way, the performance descriptor of the mutation of the i-th amino acid residue can be described as

[0101] p i =[κ i ,ι i ,δ i ] [Formula 9]

[0102] In the formula, the catalytic activity descriptor κ i =ln(k i / k0), specificity descriptor ι i =ln(E i / E0), tolerance descriptor δ i =ln(D i / D0), and the parameters with subscript 0 represent the parameters of the starting enzyme sample without mutation.

[0103] Afterwards, the trainer can form a performance description matrix with all the performance descriptors from the first amino acid residue to the last amino acid residue of the protein constituting the starting enzyme sample, i.e.

[0104] P=[p1,p2,…,p i ,…,p n-1 ,p n ] [Formula 10]

[0105] like Figure 2 As shown, after completing steps [A1b] and [A1b'] and obtaining the structure description matrix S and performance description matrix P of each starting enzyme sample, the trainer can define a mathematical mutation success rate function M that satisfies

[0106] P=M(S) [Formula A1]

[0107] And through machine learning, we train a numerical approximate solution M of the mutation success rate function M #(Step [A1c]). Afterwards, the trainer can approximate the solution M based on the value obtained during training. # , construct a relationship model between each single point mutation site and enzyme performance.

[0108] Those skilled in the art will appreciate that the above trainer is merely a non-limiting description, including but not limited to technicians who perform corresponding steps, and / or computers, controllers, processors, robotic arms and other equipment that execute corresponding computer instructions.

[0109] Please continue to refer to Figure 2 After completing the training process of the relational model and obtaining the numerical approximate solution M of the mutation success rate function M # Afterwards, the directed evolution device can obtain the starting enzyme that needs directed evolution from the prediction set in the online prediction stage, and perform structural modeling on it according to the pre-built structural model (step [A1a]). Afterwards, the directed evolution device can simulate the starting enzyme as described above to determine the trajectory data B of each amino acid residue i of the starting enzyme. i and energy data ΔG i , and then according to the trajectory data B i And the energy data ΔG i , identify multiple single-point mutation sites that can improve the performance of the starting enzyme.

[0110] Further, in some embodiments, the above simulation process can be divided into two rounds, namely, control simulation (step [A1b]) and target simulation (step [A1b']). Please refer to Table 2. The directed evolution device can perform control simulation on the starting enzyme according to the starting conditions to determine the first trajectory data B of each amino acid residue i of the starting enzyme. i and the first energy data ΔG i In addition, the directed evolution device can also perform target simulation on the starting enzyme according to the target condition to determine the second trajectory data B of each amino acid residue i of the starting enzyme. i ' and the second energy data ΔG i Here, the starting conditions can be the optimal reaction conditions before starting the directed evolution of the enzyme, including but not limited to the natural conditions for the natural enzyme, and for the intermediate enzyme (i.e., the evolved enzyme that has not yet reached the desired level) after at least one directed evolution iteration. The target simulation can be the reaction conditions determined based on the current round of directed evolution target or the overall target of directed evolution.

[0111] After completing the control simulation (step [A1b]) and the target simulation (step [A1b']), the directed evolution device can calculate the first trajectory data B based on the above [Formula 4]. i And the second trajectory data B i'Determine the trajectory descriptor β i , and based on the above [Formula 5], according to the first energy data ΔG i and the second energy data ΔG i 'Determine the energy descriptor ε i Afterwards, the directed evolution device can be based on the above [Formula 6] according to the trajectory descriptor β i and the energy descriptor ε i Determine the structural description matrix S, and combine it with the numerical approximate solution M of the mutation success rate function M obtained in advance # , and identified multiple single point mutation sites.

[0112] Specifically, the directed evolution device can firstly solve M according to the structural description matrix S and the numerical approximation solution # , make AI decisions based on the relational model to determine the corresponding prediction performance description matrix P # ,Right now

[0113]

[0114] The prediction performance description matrix P # Includes multiple prediction performance descriptors p i # ,Right now

[0115]

[0116] Each prediction performance descriptor p i # They respectively indicate the catalytic activity, specificity and tolerance after the amino acid residue i undergoes mutation. i # >1 or ι i # >1 or δ i # >1, the directed evolution device can determine that the mutation of the i-th amino acid residue may lead to improved performance. Afterwards, the directed evolution device can add all amino acid residues i that meet this condition to the mutation library of single-point mutations for selection in the subsequent combined mutation stage.

[0117] Compared with the existing technology, which is difficult to derive the mutation success rate function M through the theoretical formula of enzyme catalysis reaction, the present invention obtains the approximate numerical solution M of the mutation success rate function M through machine learning of a large amount of training data. # , and based on this value, the approximate solution M #AI decision-making is performed to determine multiple single-point mutation sites that can improve the performance of the starting enzyme. Therefore, the present invention can use the results of multiple existing theories or algorithms as the decision-making basis of the relationship model through a digital method, thereby efficiently, reliably and objectively constructing single mutants, and effectively shortening the iteration cycle and economic cost of directed evolution.

[0118] Please continue to refer to Figure 1 After completing the selection of mutation sites (step [A1]) and obtaining multiple single-point mutation sites that improve the performance of the starting enzyme, the directed evolution device can design a single mutation target primer sequence based on at least one target mutation site (step [A2]).

[0119] Please refer to Figure 3 In some embodiments of the present invention, the process of single mutation primer design (step [A2]) can also be implemented in two stages: offline training and online prediction. In the offline training stage, the trainer can first obtain experimental data of single point mutation polymerase chain reaction (PCR) of multiple starting enzyme samples from the training set. Each experimental data includes a primer sequence and the result of the mutation experiment.

[0120] In organisms, there is a definite correspondence between amino acid sequences and base sequences on DNA, which is a carrier of genetic information, and each amino acid corresponds to three bases. Organisms have specific physiological mechanisms that can produce proteins with corresponding amino acid sequences based on DNA base sequences. When one or more changes occur in the DNA base sequence, the amino acids at corresponding positions of all proteins produced by the organism based on such a changed DNA will undergo corresponding and consistent changes. Based on the above principle, by changing the DNA sequence through specific technical means, proteins with corresponding changes in amino acid sequences can be artificially created, which is called site-directed mutagenesis.

[0121] Polymerase chain reaction (PCR) is one of the methods to change DNA sequence. It uses the fact that DNA will become single-stranded when it is denatured at a high temperature of 95°C. At low temperature (usually around 60°C), primers and single-stranded DNA will combine according to the principle of base complementary pairing. Then the temperature is adjusted to the optimal reaction temperature of DNA polymerase (around 72°C), so that DNA polymerase can synthesize complementary chains along the direction from phosphate to pentose (5′→3′). There are five main substances involved in the PCR reaction, namely primers (short DNA chains), DNA polymerase, deoxyribonucleoside triphosphates (dNTPs), templates (DNA to be copied), and buffer (Mg is required). 2+ ).

[0122] The base sequence of the primer is usually exactly the same as a certain base sequence in the template. In this case, the base sequence of the large number of DNA products obtained by PCR is also exactly the same as the template. If there are individual differences between the base sequence of the primer and the corresponding base sequence of the template, this difference will be retained in the new DNA molecules generated by PCR, that is, a mutation occurs at a specific position in the large number of DNA molecules obtained.

[0123] During the training process, the trainer can first digitize the primer base sequence. Specifically, the trainer can first identify the start codon through nucleotide sequences such as AUG, GUG or UUG, and then introduce mutations in the three bases from the i-th to i+2 bases starting from the start codon.

[0124] Thus, the upstream fragment of the forward primer, i.e. fw The base sequence from i-1 to i-1 is

[0125]

[0126] The downstream segment of the forward primer, i.e. from i+2 to i+2+d fw The base sequence is

[0127]

[0128] The upstream fragment of the reverse primer, i.e. from i-1 to iu rv The base sequence is

[0129]

[0130] The downstream fragment of the reverse primer, i.e., from i+2+d rv The base sequence from i to i+2 is

[0131]

[0132] Among them, $b j It is one of the four characters 'a', 'c', 'g', 't', corresponding to the type of base at position j in the entire gene.

[0133] Afterwards, the trainer can digitize the above sequences and convert the character form of $b j Expressed in digital form,

[0134] b j =[a j ,c j ,g j ,t j ]

[0135] Among them, a j ,c j,g j ,t j = 0 or 1, and for any j, a j ,c j ,g j ,t j There is only one 1 in the set and the rest are 0.

[0136] Thus, the upstream and downstream sequences of the forward primer $u fw 、$d fw , and the upstream and downstream sequences of the reverse primer $u rv 、$d rv It can be expressed in digital form as the following matrix

[0137]

[0138]

[0139]

[0140]

[0141] Thus, the reverse primer upstream sequence matrix U for all experiments k is k rv Can form a third-order tensor

[0142] U rv(3) =[U rv 1,U rv 2,…,U rv n-1 ,U rv n ]

[0143] The matrix D of the downstream sequences of the reverse primers for all experiments k k rv Can form a third-order tensor

[0144] D rv(3) =[D rv 1,D rv 2,…,D rv n-1 ,D rv n ]

[0145] The matrix U of the upstream sequences of the forward primers of all experiments k k fw Can form a third-order tensor

[0146] U fw(3) =[U fw 1,U fw 2,…,U fwn-1 ,U n ]

[0147] The matrix D of the downstream sequences of the forward primers of all experiments k k fw Can form a third-order tensor

[0148] D fw(3) =[D fw 1,D fw 2,…,D fwv n-1 ,D n ] [Formula 12]

[0149] In addition, for the kth experiment, its experimental results can also be expressed digitally, that is, r k = 0 or 1, where 0 indicates that the experiment did not obtain the expected mutation, and 1 indicates that the experiment successfully obtained the expected mutation. k Then a vector r = [r1, r2, ..., r n-1 ,r n ].

[0150] After that, the trainer can define a mathematical mutation success rate function Q that satisfies

[0151] r=Q(U rv(3) ,D rv(3) ,U fw(3) ,D fw(3) ) [Formula A2]

[0152] And through machine learning, we train a numerical approximate solution Q of the above mutation success rate function Q # After that, the trainer can approximate the solution of Q based on the value obtained during training. # , construct a mutation success rate model of each single point mutation site and enzyme performance.

[0153] Those skilled in the art will appreciate that the above trainer is merely a non-limiting description, including but not limited to technicians who perform corresponding steps, and / or computers, controllers, processors, robotic arms and other equipment that execute corresponding computer instructions.

[0154] Please continue to refer to Figure 3 , after completing the training process of the mutation success rate model, and obtaining the numerical approximate solution Q of the mutation success rate function Q #Afterwards, the directed evolution device can determine at least one target mutation site in the online prediction stage according to the multiple single-point mutation sites i that can improve the performance of the starting enzyme determined in step [A1]. Afterwards, the directed evolution device can introduce mutations at the three bases from the i-th to the i+2-th bases of the entire gene, starting from the start codon, and replace these three bases with the codons of the target amino acid residue to be mutated according to the codon preference of the corresponding target host to obtain the mutated gene. Here, different types of amino acid residues can correspond to different codons. Taking Escherichia coli as an example, please refer to Table 3.

[0155] Table 3

[0156] Amino acid residues Forward codon Reverse codon X(any amino acid residue) 5′-NNK-3′ 5′-MNN-3′ A (alanine residue) 5′-GCG-3′ 5′-CGC-3′ C (cysteine ​​residue) 5′-TGC-3′ 5′-GCA-3′ D (aspartic acid residue) 5′-GAT-3′ 5′-ATC-3′ E (glutamic acid residue) 5′-GAA-3′ 5′-TTC-3′ F (phenylalanine residue) 5′-TTT-3′ 5′-AAA-3′ G (glycine residue) 5′-GGC-3′ 5′-GCC-3′ H (histidine residue) 5′-CAT-3′ 5′-ATG-3′ I (isoleucine residue) 5′-ATT-3′ 5′-AAT-3′ K (lysine residue) 5′-AAA-3′ 5′-TTT-3′ L (leucine residue) 5′-CTG-3′ 5′-CAG-3′ M (methionine residue) 5′-ATG-3′ 5′-CAT-3′ N (asparagine residue) 5′-AAC-3′ 5′-GTT-3′ P (proline residue) 5′-CCG-3′ 5′-CGG-3′ Q (glutamine residues) 5′-CAG-3′ 5′-CTG-3′ R (arginine residue) 5′-CGC-3′ 5′-GCG-3′ S (serine residue) 5′-AGC-3′ 5′-GCT-3′ T (Threonine residue) 5′-ACC-3′ 5′-GGT-3′ V (valine residue) 5′-GTG-3′ 5′-CAC-3′ W (tryptophan residue) 5′-TGG-3′ 5′-CCA-3′ Y (tyrosine residue) 5′-TAT-3′ 5′-ATA-3′

[0157] The directed evolution device can start from the i-th base in the 5′→3′ sequence of the mutated gene and take u bases forward according to the template sequence and the at least one target mutation site i. fw Bases to get the forward primer upstream sequence $u fw ; Start from the i+2th base in the mutated gene and take d fw Bases to get the forward primer downstream sequence $d fw ; Starting from the ith base in the 3′→5′ complementary sequence of the mutated gene (note, not the reverse complementary sequence of 5′→3′), take u bases forward rv bases and reverse the sequence to get the upstream sequence of the reverse primer $u rv ; Start from the i+2th base in the mutated gene and take d rv Bases and reverse sequence to get the downstream sequence of the reverse primer $d rv .

[0158] Afterwards, the directed evolution device can be used to fw ,d fw 、u rv ,d rv These four optional lengths take values ​​of 0, 1, 2, ..., 50, etc., and exhaust their combinations to generate several groups of $u fw 、$d fw 、$u rv 、$d rv Then, the directed evolution device can perform digital processing on each set of sequences as described above, and approximate the solution Q according to the numerical value of the mutation success rate function Q obtained in advance. # , determine the target primer sequence corresponding to the at least one target mutation site i.

[0159] Specifically, the directed evolution device can convert each set of digitally processed sequences $u fw 、$d fw、$u rv and $d rv Substitute the numerical approximate solution Q # , to calculate the corresponding predicted experimental result r. Here, the predicted experimental result r indicates whether the corresponding combination obtains the expected mutation. Then, the directed evolution device can find out which meets the first preset range (for example: 0.9≤r≤1.0), and u fw ,d fw 、u rv ,d rv The primer sequences that meet the requirements are determined by assembling them in the following manner:

[0160] Forward primer sequence: 5′-$u fw +Forward codon+$d fw -3′

[0161] Reverse primer sequence: 5′-$d fw +Reverse codon+$u fw -3′

[0162] Furthermore, if mutations need to be performed at multiple single points respectively, the directed evolution device can repeat the above process to generate a primer sequence for each single point mutation.

[0163] Compared with the prior art which mainly relies on empirical formulas to design primers, the present invention obtains an approximate numerical solution Q of the mutation success rate function Q by machine learning of a large amount of training data. # , and based on this value, the approximate solution Q # AI decision-making is performed to determine the primer sequence that can generate the corresponding single mutant. Therefore, the present invention can use the results of multiple existing theories or algorithms as the decision basis of the relationship model through a digital method, so as to efficiently, reliably and objectively construct single mutants, and effectively shorten the iteration cycle and economic cost of directed evolution.

[0164] Please continue to refer to Figure 1 After completing the single mutation primer design (step [A2]) and obtaining the target primer sequence corresponding to each single point mutation site, the directed evolution device can construct a single mutant according to the target primer sequence of the single mutation (step [A3]), and perform a catalytic reaction and analysis on it (step [B1]). Those skilled in the art can complete the relevant operations of step [A3] and step [B1] based on the existing technology, which will not be described in detail here. Afterwards, the directed evolution device can make a positive judgment based on the result of the catalytic reaction, and screen multiple single point mutation sites that can improve the performance of the starting enzyme based on the single point mutation sites of multiple positive single mutants (step [B2]).

[0165] Specifically, in the process of making a positive judgment, the directed evolution device can first obtain the result data such as the conversion rate data x and the evolution data y of the catalytic reaction of the single mutant j. Here, the conversion rate data x may include the conversion rate (conv.). The evolution data y may include enantiomeric excess (ee), and may also include a variety of data such as diastereomeric excess value (de), positional isomer excess value (re), and pretreatment residual enzyme activity (ra), which respectively reflect the catalytic activity, specificity and tolerance of the enzyme. However, these data are only direct data of the reaction, and they are non-orthogonal to each other. Therefore, it is usually unreliable to directly derive the intrinsic value of the reaction enzyme performance (for example: the specific reaction rate constant k, specificity constant E, and inactivation rate constant D in step [A1c]) from a large amount of such raw data as a method of judging whether the mutant has improved performance.

[0166] In addition, for the specific case of the above-mentioned non-orthogonality, if the conversion rate (conv.) of a mutant is lower than that of the control, but its enantiomeric excess (ee) is higher than that of the control, the directed evolution device will find it difficult to determine whether the performance of the mutant is improved.

[0167] To this end, for each mutant and control, the prior art needs to measure multiple sets of data, namely, multiple conversion rates (conv.), enantiomeric excess (ee) and pretreatment residual enzyme activity (ra), and draw curves respectively to determine whether the mutant has improved performance compared to the control. However, in the directed evolution of enzymes, each mutant library screening usually contains a large number of mutants (hundreds to tens of thousands), so it is unrealistic to measure a series of data points for each mutant and draw a curve.

[0168] In order to overcome the above defects, in some embodiments of the present invention, the directed evolution device can pre-construct the interaction relationship function f of multiple performance factors and find its numerical approximate solution f # Afterwards, the directed evolution device can approximate the solution f based on the acquired conversion rate data x, evolution data y, and the numerical value of the interaction relationship function f obtained in advance. # , to make a positive judgment on single mutant j.

[0169] Taking the two factors of conversion rate (conv.) and enantiomeric excess (ee) as an example, the directed evolution device can assume that the interaction between the two factors is y = f(x), then mathematically there is a Taylor series expansion of this relationship, that is,

[0170]

[0171] In some embodiments, the directed evolution device can select i≤3, i.e., a curve below cubic, according to the actual accuracy requirement, and perform Taylor series expansion. Afterwards, the directed evolution device can adjust the control experimental conditions (e.g., two groups of control with high enzyme dosage and low enzyme dosage) to obtain two groups of control data under high conversion rate (conv.) and low conversion rate (conv.), and use these two groups of control data for mathematical fitting to obtain an approximate numerical solution f of the above interaction relationship f. # , Right now

[0172] y=f # (x) = λ3x 2 +λ2x 2 +λ1x+λ0

[0173] Thus, for any single mutant j, if

[0174] y j >f # (x j ) or x j >(f # ) -1 (y)

[0175] Then the directed evolution device can determine the data point corresponding to the single mutant j (i.e. Figure 4 The dark circles in the figure are located on the fitted interaction curve f # , thus determining that the single mutant j is a positive single mutant.

[0176] On the contrary, for any single mutant j, if

[0177] y j ≤f # (x j ) and x j ≤(f # ) -1 (y)

[0178] Then the directed evolution device can determine the data point corresponding to the single mutant j (i.e. Figure 4 The light circles or dark boxes in the figure are located on the fitted interaction curve f # above, or to its lower left, thus determining that the single mutant j is a negative single mutant.

[0179] After that, the directed evolution device can screen out the negative single mutants that cannot improve the performance of the starting enzyme to reduce the data processing load in the subsequent combined mutation stage. In addition, the directed evolution device can also solve the following set of equations based on the positive mutants determined by the above rules to calculate the eigenvalues ​​of the starting enzyme performance for use in the subsequent combined mutation stage.

[0180]

[0181] In the formula, conv. (τ)(t) The conversion rate is measured after the initial enzyme is pretreated for τ time and the reaction time is t; ee Product is the enantiomeric excess of the product; [Enzyme]0 is the initial enzyme concentration; k, E, and D are the catalytic rate constant, specificity constant, and inactivation rate constant of the initial enzyme, respectively, that is, the intrinsic value of the initial enzyme performance.

[0182] Compared with the prior art, which finds it difficult to directly derive the eigenvalue indicating the performance of the reaction enzyme based on non-orthogonal experimental data, the present invention assumes the interaction function f of multiple performance factors and obtains the numerical approximate solution f that conforms to the experimental data of the starting enzyme through digital methods such as Taylor series expansion. # , so as to efficiently, reliably and objectively judge the behavior of each single mutant j. Therefore, the present invention can efficiently, reliably and objectively screen single mutants through digital methods, thereby effectively shortening the iteration cycle and economic cost of directed evolution.

[0183] Those skilled in the art will appreciate that the above method is based on the numerical approximation solution M of the mutation success rate function M. # The scheme of making AI decisions to determine multiple single-point mutation sites (step [A1]), and then making positive determinations based on the determined multiple single-point mutation sites (steps [A2] to [B2]) is only a non-limiting implementation method provided by the present invention, which is intended to clearly demonstrate the main concept of the present invention and provide a specific scheme that is easy for the public to implement, but is not intended to limit the scope of protection of the present invention.

[0184] Optionally, in other embodiments, the directed evolution device may skip step [A1] and directly perform the steps of single mutation primer design (step [A2]), mutant construction (step [A3]), catalytic reaction and analysis (step [B1]), and data processing and analysis (step [B2]) based on all sites of the starting enzyme, thereby constructing and screening multiple positive single-point mutation sites that can improve the performance of the starting enzyme with a relatively large data processing load (still far less than the data processing load of the prior art).

[0185] For further information, please refer to Figure 5In some preferred embodiments, when the starting enzyme samples in the training set are sufficient, the prediction accuracy of the AI ​​decision (step [A1]) can meet the accuracy requirements of the directed evolution, and the directed evolution device can also preferably skip the single point mutation and single point mutation screening operations of steps [A2] to [B2], and directly perform the combined mutation stage of the directed evolution based on the multiple single point mutation sites determined by the AI ​​decision (step [A1]). By omitting the single point mutation and single point mutation screening process, the present invention can further save the time and corresponding experimental costs of this part of the experiment.

[0186] like Figure 1 and Figure 5 As shown, after completing the single-point mutation stage of the starting enzyme and determining multiple single-point mutation sites that can improve the performance of the starting enzyme, the directed evolution device can perform a combined mutation stage based on the multiple single-point mutation sites determined in step [A1] or step [B2] to determine at least one combined mutation that can improve the performance of the starting enzyme as a whole (step [A'1]).

[0187] Please refer to Figure 6 , Figure 6 A schematic diagram of a process for determining combined mutations that conform to an evolutionary direction according to some embodiments of the present invention is shown.

[0188] like Figure 6 As shown, in some embodiments, it is assumed that the directed evolution device has identified i1, i2, ..., i j ,…,i m-1 ,i m These m single-point mutation sites that can improve at least one aspect of the enzyme performance, the directed evolution device can be used to modify each site i j Digital processing is performed to convert the site i j Constructed into a vector, namely the mutation descriptor s j , to abstractly describe possible mutations

[0189] s j =[size j ,polarity j ,charge j ,rigidity j ] [Formula 13]

[0190] In the formula, size j Indicates the change in the size of the side chain after mutation, 1 for increase, 0 for no change, and -1 for decrease; polarity j Indicates the change in side chain polarity after mutation, 1 for increase, 0 for no change, and -1 for decrease; charge jIndicates the side chain charge after mutation, 0 if unchanged, -1 if charge reversed; rigidity j Indicates the rigidity of the main chain after mutation, with rigidity taking 1 and flexibility taking 0.

[0191] Thus, when one or more of the m single-point mutation sites mutate, a combined mutation is formed. There are many combinations of these combined mutations, and the total number of possible combined mutations can be expressed as

[0192]

[0193] The directed evolution device can select a representative subset based on an orthogonal table or other methods, which includes l combinations of mutations (m≤l≤L) and are numbered as 1, 2,…, k,…l-1, l.

[0194] Thus, for any k-th combined mutation in the above subset, the mutation status of each site can be described by the following mutation description matrix:

[0195] S k =[s k_1 ,s k_2 ,…s k_j ,…,s k_m-1 ,s k_m ] [Formula 14]

[0196] For all selected combined mutations, the corresponding mutation description matrix S k The mutation description tensor S can be further constructed, namely

[0197] S=[S1,S2,…S k ,…,S l-1 ,S l ] [Formula 15]

[0198] In addition, if Figure 6 As shown, in the process of determining at least one combination mutation that can improve the performance of the starting enzyme as a whole (step [A'1]), the directed evolution device can also perform structural modeling on the unmutated starting enzyme and the various selected combination mutations k as described above, and perform control simulation and target simulation in the manner of steps [A1b][A1b′] to determine the catalytic activity, specificity and tolerance of the starting enzyme and each combination mutation k. Here, the ligand of the control simulation (step [A1b]) can select the product isomer, and the ligand of the target simulation step [A1b'] can select the target product. Afterwards, the directed evolution device can determine the simulation performance descriptors of each combination mutation k based on the catalytic activity parameters, specificity parameters and tolerance parameters determined by the simulation. And according to the simulation performance descriptor of each combination mutation k Construct the simulation performance description matrix P of each combination mutation k # .

[0199] Specifically, the directed evolution device can first perform control simulation and target simulation on the starting enzyme to determine the third energy data ΔG0 and the fourth energy data ΔG′0 of the starting enzyme, and respectively perform control simulation and target simulation on each combination mutation k to determine the fifth energy data ΔG k and the sixth energy data ΔG′ k Afterwards, the directed evolution device can calculate the energy of the fourth energy data ΔG′0 and the sixth energy data ΔG′ k , determine the catalytic activity descriptor of the combined mutation k Right now

[0200]

[0201] Wherein, molar gas constant R = 8.314 J / mol / K, T is thermodynamic temperature, unit K.

[0202] In addition, the directed evolution device can be based on the third energy data ΔG0, the fourth energy data ΔG′0, the fifth energy data ΔG k , and the sixth energy data ΔG′ k , determine the specific descriptor for the combined mutation k Right now

[0203]

[0204] Wherein, molar gas constant R = 8.314 J / mol / K, T is thermodynamic temperature, unit K.

[0205] In addition, the directed evolution device can also calculate the first average RMSD value of the starting enzyme given by the target simulation (step [A1b']) And the first average RMSD value of each combination mutation k Determine tolerance descriptors for combined mutations k Right now

[0206]

[0207] Afterwards, for the kth selected combination mutation, its simulation performance descriptor can be expressed as

[0208]

[0209] The simulation performance descriptors of all the selected l combined mutations are A simulation performance description matrix can be formed, namely

[0210]

[0211] Afterwards, the directed evolution device can define that there is a tensor X=[χ1,χ2,…χ j ,…,χ m-1 ,x m ],satisfy

[0212] P # =S×X [Formula 18]

[0213] In the formula,

[0214]

[0215] χ1,χ2,…χ j ,…,χ m-1 ,x m is the relationship matrix of each single point mutation site j.

[0216] After that, the directed evolution device can consider any mutation descriptor s = [size, polarity, charge, rigidity], and traverse any value allowed by [Formula 13] for the mutated side chain size (size), the mutated side chain polarity (polarity), the mutated side chain charge (charge), and the mutated main chain rigidity (rigidity) respectively, so as to describe the tensor S according to the above combination mutation and the simulation performance description matrix P # Perform mathematical fitting and find the value that satisfies any mutation descriptor s

[0217] All relationship matrices χ j , and determine or All mutation sites of j.

[0218] like or or This indicates that the mutation described by the mutation descriptor s at site j can improve the performance of the starting enzyme as a whole and should be combined into the combined mutation. The combination of all mutation sites j determines the solution that meets the evolutionary direction of enhancing catalytic activity; according to the solution that satisfies The combination of all mutation sites j determines the solution that meets the evolutionary direction of enhancing specificity; and / or according to satisfying The combination of all mutation sites j determines the solution that conforms to the evolutionary direction of enhancing tolerance.

[0219] Please continue to refer to Figure 1 and Figure 5 After completing the selection of the combined mutation site (step [A'1]) and obtaining a solution that conforms to at least one evolutionary direction, the directed evolution device can determine multiple candidate combined mutations according to the solution that conforms to at least one evolutionary direction provided in step [A'1], and determine at least one target mutation site according to each candidate combined mutation. Afterwards, the directed evolution device can design target primer sequences corresponding to each candidate combined mutation according to the template sequence and the at least one target mutation site (step [A'2]).

[0220] Specifically, similar to step [A2], the process of designing the target primer sequence corresponding to each candidate combination mutation (step [A'2]) can also be implemented in two stages: offline training and online prediction. In the offline training stage, the trainer can first obtain the experimental data of single-point mutation polymerase chain reaction (PCR) of multiple starting enzyme samples from the training set, wherein each experimental data contains a primer sequence and the result of the mutation experiment. Compared with the single mutation primer design, the primer design of the combination mutation does not require a reverse primer, and each PCR experiment may have a situation where the mutation contained in one, multiple or all primers is introduced into the PCR product, that is, different combination mutants are obtained.

[0221] Afterwards, the trainer can introduce mutations at three bases from the start codon to the bases i to i+2 to determine the upstream sequence $u of the forward primer after mutation of each starter enzyme sample. fw and the downstream sequence of the forward primer $d fw , and the upstream sequence $u of the forward primer after mutation of each starting enzyme sample is mutated as described in step [A2] fw and the downstream sequence of the forward primer $d fw Digital processing is performed to obtain the upstream matrix U of the forward primer after mutation of each starting enzyme sample fw and the forward primer downstream matrix D fw .

[0222] For the kth experiment, suppose it uses m k primers, then the corresponding primer upstream sequence $u fw and primer downstream sequence $d fw Can be digitally represented as a series of matrices and And form two third-order tensors respectively and

[0223] Afterwards, the number of mutations actually introduced by this experiment is recorded as w k (0≤w k ≤m k), and define the result of the kth experiment, i.e., the mutation rate r k =w k / m k , then 0≤r k ≤1, where 0 indicates that no mutation has occurred, 1 indicates that mutations in all primers have been introduced into the kth combined mutant, and values ​​between the two indicate that some mutations have been introduced into the kth combined mutant.

[0224] Thus, all experimental results r k We can construct a vector r = [r1, r2, ..., r n-1 ,r n ], all third-order tensors Can form a fourth-order tensor U (4) , all third-order tensors Can form a fourth-order tensor D (4) , the trainer can define a mathematical mutation success rate function Q that satisfies

[0225] r=Q(U (4) ,D (4) ) [Formula A'2]

[0226] And through machine learning, we train a numerical approximate solution Q of the above mutation success rate function Q # After that, the trainer can approximate the solution of Q based on the value obtained during training. # , construct a mutation success rate model of each candidate combination mutation and enzyme performance.

[0227] After completing the training process of the mutation success rate model and obtaining the numerical approximate solution Q of the mutation success rate function Q # Afterwards, the directed evolution device can first determine at least one target mutation site in the online prediction stage based on the multiple candidate combination mutations that can improve the performance of the starting enzyme as a whole as determined in step [A'1]. Afterwards, for a given target mutation, the directed evolution device can introduce mutations in three bases from the i-th to the i+2-th bases of the entire gene, starting from the start codon, based on the template sequence and at least one target mutation site i, and replace these three bases with the codons of the target amino acid residue to be mutated according to the codon preference of the corresponding target host, so as to obtain the mutated gene.

[0228] Still taking the E. coli shown in Table 3 as an example, the directed evolution device can first start from the i-th base in the 5′→3′ sequence of the mutated gene and take u bases forward. fw bases to obtain the upstream sequence of the forward primer $u fw , and start from the i+2th base in the mutated gene and take dfw bases to obtain the downstream sequence of the forward primer $d fw After that, the directed evolution device can be used to fw ,d fw These two optional lengths take values ​​such as 0, 1, 2, ..., 50, and exhaust their combinations to generate the upstream sequence $u of the forward primer fw and the downstream sequence of the forward primer $d fw Then, the directed evolution device can perform various combinations of the upstream sequence $u of the forward primer. fw and the downstream sequence of the forward primer $d fw Perform digital processing and approximate the solution Q based on the numerical value of the mutation success rate function Q obtained in advance # , determining a target primer sequence corresponding to at least one target mutation site.

[0229] Specifically, the directed evolution device can convert each set of digitally processed sequences $u fw and $d fw Substitute the numerical approximate solution Q # , to calculate the corresponding predicted experimental result r. Here, the predicted experimental result r indicates whether the corresponding combination obtains the expected mutation. Afterwards, the directed evolution device can find out which combination satisfies the second preset range (for example, 0.3 ≤ r ≤ 0.5) and u fw and fw The sequences are the smallest and are spliced ​​in the following way to determine the primer sequences that meet the requirements:

[0230] Primer sequence: 5′-$u fw + codon + $d fw -3′

[0231] Furthermore, the evolution device may repeat the above process for each candidate single point mutation in the combined mutation, so as to generate primer sequences for each single point mutation in each combined mutation.

[0232] Compared with the prior art which mainly relies on empirical formulas to design primers, the present invention obtains an approximate numerical solution Q of the mutation success rate function Q by machine learning of a large amount of training data. # , and based on this value, the approximate solution Q # AI decision-making is performed to determine the primer sequences that can generate the corresponding combination mutants. Therefore, the present invention can use the results of multiple existing theories or algorithms as the decision basis of the relationship model through a digital method, thereby efficiently, reliably and objectively constructing combination mutants, and effectively shortening the iteration cycle and economic cost of directed evolution.

[0233] Please continue to refer to Figure 1 and Figure 5 After completing the design of the combination mutation primers (step [A'2]) and obtaining the target primer sequences corresponding to each combination mutation, the directed evolution device can construct a candidate combination mutant according to the target primer sequence of the combination mutation (step [A'3]), and perform a catalytic reaction and analysis on it (step [B'1]). Those skilled in the art can complete the relevant operations of step [A'3] and step [B'1] based on the existing technology, which will not be described in detail here. Afterwards, the directed evolution device can make a positive judgment based on the result of the catalytic reaction, and preliminarily screen a plurality of positive combination mutations that can improve the performance of the starting enzyme as a whole based on the combined mutation sites of the plurality of positive combination mutants (step [B'2]). The specific process is similar to step [B2], which will not be described in detail here.

[0234] Please refer to further Figure 7 , Figure 7 A schematic diagram of a process for determining combined mutations that conform to an evolutionary direction according to some embodiments of the present invention is shown.

[0235] like Figure 7 As shown, in some embodiments of the present invention, after preliminarily screening out multiple positive combination mutations through step [B'2], the directed evolution device can also combine the result data of the catalytic reaction of each positive combination mutant to further screen solutions that meet the expected evolutionary direction (step [B'2]).

[0236] Specifically, the directed evolution device can first perform sequencing, sequence alignment, digitization and other operations on each positive combination mutation, and determine the mutation descriptors indicating the change in side chain size after mutation, the change in side chain polarity after mutation, the charge of the side chain after mutation and the rigidity of the main chain after mutation according to each mutation site j in each positive combination mutation, that is,

[0237] s j =[size j ,polarity j ,charge j ,rigidity j ]

[0238] Afterwards, the directed evolution device can sequentially combine the mutation descriptors s j , to obtain the mutation description matrix of each combined mutation, that is,

[0239] S k =[s k_1 ,s k_2 ,…s k_j ,…,s k_m-1 ,s k_m ] [Formula 20]

[0240] Afterwards, the directed evolution device can sequentially combine the mutation description matrices S k , to construct the mutation description tensor of all combined mutations, that is,

[0241] S=[S1,S2,…S k ,…,S l-1 ,S l ] [Formula 21]

[0242] In addition, if Figure 7 As shown, in the process of further screening solutions that meet the expected evolutionary direction (step [B'2]), the directed evolution device can also obtain the result data of the catalytic reaction of the starting enzyme and each positive combination mutation, and calculate the first catalytic rate constant k0, the first specificity constant E0 and the first inactivation rate constant D0 of the starting enzyme based on [Formula B2], as well as the second catalytic rate constant k of each positive combination mutation k k , the second specificity constant E k and the second inactivation rate constant D k .

[0243] Afterwards, the directed evolution device can be used to generate the first catalytic rate constant k0 and the second catalytic rate constant k k , determine the catalytic activity descriptor κ of each positive combination mutation k k ,Right now

[0244] κ k =ln(k k / k0)

[0245] In addition, the directed evolution device can also be based on the first specific constant E0 and the second specific constant E k , determine the specific descriptor ι for each positive combination mutation k k ,Right now

[0246] ι k =ln(E k / E0)

[0247] In addition, the directed evolution device can also be based on the first inactivation rate constant D0 and the second inactivation rate constant D k , determine the tolerance descriptor δ of each positive combination mutation k k ,Right now

[0248] δ k =ln(D k / D0)

[0249] Then, the directed evolution device can be used according to the above catalytic activity descriptor κ k , specific descriptor ι kand the tolerance descriptor δ k , determine the performance descriptor p of each positive combination mutation k k ,Right now

[0250] p k =[κ k ,ι k ,δ k ] [Formula 22]

[0251] Then according to the performance descriptor p of each positive combination mutation k k , construct the simulation performance description matrix P, that is

[0252] P=[p1,p2,…p i ,…,p l-1 ,p l ] [Formula 23]

[0253] Afterwards, the directed evolution device can define that there is a tensor X=[χ1,χ2,…χ j ,…,χ m-1 ,x m ],satisfy

[0254] P=S×X [Formula 24]

[0255] In the formula,

[0256]

[0257] χ1,χ2,…χ j ,…,χ m-1 ,x m is the relationship matrix of each single point mutation site j.

[0258] After that, the directed evolution device can consider any mutation descriptor s = [size, polarity, charge, rigidity], and traverse any value allowed by [Formula 13] for the mutated side chain size (size), the mutated side chain polarity (polarity), the mutated side chain charge (charge), and the mutated main chain rigidity (rigidity), respectively, to perform mathematical fitting based on the above-mentioned combined mutation description tensor S and the above-mentioned performance description matrix P, and obtain the value that satisfies any mutation descriptor s.

[0259] All relationship matrices χ j , and determine or All mutation sites of j.

[0260] like or or This indicates that the mutation described by the mutation descriptor s at site j can improve the performance of the starting enzyme as a whole and should be combined into the combined mutation. The optimal solution that meets the evolutionary direction of enhancing catalytic activity is determined by combining all mutation sites j. The optimal solution that meets the evolutionary direction of enhancing specificity is determined by combining all mutation sites j; and / or by satisfying The combination of all mutation sites j determines the optimal solution that conforms to the evolutionary direction of enhancing tolerance.

[0261] Afterwards, the directed evolution device can select the optimal solution that meets the corresponding evolution direction as the evolution enzyme for this round of directed evolution iteration based on the expected one or more evolution directions.

[0262] Please refer to Figure 8 , compared to traditional directed evolution schemes that rely entirely on experiments, and Figure 8 The traditional directed evolution scheme shown partially relies on data processing technologies such as simulation and curve fitting. The present invention can perform full-process AI decision-making based on a pre-constructed and trained relationship model M, mutation success rate model Q, interaction relationship function f and / or other mathematical fitting technologies, thereby effectively avoiding the influence of human tendencies on the directed evolution results of the enzyme, and greatly reducing the experimental work of the directed evolution process of the enzyme, so as to shorten the directed evolution cycle of the enzyme and reduce its R&D cost.

[0263] Those skilled in the art will appreciate that the above description is based on the mutation description tensor S and the simulation performance description matrix P # The scheme of performing mathematical fitting to determine multiple candidate combination mutations (step [A'1]), and then performing positive judgment and expected evolutionary direction screening based on the determined multiple candidate combination mutations (steps [A'2] to [B'2]) is only a non-limiting implementation method provided by the present invention, which is intended to clearly demonstrate the main concept of the present invention and provide a specific scheme that is easy for the public to implement, but is not used to limit the scope of protection of the present invention.

[0264] Optionally, in some other embodiments, the directed evolution device may skip step [A'1] and directly perform the steps of combined mutation primer design (step [A'2]), mutant construction (step [A'3]), catalytic reaction and analysis (step [B'1]), and data processing and analysis (step [B'2]) based on the multiple single-point mutation sites determined by step [A1], or the single-point mutation sites of the multiple positive single mutants determined by step [A2], so as to construct and screen a plurality of positive combination mutations that can improve the performance of the starting enzyme as a whole with a relatively large data processing load (still far less than the data processing load of the prior art).

[0265] For further information, please refer to Fig. 9 In some preferred embodiments, when the initial enzyme samples in the training set are sufficient, and the mathematical fitting accuracy of step [A'1] can meet the accuracy requirements of directed evolution, the directed evolution device can also preferably skip the combined mutation and combined mutation screening operations of steps [A'2] to [B'2], and directly determine the evolved enzyme for this round of directed evolution iteration based on the optimal solution that meets the evolution direction determined by mathematical fitting (step [A'1]). By omitting the process of combined mutation and combined mutation screening, the present invention can further save the time and corresponding experimental costs of this part of the experiment.

[0266] For further information, please refer to Fig.10 In some more preferred embodiments, when the initial enzyme samples in the training set are sufficient, the prediction accuracy of step [A1] and the mathematical fitting accuracy of step [A'1] can both meet the accuracy requirements of directed evolution, and the directed evolution device can preferably skip the single point mutation and single point mutation screening operations of steps [A2] to [B2], as well as the combined mutation and combined mutation screening operations of steps [A'2] to [B'2], and perform combined mutation of directed evolution based on the multiple single point mutation sites determined in step [A1], and then directly determine the evolved enzyme of this round of directed evolution iteration based on the optimal solution that meets the evolution direction determined in step [A'1]. By omitting the single point mutation and single point mutation screening, as well as the combined mutation and combined mutation screening process, the present invention can further save the time of these two parts of the experiment and the corresponding experimental costs.

[0267] In addition, if Figure 1 , Figure 5 , Fig. 9 and Fig.10As shown, after determining the evolved enzyme of this round of directed evolution iteration, the technician may also preferably conduct relevant verification experiments on the evolved enzyme to determine whether its performance meets expectations. If the performance of the evolved enzyme has met expectations, the directed evolution device may use the evolved enzyme as the final product of the directed evolution process of the enzyme and end the entire directed evolution process. Conversely, if the performance of the evolved enzyme does not meet expectations, the directed evolution device may use the evolved enzyme as the starting enzyme for the next generation of evolution to perform the next round of directed evolution iteration of the enzyme until the evolved enzyme with the expected performance is finally obtained.

[0268] Although the above methods are illustrated and described as a series of actions for simplicity of explanation, it should be understood and appreciated that these methods are not limited by the order of the actions, because according to one or more embodiments, some actions may occur in a different order and / or concurrently with other actions from those illustrated and described herein or not illustrated and described herein but understandable to those skilled in the art.

[0269] The previous description of the disclosure is provided to enable any person skilled in the art to make or use the disclosure. Various modifications to the disclosure will be apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the spirit or scope of the disclosure. Thus, the disclosure is not intended to be limited to the examples and designs described herein, but should be granted the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for directed evolution of an enzyme, characterized in that: The following steps are included: Identify multiple single-point mutation sites that improve the performance of the starting enzyme; Generating a plurality of combined mutations according to each of the single point mutation sites; According to each mutation site j in each of the combined mutations, a mutation descriptor s indicating a change in side chain size after mutation, a change in side chain polarity after mutation, a side chain charge after mutation, and a main chain rigidity after mutation is determined. j ; According to the mutation descriptor s of at least one mutation site j constituting the combined mutation k j , generate the mutation description matrix S of the combined mutation k k ; According to the mutation description matrix S of each combination mutation k k , construct the corresponding combined mutation description tensor S; Constructing a performance description matrix P according to the catalytic activity, specificity and tolerance of the starting enzyme and each of the combined mutations; as well as Mathematical fitting is performed based on the combined mutation description tensor S and the performance description matrix P to determine a solution that conforms to the evolutionary direction.

2. The directed evolution method according to claim 1, characterized in that The step of determining multiple single-point mutation sites that improve the performance of the starting enzyme comprises: The starting enzyme is simulated to determine the trajectory data B of each amino acid residue i of the starting enzyme. i and energy data ΔG i ;as well as According to the trajectory data B i and the energy data ΔG i , determine multiple single-point mutation sites that improve the performance of the starting enzyme.

3. The directed evolution method according to claim 2, characterized in that: The starting enzyme is simulated to determine the trajectory data B of each amino acid residue i of the starting enzyme. i and energy data ΔG i The steps include: The starting enzyme is subjected to a control simulation according to the starting conditions to determine the first trajectory data B of each amino acid residue i of the starting enzyme. i and the first energy data ΔG i ;as well as The starting enzyme is subjected to target simulation according to the target condition to determine the second trajectory data B of each amino acid residue i of the starting enzyme. i ' and the second energy data ΔG i '.

4. The directed evolution method according to claim 3, characterized in that: According to the trajectory data B i and the energy data ΔG i The step of determining multiple single-point mutation sites that improve the performance of the starting enzyme comprises: According to the first trajectory data B i And the second trajectory data B i ', determine the trajectory descriptor β i ; According to the first energy data ΔG i and the second energy data ΔG i ', determine the energy descriptor ε i ; According to the trajectory descriptor β i and the energy descriptor ε i , determine the structure description matrix S; and According to the structure description matrix S and the numerical approximate solution M of the mutation success rate function M obtained in advance # , determine the multiple single point mutation sites.

5. The directed evolution method according to claim 4, characterized in that: The numerical approximate solution M of the mutation success rate function M obtained in advance according to the structural description matrix S # , the step of determining the multiple single point mutation sites comprises: According to the structural description matrix S and the numerical approximate solution M # , determine the corresponding prediction performance description matrix P # , where the prediction performance description matrix P # Includes multiple prediction performance descriptors p i # , the prediction performance descriptor p i # Indicates the catalytic activity, specificity and tolerance of amino acid residue i after mutation; and According to each of the prediction performance descriptors p i # The indicated catalytic activity, specificity and tolerance were determined by determining the multiple single point mutation sites.

6. The directed evolution method according to claim 5, characterized in that According to the structural description matrix S and the numerical approximate solution M # , determine the corresponding prediction performance description matrix P # Previously, the directed evolution method also included the following steps: Obtain multiple starting enzyme samples from the training set; Performing control simulation and target simulation on each of the starting enzyme samples respectively, and determining a structure description matrix S of each of the starting enzyme samples according to the simulation results; The catalytic activity, specificity and tolerance of each amino acid residue i of each starting enzyme sample after mutation are determined according to the experimental data, so as to determine the performance descriptor p of each amino acid residue i. i , and construct a performance description matrix P for each of the starting enzyme samples; and Define the mutation success rate function P = M (S), and use machine learning to find the numerical approximate solution M of the mutation success rate function M according to the structure description matrix S and performance description matrix P of each of the starting enzyme samples. # .

7. The directed evolution method according to claim 2, characterized in that: The step of determining multiple single-point mutation sites that improve the performance of the starting enzyme comprises: Designing a single mutation target primer sequence according to at least one target mutation site; constructing a single mutant according to the target primer sequence of the single mutation; According to the constructed single mutant, a catalytic reaction is performed; Performing a positive determination based on the result of the catalytic reaction; and According to the single point mutation sites of multiple positive single mutants, multiple single point mutation sites that improve the performance of the starting enzyme are determined.

8. The directed evolution method according to claim 7, characterized in that: The step of designing a single mutation target primer sequence according to at least one target mutation site comprises: According to the template sequence and the at least one target mutation site, determine the upstream sequence $u of the forward primer after the initiator enzyme is mutated fw , forward primer downstream sequence $d fw , reverse primer upstream sequence $u rv , and the reverse primer downstream sequence $d rv ; According to multiple optional lengths, list the upstream sequence $u of the forward primer fw , the forward primer downstream sequence $d fw , the reverse primer upstream sequence $u rv And the reverse primer downstream sequence $d rv various combinations of; and For each combination of the forward primer upstream sequence $u fw , the forward primer downstream sequence $d fw , the reverse primer upstream sequence $u rv And the reverse primer downstream sequence $d rv Digital processing is performed, and the numerical approximation solution Q is obtained based on the mutation success rate function Q obtained in advance. # , determining a target primer sequence corresponding to the at least one target mutation site.

9. The directed evolution method according to claim 8, characterized in that The forward primer upstream sequence $u fw , the forward primer downstream sequence $d fw , the reverse primer upstream sequence $u rv And the reverse primer downstream sequence $d rv The forward primer upstream sequence $u of each of the combinations includes a plurality of bases j. fw , the forward primer downstream sequence $d fw , the reverse primer upstream sequence $u rv And the reverse primer downstream sequence $d rv The steps to digitize include: Vector b in digital form j =[a j ,c j ,g j ,t j ] represents base j, where a j ,c j ,g j ,t j There is only one base that takes the value of 1, and the rest take the value of 0, indicating the base type of base j.

10. The directed evolution method according to claim 9, characterized in that The numerical approximate solution Q based on the mutation success rate function Q obtained in advance # , the step of determining a target primer sequence corresponding to the at least one target mutation site comprises: The sequence $u of various combinations of the digital processing fw 、$d fw 、$u rv and $d rv , substitute the numerical value to approximate the solution Q # , to calculate the corresponding predicted experimental result r, wherein the predicted experimental result r indicates whether the combination obtains the expected mutation; and The target primer sequence is determined according to the combination in which the predicted experimental result r reaches the first preset range and the sequence length is the smallest.

11. The directed evolution method according to claim 10, characterized in that: According to the numerical approximation solution Q # Before determining the target primer sequence corresponding to the at least one target mutation site, the directed evolution method further comprises the following steps: Obtain multiple starting enzyme samples from the training set; Determine the upstream sequence $u of the forward primer after mutation of each of the starting enzyme samples fw , forward primer downstream sequence $d fw , reverse primer upstream sequence $u rv , and the reverse primer downstream sequence $d rv ; The upstream sequence $u of the forward primer after mutation of each of the starting enzyme samples fw , forward primer downstream sequence $d fw , reverse primer upstream sequence $u rv , and the reverse primer downstream sequence $d rv Digital processing is performed to obtain the forward primer upstream matrix U after each of the starting enzyme samples is mutated fw , forward primer downstream matrix D fw , reverse primer upstream matrix U rv , and the reverse primer downstream matrix D rv ; Obtaining the mutation experiment result r of each of the starting enzyme samples to form a mutation experiment result vector r, wherein the value of the mutation experiment result r is 0 or 1, indicating that the expected mutation is not obtained or the expected mutation is obtained, respectively; and Define the mutation success rate function r = Q(U rv(3) ,D rv(3) ,U fw(3) ,D fw(3) ), and adopt machine learning method, according to the forward primer upstream matrix U of each starting enzyme sample fw The third-order tensor U fw(3) , the forward primer downstream matrix D of each of the starting enzyme samples fw The third-order tensor D fw(3) , the reverse primer upstream matrix U of each of the starting enzyme samples rv The third-order tensor U rv(3) , the reverse primer downstream matrix D of each of the starting enzyme samples rv The third-order tensor D rv(3) , and the mutation experiment result vector r, find the numerical approximate solution Q of the mutation success rate function Q # .

12. The directed evolution method according to claim 7, characterized in that: The step of making a positive determination according to the result of the catalytic reaction comprises: Obtaining result data of the catalytic reaction of the single mutant j, wherein the result data includes conversion rate data x, and evolution data y indicating the catalytic activity, specificity and / or tolerance of the mutant j; and According to the conversion rate data x, the evolution data y, and the numerical approximate solution f of the interaction relationship function f obtained in advance # , perform a positive determination of the single mutant j.

13. The directed evolution method according to claim 12, characterized in that: The numerical approximate solution f based on the conversion rate data x, the evolution data y, and the pre-calculated interaction relationship function f # The steps of performing a positive determination of the single mutant j include: If y j >f # (x j ) or x j >(f # ) -1 (y), then the single mutant j is determined to be a positive single mutant; and If y j ≤f # (x j ) and x j ≤(f # ) -1 (y), then the single mutant j is determined to be a negative single mutant.

14. The directed evolution method according to claim 1, wherein: The step of constructing a performance description matrix P according to the catalytic activity, specificity and tolerance of the starting enzyme and each of the combined mutations comprises: Performing control simulation and target simulation on the starting enzyme and each of the combined mutations k respectively to determine the catalytic activity, specificity and tolerance of each of the combined mutations k; Determine the simulation performance descriptor of each of the combined mutations k according to the catalytic activity, the specificity and the tolerance as well as According to the simulation performance descriptor of each combination mutation k Construct simulation performance description matrix P # .

15. The directed evolution method according to claim 14, characterized in that: The step of performing control simulation and target simulation on the starting enzyme and each of the combined mutations k to determine the catalytic activity, specificity and tolerance of each of the combined mutations k comprises: The control simulation and the target simulation are performed on the starting enzyme and each of the combination mutations k, respectively, to determine the third energy data ΔG0 and the fourth energy data ΔG′0 of the starting enzyme, and the fifth energy data ΔG k and the sixth energy data ΔG′ k ; According to the fourth energy data ΔG′0 and the sixth energy data ΔG′ k , determine the catalytic activity descriptor of the combined mutation k According to the third energy data ΔG0, the fourth energy data ΔG′0, the fifth energy data ΔG k and the sixth energy data ΔG′ k , determine the specific descriptor of the combined mutation k as well as The first average RMSD value of the starting enzyme given by the target simulation And the first average RMSD value of each combination mutation k Determine the tolerance descriptor of the combined mutation k 16. The directed evolution method according to claim 14, characterized in that: The step of performing mathematical fitting according to the combined mutation description tensor S and the performance description matrix P to determine a solution that conforms to the evolutionary direction comprises: Define a tensor X that satisfies P # =S×X, wherein the tensor X includes the relationship matrix X corresponding to each mutation site j j ; According to the combined mutation description tensor S and the simulation performance description matrix P # Perform mathematical fitting to obtain the relationship matrix χ j , where the relationship matrix χ j For any mutation descriptor s that satisfies Traverse the range of values ​​of the side chain size after mutation, the side chain polarity after mutation, the side chain charge after mutation, and the main chain rigidity after mutation to determine or The mutation site j; and According to satisfaction The combination of all mutation sites j that satisfy The combination of all mutation sites j, and / or satisfy The combination of all mutation sites j determines the solution that conforms to the evolutionary direction.

17. The directed evolution method according to claim 1 or 16, characterized in that: The step of generating a plurality of combined mutations according to each of the single point mutation sites comprises: Generating a plurality of candidate combination mutations according to one or more of the single-point mutation sites, or determining a plurality of candidate combination mutations according to a solution that conforms to the evolutionary direction; Determine at least one target mutation site according to each of the candidate combination mutations; According to the template sequence and the at least one target mutation site, respectively designing target primer sequences for each of the candidate combination mutations; According to the target primer sequences of each candidate combination mutation, respectively construct corresponding candidate combination mutants; According to each of the candidate combination mutants, respectively, a catalytic reaction is performed; and A positive determination is performed based on the results of each of the catalytic reactions to determine a plurality of positive combination mutations.

18. The directed evolution method according to claim 17, characterized in that: The step of designing target primer sequences for each candidate combination mutation according to the template sequence and the at least one target mutation site comprises: According to the template sequence and the at least one target mutation site, determine the upstream sequence $u of the forward primer after the initiator enzyme is mutated fw , and the downstream sequence of the forward primer $d fw ; According to multiple optional lengths, list the upstream sequence $u of the forward primer fw And the forward primer downstream sequence $d fw Various combinations of For each combination of the forward primer upstream sequence $u fw And the forward primer downstream sequence $d fw Digital processing is performed, and the numerical approximation solution Q is obtained based on the mutation success rate function Q obtained in advance. # , determining a target primer sequence corresponding to the at least one target mutation site; and According to each of the single point mutation sites in each of the candidate combination mutations, the above operations are repeatedly performed to generate target primer sequences for each of the candidate combination mutations.

19. The directed evolution method according to claim 18, characterized in that: The numerical approximate solution Q based on the mutation success rate function Q obtained in advance # , the step of determining a target primer sequence corresponding to the at least one target mutation site comprises: The sequence $u of various combinations of the digital processing fw and $d fw , substitute the numerical value to approximate the solution Q # , to calculate the corresponding predicted experimental result r, wherein the predicted experimental result r indicates whether the combination obtains the expected mutation; and The target primer sequence is determined according to the combination in which the predicted experimental result r reaches the second preset range and the sequence length is the smallest.

20. The directed evolution method according to claim 19, characterized in that: According to the numerical approximation solution Q # Before determining the target primer sequence corresponding to the at least one target mutation site, the directed evolution method further comprises the following steps: Obtain multiple starting enzyme samples from the training set; Determine the upstream sequence $u of the forward primer after mutation of each of the starting enzyme samples fw , and the downstream sequence of the forward primer $d fw ; The upstream sequence $u of the forward primer after mutation of each of the starting enzyme samples fw and the downstream sequence of the forward primer $d fw Digital processing is performed to obtain the forward primer upstream matrix U after each of the starting enzyme samples is mutated fw and the forward primer downstream matrix D fw ; Obtaining the mutation experiment result r of each of the starting enzyme samples to form a mutation experiment result vector r, wherein the mutation experiment result r=w / m, 0≤r≤1, indicating that w mutations in each of the starting enzyme samples are introduced into the candidate combination mutant, and m is the total number of primers; and Define the mutation success rate function r = Q(U (4) ,D (4) ), and adopt machine learning method, according to the forward primer upstream matrix U of each starting enzyme sample fw The fourth-order tensor U (4) , the forward primer downstream matrix D of each of the starting enzyme samples fw The fourth-order tensor D (4) , and the mutation experiment result vector r, find the numerical approximate solution Q of the mutation success rate function Q # .

21. The directed evolution method according to claim 17, wherein: The step of constructing a corresponding combined mutation description tensor S according to the multiple combined mutations includes: According to each mutation site j in each of the positive combined mutations, a mutation descriptor s indicating a change in side chain size after mutation, a change in side chain polarity after mutation, a side chain charge after mutation, and a main chain rigidity after mutation is determined. j ; According to the mutation descriptor s of at least one mutation site j constituting the combined mutation k j , generate the mutation description matrix S of the combined mutation k k ;as well as According to the mutation description matrix S of each combination mutation k k , construct the combined mutation description tensor S.

22. The directed evolution method according to claim 21, characterized in that: The step of constructing a performance description matrix P according to the catalytic activity, specificity and tolerance of the starting enzyme and each of the combined mutations comprises: Obtain the result data of the catalytic reaction of the starting enzyme and each of the positive combination mutations to calculate the first catalytic rate constant k0, the first specificity constant E0 and the first inactivation rate constant D0 of the starting enzyme, and the second catalytic rate constant k of each of the positive combination mutations k k , the second specificity constant E k and the second inactivation rate constant D k ; According to the first catalytic rate constant k0 and the second catalytic rate constant k k , determine the catalytic activity descriptor κ of the positive combination mutation k k ; According to the first specific constant E0 and the second specific constant E k , determine the specific descriptor ι of the positive combination mutation k k ; According to the first deactivation rate constant D0 and the second deactivation rate constant D k , determine the tolerance descriptor δ of the positive combination mutation k k ; According to the catalytic activity descriptor κ k , the specific descriptor ι k and the tolerance descriptor δ k , determine the performance descriptor p of each positive combination mutation k k ;as well as According to the performance descriptor p of each positive combination mutation k k , construct the simulation performance description matrix P.

23. The directed evolution method according to claim 22, wherein: The step of performing mathematical fitting according to the combined mutation description tensor S and the performance description matrix P to determine a solution that conforms to the evolutionary direction comprises: Define a tensor X that satisfies P=S×X, wherein the tensor X includes a relationship matrix χ corresponding to each mutation site j j ; Mathematical fitting is performed based on the combined mutation description tensor S and the performance description matrix P to obtain each of the relationship matrices x j , where the relationship matrix χ j For any mutation descriptor s, s×χ j =[κ j ,ι j ,δ j ]; Traverse the range of values ​​of the side chain size after mutation, the side chain polarity after mutation, the side chain charge after mutation, and the main chain rigidity after mutation to determine κ j >0,ι j >0 or δ j >0 mutation site j; and According to satisfaction The combination of all mutation sites j that satisfy The combination of all mutation sites j, or satisfy The combination of all mutation sites j determines the solution that conforms to the evolutionary direction.

24. The directed evolution method according to claim 1, wherein: After determining a solution that conforms to the evolutionary direction, the directed evolution method further comprises the following steps: Determine the evolution enzyme for this round of directed evolution according to the solution that conforms to the evolution direction; Determining whether the evolved enzyme meets expectations; and In response to the judgment that the evolved enzyme does not achieve the expected result, the evolved enzyme is used as the starting enzyme for the next round of directed evolution.

25. An enzyme directed evolution device, characterized in that: include: Memory; as well as A processor connected to the memory and configured to implement the directed evolution method of the enzyme according to any one of claims 1 to 24.

26. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by a processor, the directed evolution method of an enzyme according to any one of claims 1 to 24 is implemented.

Images