Design method and device of protein enzyme, electronic equipment and storage medium
By constructing protein and small molecule complexes and using protein sequence design models to modify enzyme sites, the problem of poor activity and stability in enzyme modification design is solved, and the effect of rapid and efficient generation of high-efficiency enzymes is achieved.
Patent Information
- Application Number
- CN202411585032.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-07
- Publication Date
- 2025-02-18
AI Technical Summary
In the prior art, the activity and stability of the enzyme transformation design are poor, resulting in high cost and low efficiency of the enzyme transformation process.
By determining the known substrate to be catalyzed, constructing a constructed complex composed of proteins and small molecules, and using the trained protein sequence design model to modify the protein pocket and some surrounding residues in the constructed complex, we obtain Target protein enzyme sequence.
It realizes rapid and efficient generation of enzymes with improved activity and stability, reduces the workload of directed evolution and other projects, and improves the efficiency of enzyme transformation design.
Smart Images
Figure CN119479800A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of biology and computer technology, and in particular to a design method, device, electronic equipment and storage medium for a protein enzyme. Background Art
[0002] Proteases are important in the fields of medicine, chemical manufacturing, biotechnology, energy, agriculture, consumer products, etc. For example, antibodies can be designed to enhance their binding and specificity as therapeutic drugs, while the stability and activity of enzymes can be improved under process conditions to obtain more environmentally friendly and efficient chemical synthesis. Protein engineering or enzyme engineering is a technical field in which the genes of proteins are purposefully designed and modified to meet specific needs.
[0003] Traditional enzyme modification technology has the disadvantages of huge mutation sampling space, high experimental costs and reliance on high-throughput technology, which greatly limits the process of enzyme modification design. Compared with the single-pathway enzyme activity improvement model of traditional directed evolution, machine learning can combine existing sequence-function data to comprehensively calculate the global activity optimization path, propose the optimal sequence activity model, and guide the further activity modification of the enzyme.
[0004] Therefore, how to use machine learning to construct a suitable protein-small molecule complex system based on the provided substrate is an urgent problem to be solved. Summary of the invention
[0005] The present invention provides a protein enzyme design method, device, electronic device and storage medium, which are used to solve the defects of poor activity and stability of enzyme modification design in the prior art.
[0006] The present invention provides a method for designing a protein enzyme, comprising: Identify known substrates to be catalyzed; Based on the known substrate or the related enzyme catalytic reaction mechanism, a reference complex is retrieved from a pre-constructed protein enzyme database, and the known substrate is docked with the protein enzyme in the reference complex to obtain a constructed complex; Based on the trained protein sequence design model, the protein pocket in the structural complex and some of the residues around it are modified to obtain the target protein enzyme sequence.
[0007] According to the method for designing a protein enzyme provided by the present invention, the protein pocket in the structural complex and some of the residues around it are modified based on the trained protein sequence design model to obtain the target protein enzyme sequence, including: Masking the protein pocket in the structural complex and some of the residues around it to obtain a masked protein sequence; The construction complex and the mask protein sequence are input into the trained protein sequence design model to obtain a plurality of protein enzyme sequences output by the protein sequence design model, and the target protein enzyme sequence is determined from the plurality of protein enzyme sequences.
[0008] According to the protein enzyme design method provided by the present invention, the construction complex and the mask protein sequence are input into the trained protein sequence design model to obtain a plurality of protein enzyme sequences output by the protein sequence design model, comprising: Inputting the structural complex into the encoder in the protein sequence design model to encode the structural information of the structural complex to obtain structural encoding information; The structural encoding information and the masked protein sequence are input into a decoder in the protein sequence design model to perform protein enzyme sequence prediction to obtain the multiple protein enzyme sequences.
[0009] According to the method for designing a protein enzyme provided by the present invention, the method further comprises: Based on the protein sequence design model, other residues in the target protein enzyme sequence are designed to obtain an optimized protein enzyme sequence, wherein the other residues are residues in the target protein enzyme sequence except the pocket and some residues around it.
[0010] According to the method for designing a protein enzyme provided by the present invention, the method designs other residues in the target protein enzyme sequence based on the protein sequence design model to obtain an optimized protein enzyme sequence, including: Determining multiple sequence alignment information of the target protein enzyme sequence; Based on the protein sequence design model and the multiple sequence alignment information, other residues in the target protein enzyme sequence are designed to obtain the optimized protein enzyme sequence.
[0011] According to the protein enzyme design method provided by the present invention, the other residues in the target protein enzyme sequence are designed based on the protein sequence design model and the multiple sequence alignment information to obtain the optimized protein enzyme sequence, including: Based on the protein sequence design model and the multiple sequence alignment information, other residues in the target protein enzyme sequence are designed to obtain multiple candidate protein enzyme sequences; Based on at least one of the similarity between the candidate protein enzyme sequence and the original sequence, the similarity between the candidate protein enzyme sequences, and the score of the candidate protein enzyme sequence, the candidate protein enzyme sequence is screened to obtain the optimized protein enzyme sequence.
[0012] According to the method for designing a protein enzyme provided by the present invention, docking the known substrate with the protein enzyme in the reference complex to obtain a constructed complex comprises: docking the known substrate with the protein enzyme in the reference complex to obtain configurations of multiple known substrates; The structured complex is determined based on the energies of the complexes respectively corresponding to the configurations of the plurality of known substrates.
[0013] The present invention also provides a design device for protein enzyme, comprising: a substrate determination unit, for determining a known substrate to be catalyzed; A complex determination unit, for retrieving a reference complex from a pre-constructed protein enzyme database based on the known substrate or a related enzyme catalytic reaction mechanism, and docking the known substrate with the protein enzyme in the reference complex to obtain a constructed complex; The transformation unit is used to design a model based on the trained protein sequence, transform the protein pocket in the structural complex and some residues around it to obtain the target protein enzyme sequence.
[0014] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for designing a protein enzyme as described above is implemented.
[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for designing a protein enzyme as described in any one of the above is implemented.
[0016] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method for designing a protein enzyme as described above is implemented.
[0017] The protein enzyme design method, device, electronic device and storage medium provided by the present invention use the known substrate to be catalyzed to construct a new structural complex composed of protein and small molecules, and then based on the trained protein sequence design model, the protein pocket in the structural complex and some residues around it are modified to obtain the target protein enzyme sequence. The deep learning model is used to realize the mutation of the enzyme site, and the rapid and high-consumption enzyme generation is realized, which greatly reduces the workload of engineering such as directed evolution. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0019] Figure 1 This is one of the schematic diagrams of the process for designing the protein enzyme provided by the present invention.
[0020] Figure 2 It is a structural schematic diagram of the protein sequence design model provided by the present invention.
[0021] Figure 3 This is the second schematic diagram of the process of designing the protein enzyme provided by the present invention.
[0022] Figure 4 This is the third schematic diagram of the process of designing the protein enzyme provided by the present invention.
[0023] Figure 5 It is a schematic diagram of the structure of the design device of the protein enzyme provided by the present invention.
[0024] Figure 6 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0025] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0026] At its core, protein engineering is a design problem: the goal is to generate or alter the amino acid sequence of a protein to encode a desired function.
[0027] Protein engineering is challenging because accurate biophysical prediction methods for determining protein fitness are rare or non-existent, and the search space of possible proteins is astronomically large. Worse, functional proteins are scarce in the space of all protein sequences, and searching for the optimal sequence among this repertoire is NP-hard, as there is no known polynomial-time solution. Much attention is currently being paid to engineering enzymes, with applications ranging from chemical synthesis and plastic degradation to diagnostic protein therapeutics and gene editing. The design and optimization of enzymes presents some unique challenges: catalysis is more complex than binding and may involve multiple substrates, cofactors, and elementary steps.
[0028] Enzyme engineering often uses directed evolution (DE) to optimize these characteristics. At a high level, enzyme engineering involves finding an enzyme with a certain initial activity level that satisfies some but not all desired properties, and then using DE to improve fitness. Thus, the first step in the enzyme engineering workflow involves identifying an enzyme with some measurable fitness. New enzymes that do not exist in nature are synthesized through in vitro evolution techniques to suit specific industrial applications.
[0029] Traditional enzyme modification technology has the disadvantages of huge mutation sampling space, high experimental costs, and reliance on high-throughput technology, which greatly limits the process of enzyme modification design. The directed evolution technology of enzymes imitates the process of gene mutation and natural selection in the natural evolution process. By introducing mutations into genes and directly or indirectly screening mutants, after multiple rounds of iterations, mutants that meet expectations such as improved enzyme activity, improved substrate specificity, and improved enzyme thermal stability are finally obtained. Compared with natural evolution, directed evolution technology can perform multiple rounds of mutations and screening on enzymes in a short period of time, quickly search the sequence mutation space, and is not limited by whether the protein structure or mechanism is clear. It is an effective and efficient means of modifying proteins.
[0030] Although directed evolution has been widely used in enzyme modification and has achieved great success, the defects of this method are also obvious. For a given protein, the single-point saturated mutation space is 20^N (N is the length of the protein sequence), and the mutation space of multi-point combination mutation is larger. The mutation library used in directed evolution only occupies a small part of all possible sequences. Using a larger library and more screening can certainly solve this problem, but the workload that comes with it will be a huge obstacle.
[0031] Compared with directed evolution, rational and semi-rational enzyme modification methods use sequence, structure, function and other information as prior knowledge, which greatly narrows the range of amino acids to be considered, reduces experimental workload, and increases the probability of beneficial mutations. The selection of enzyme modification regions is the first step to be faced in the enzyme modification design process. Correct selection of modification sites can greatly improve the efficiency of enzyme modification. Generally, the selection methods of modification regions can be divided into structure-based and multiple sequence alignment MSA (multi sequences alignment)-based methods. The structure-based selection method is to calculate the distance to the substrate and then divide the protein residues into different regions. The MSA-based method analyzes, locates and identifies functional regions in protein sequences. Exploring the conservation and ancestral evolutionary relationship of protein amino acids is the main aspect of the application of sequence evolution information. In short, using these structural or MSA prior knowledge and adopting rational modification methods can reduce the difficulty of enzyme modification.
[0032] With the development of next-generation sequencing technology, high-throughput screening methods, protein modification databases and artificial intelligence, enzyme modification design driven by statistical data is becoming a potential and effective solution to these challenges. In recent years, machine learning-assisted enzyme modification methods have been initially proposed and have achieved some considerable achievements in enzyme activity modification, stereoselectivity modification and thermal stability modification. This method of learning from existing enzyme modification data and supplemented by experimental verification and mathematical model verification has greatly improved the efficiency of enzyme modification. Recently, deep learning-based sequence design models such as LigandMPNN use complex structures to redesign protein sequences, which can still improve protein activity and stability in the presence of more mutation sites.
[0033] Traditional enzyme modification technology has the disadvantages of huge mutation sampling space, high experimental costs and reliance on high-throughput technology, which greatly limits the process of enzyme modification design. Compared with the single-pathway enzyme activity improvement model of traditional directed evolution, machine learning can combine existing sequence-function data to comprehensively calculate the global activity optimization path, propose the optimal sequence activity model, and guide the further activity modification of the enzyme.
[0034] Recent deep learning-based technologies can quickly design sequences for known complex structures, and have a high success rate in the modification of proteases. Therefore, how to construct a suitable protein-small molecule complex system based on the provided substrate and successfully apply the deep learning model to the construction of the complex system is an urgent problem to be solved.
[0035] Based on this, the embodiment of the present invention provides a method for designing a protein enzyme, which uses a known substrate to be catalyzed to construct a new structural complex composed of a protein and a small molecule, and then based on the trained protein sequence design model, the protein pocket in the structural complex and some of the residues around it are modified to obtain the target protein enzyme sequence. The deep learning model is used to realize the mutation of the enzyme site, which realizes the rapid and high-consumption enzyme generation, greatly reducing the workload of engineering such as directed evolution.
[0036] The embodiments of the present invention can be applied to the design scenario of a new protein enzyme that catalyzes a given substrate. The execution subject of the method can be an electronic device such as a terminal device, a computer, a server, a server cluster, or a specially designed protein enzyme design device, or it can be a protein enzyme design device set in the electronic device. The device can be implemented by software, hardware, or a combination of both.
[0037] Figure 1 FIG. 1 is one of the flow charts of the method for designing protein enzymes provided by the present invention, such as Figure 1 As shown, the method comprises the following steps: Step 110, determining a known substrate to be catalyzed; Step 120, based on known substrates or related enzyme catalytic reaction mechanisms, a reference complex is retrieved from a pre-constructed protein enzyme database, and the known substrate is docked with the protein enzyme in the reference complex to obtain a constructed complex; Step 130, based on the trained protein sequence design model, the protein pocket in the structure complex and some of the residues around it are modified to obtain the target protein enzyme sequence.
[0038] Specifically, the known substrate to be catalyzed is the substrate that is combined with the protein enzyme to be designed to undergo a catalytic reaction. After the known substrate is determined, a new complex system consisting of proteins and small molecules is constructed according to the known substrate to be combined, that is, a complex is constructed.
[0039] In order to generate the structural complex, we first build a database of protein enzymes. To build this database, we need to search for published enzyme structures or sequence data in relevant literature and classify them according to the reaction mechanism catalyzed by the enzyme. To this end, we first use tools to classify the PDB (protein data bank) database according to the substrates catalyzed or bound. At the same time, we further classify the data for each structural data using homologous sequence comparison data to ensure that proteins in the same class have similar functions.
[0040] On this basis, based on known substrates or related enzyme catalytic reaction mechanisms, reference complexes are retrieved from the pre-built protein enzyme database. That is, the substrate can be searched or searched according to the related reaction mechanism to ensure that the substrate in the reference complex is similar to the known substrate. Try to find a complete protein and small molecule complex data with structural data in the protein enzyme database.
[0041] Next, the known substrate can be put into the complex with reference to the reference substrate to obtain a new complex data, that is, the known substrate is docked with the protein enzyme in the reference complex to obtain the constructed complex.
[0042] In some embodiments, step 120 docks the known substrate with the protein enzyme in the reference complex to obtain the constructed complex, which specifically includes: Step 121, docking the known substrate with the protein enzyme in the reference complex to obtain configurations of multiple known substrates; Step 122, determining the structure of the complex based on the energies of the complexes corresponding to the configurations of the plurality of known substrates.
[0043] Since the reference substrate and the known substrate cannot be exactly the same, putting the known substrate in may cause new problems, such as the new substrate may clash with the protein. To solve this problem, the embodiment of the present invention uses a docking tool to regenerate the conformation of the small molecule.
[0044] The reference substrate and the known substrate have similar functional groups. During the docking process, it is necessary to ensure that the structure of the overlapping functional group part remains unchanged and optimize the placement of the known substrate. In this embodiment, R-dock is first used to perform docking of proteins and small molecules to obtain the configurations of multiple known substrates and design a variety of small molecule configurations. Then, AutoDock-Vina is used to calculate the energy of the complex. Based on the energy of the complex corresponding to the configurations of multiple known substrates, a relatively reasonable structure is screened out from the configurations of multiple known substrates to obtain a structural complex.
[0045] After obtaining the structural complex, that is, constructing a complex system of small molecule substrates, coenzyme factors and proteins, it is necessary to further transform the structural complex, mainly the protein.
[0046] Here, the embodiment of the present invention uses a protein sequence design model PTSeqD model combined with small molecules, that is, a protein sequence design model, to transform the protein pocket and some of the residues around it in the constructed complex to obtain the target protein enzyme sequence. In the existing complex system, if the residues in the pocket part are not considered, the obtained complex and protein have a good fit. However, due to the change of the functional group in the pocket part, the residues around the protein pocket need to be changed to adapt to the changes in this part to achieve efficient catalysis of the enzyme. Therefore, it is necessary to fix most of the residues and only design the residues around the pocket to ensure the rationality of the interaction between the protein and the small molecule.
[0047] In some embodiments, step 130 specifically includes: Step 131, masking the protein pocket in the complex and some of the residues around it to obtain a masked protein sequence; Step 132, input the constructed complex and the mask protein sequence into the trained protein sequence design model, obtain multiple protein enzyme sequences output by the protein sequence design model, and determine the target protein enzyme sequence from the multiple protein enzyme sequences.
[0048] Specifically, for the design of the target protein enzyme sequence, the trained protein sequence design model uses a graph neural network to encode the structural information of proteins and small molecules, and then predicts the sequence in the mask-predict task during decoding. Therefore, the protein pocket in the complex and some of the residues around it are first masked to obtain a masked protein sequence.
[0049] The input of the protein sequence design model is the structural information of the constructed complex and the mask protein sequence, wherein the structural information of the constructed complex serves as the generation condition of the target protein enzyme sequence.
[0050] The output of the protein sequence design model is multiple protein enzyme sequences, which can be screened using chemical prior knowledge to ensure that the mutation of the residues is reasonable. Finally, it is verified whether the obtained sequence can complete the catalysis of the known substrate and the target protein enzyme sequence is determined from the multiple protein enzyme sequences generated.
[0051] In some embodiments, in step 132, the construction complex and the mask protein sequence are input into a trained protein sequence design model to obtain a plurality of protein enzyme sequences output by the protein sequence design model, specifically including: Step 132-1, inputting the structured complex into an encoder in the protein sequence design model, encoding the structural information of the structured complex, and obtaining structural encoding information; Step 132-2, input the structure encoding information and the masked protein sequence into the decoder in the protein sequence design model to perform protein enzyme sequence prediction to obtain multiple protein enzyme sequences.
[0052] Specifically, Figure 2 It is a schematic diagram of the structure of the protein sequence design model provided by the present invention, such as Figure 2 As shown in FIG. 1 , the protein sequence design model includes an encoder and a decoder. The encoder input is the structural complex. The encoder encodes the structural information of the complex to obtain the structural encoding information. The structural encoding information is input into the decoder as a condition. The decoder input is the structural encoding information and the masked protein sequence, and the output is multiple protein enzyme sequences.
[0053] During the training process of the protein sequence design model, the proportion of the mask is selected with random probability each time, and then the sequence is masked, and the model learns the types of residues that are masked. In the sampling stage, the model needs multiple iterations to generate a stable protein sequence. In actual use, you can select the sequence that needs to be designed or predicted, mask it, and then predict its sequence representation. If the sequence is designed from scratch, all sequences are masked at the beginning and then iterated for several rounds to obtain the predicted sequence.
[0054] The method provided in the embodiment of the present invention uses the known substrate to be catalyzed to construct a new structural complex composed of a protein and a small molecule, and then based on the trained protein sequence design model, the protein pocket in the structural complex and some of the residues around it are modified to obtain the target protein enzyme sequence. The deep learning model is used to realize the mutation of the enzyme site, which realizes the rapid and high-consumption enzyme generation, greatly reducing the workload of engineering such as directed evolution.
[0055] Based on any of the above embodiments, based on the trained protein sequence design model, the protein pocket in the structure complex and some of the residues around it are modified to obtain the target protein enzyme sequence, that is, after step 130, the following steps are included: Step 140, based on the protein sequence design model, design other residues in the target protein enzyme sequence to obtain an optimized protein enzyme sequence, wherein the other residues are residues in the target protein enzyme sequence except for the pocket and some residues around it.
[0056] Specifically, the above embodiment introduces the use of protein sequence design models to modify the pockets of the obtained protein small molecule complex. Generally, the protein obtained by modifying the pocket may be able to catalyze the corresponding substrate, but it may not meet the needs of actual engineering. Therefore, the protein sequence can be further optimized to improve the activity or stability of the protease to ensure that the designed enzyme can be used.
[0057] Generally, the enzyme can be subjected to point mutation and then screened to transform the enzyme to achieve the purpose of improving its stability or activity. In this embodiment, after obtaining the target protein enzyme sequence, other residues in the target protein enzyme sequence are designed to obtain an optimized protein enzyme sequence, and the protein sequence design model PTSeqD model is still used for sequence design instead of point mutation. Related experiments have shown that the reasonable use of the PTSeqD model to design the sequence of protein structures bound to small molecules can improve its stability. For the design of the target protein enzyme sequence, most of the residues are fixed to design the residues around the pocket. For the design of the optimized protein enzyme sequence, it is necessary to fix the pocket after the mutation in the previous step to design other residues. Here, other residues are residues in the target protein enzyme sequence except for the pocket and some residues around it. In this way, the optimized protein enzyme sequence after optimizing the target protein enzyme sequence can be obtained.
[0058] Based on any of the above embodiments, other residues in the target protein enzyme sequence are designed based on the protein sequence design model to obtain an optimized protein enzyme sequence, that is, step 140 specifically includes: Step 141, determining multiple sequence alignment information of the target protein enzyme sequence; Step 142, based on the protein sequence design model and the multiple sequence alignment information, other residues in the target protein enzyme sequence are designed to obtain an optimized protein enzyme sequence.
[0059] Specifically, considering the long-term directed evolution of proteins, the rational use of the multiple sequence alignment (MSA) information of the sequence can help design the stability of the sequence. Therefore, in the embodiment of the present invention, the multiple sequence alignment information of the target protein enzyme sequence is first determined. When designing other residues in the target protein enzyme sequence, the input of the protein sequence design model also includes multiple sequence comparison information. The model can introduce MSA prior knowledge to improve stability. Perform an MSA search on the sequence to calculate the output probability of each site. When the PTSeqD model is iteratively designed, the probability of each residue to be designed is calculated by superimposing the probability of the model itself and the probability obtained by MSA prior knowledge, and then sampling is performed to obtain the type of residue. The introduction of MSA information can well ensure the rationality of the designed sequence.
[0060] Based on any of the above embodiments, other residues in the target protein enzyme sequence are designed based on the protein sequence design model and multiple sequence alignment information to obtain an optimized protein enzyme sequence, that is, step 142 specifically includes: Step 142-1, designing other residues in the target protein enzyme sequence based on the protein sequence design model and multiple sequence alignment information to obtain multiple candidate protein enzyme sequences; Step 142-2, based on at least one of the similarity between the candidate protease sequence and the original sequence, the similarity between the candidate protease sequences, and the score of the candidate protease sequence, the candidate protease sequence is screened to obtain an optimized protease sequence.
[0061] Specifically, a large number of protein sequences, i.e., multiple candidate protein sequences, can be generated by using a protein sequence design model and multiple sequence alignment information. During the generation process, the proportion of the mask can be controlled each time the sample is sampled to achieve the goal of controlling the number of mutation sites of the protein.
[0062] In order to further obtain optimized protein sequences from candidate protein enzyme sequences, screening can be performed from three aspects: the similarity between the candidate protein enzyme sequences and the original sequences, the similarity between the candidate protein enzyme sequences, and the scores of the candidate protein enzyme sequences.
[0063] For example, the first screening can be performed by combining the similarity between the candidate protein enzyme sequence and the original sequence, as well as the similarity between the candidate protein enzyme sequences, and then the protein enzyme sequences obtained after the first screening are scored for further screening. Here, a model that scores protein structure and generated sequence, such as the proteinMPNN model, can be used to evaluate the generated sequence. Then, further screening is performed based on the score, and finally a suitable number of protein sequences are selected for testing, and the protein sequences with better test results are used as the final optimized protein enzyme sequences.
[0064] Based on any of the above embodiments, Figure 3 FIG. 2 is a flow chart of the method for designing protein enzymes provided by the present invention. Figure 3 As shown, a method for designing a protein enzyme is provided, comprising: S1, Construction of protein enzyme database.
[0065] S2, constructing the complex structure of known substrates and proteins. Specifically, it includes: based on the known substrates or related enzyme catalytic reaction mechanisms, retrieving reference complexes from a pre-constructed protein enzyme database, docking the known substrates with the protein enzymes in the reference complexes, and obtaining the configurations of multiple known substrates; based on the energies of the complexes corresponding to the configurations of multiple known substrates, determining the constructed complexes.
[0066] S3, design protein pocket. Based on the trained protein sequence design model, the protein pocket and some of the residues around it in the complex are modified to obtain the target protein enzyme sequence. Specifically include: The protein pocket and some of the residues around it in the structural complex are masked to obtain a masked protein sequence; the structural complex is input into an encoder in a protein sequence design model to encode the structural information of the structural complex to obtain structural encoding information; the structural encoding information and the masked protein sequence are input into a decoder in the protein sequence design model to predict the protein enzyme sequence to obtain multiple protein enzyme sequences; and a target protein enzyme sequence is determined from the multiple protein enzyme sequences.
[0067] S4, De novo protein sequence design. Figure 4 FIG. 3 is a schematic diagram of the process for designing a protein enzyme provided by the present invention. Figure 4 As shown, redesigning the protein sequence specifically includes: Determine the multiple sequence alignment information of the target protein enzyme sequence; design other residues in the target protein enzyme sequence based on the protein sequence design model and the multiple sequence alignment information to obtain multiple candidate protein enzyme sequences; screen the candidate protein enzyme sequences based on at least one of the similarity between the candidate protein enzyme sequence and the original sequence, the similarity between the candidate protein enzyme sequences, and the score of the candidate protein enzyme sequence to obtain an optimized protein enzyme sequence.
[0068] The embodiment of the present invention proposes a method for designing a new protease to efficiently catalyze a given substrate, obtains a new protein and a given substrate system from a reference substrate and protein complex system, and uses a deep learning model to modify the pocket region to achieve catalysis of the given substrate. Rapid and high-consumption enzyme generation is achieved, greatly reducing the workload of engineering such as directed evolution. In addition, the complex structure of proteins and small molecules and the prior knowledge of MSA are used to improve the activity or stability of the enzyme, ensuring the rationality of the final generation result and the activity of the enzyme.
[0069] The following is a description of the protein enzyme design device provided by the present invention. The protein enzyme design device described below and the protein enzyme design method described above can be referenced to each other.
[0070] Based on any of the above embodiments, Figure 5 Schematic diagram of the structure of the design device of the protein enzyme provided by the present invention, such as Figure 5 As shown, the device comprises: A substrate determination unit 510, for determining a known substrate to be catalyzed; The complex determination unit 520 is used to retrieve a reference complex from a pre-constructed protein enzyme database based on the known substrate or the related enzyme catalytic reaction mechanism, and dock the known substrate with the protein enzyme in the reference complex to obtain a constructed complex; The transformation unit 530 is used to design a model based on the trained protein sequence, and to transform the protein pocket in the structural complex and some of the residues around it to obtain the target protein enzyme sequence.
[0071] The device provided by the embodiment of the present invention uses the known substrate to be catalyzed to construct a new structural complex composed of proteins and small molecules, and then based on the trained protein sequence design model, the protein pocket in the structural complex and some residues around it are modified to obtain the target protein enzyme sequence. The deep learning model is used to realize the mutation of the enzyme site, which realizes the rapid and high-consumption enzyme generation, greatly reducing the workload of engineering such as directed evolution.
[0072] Based on any of the above embodiments, the transformation unit is specifically used for: Masking the protein pocket in the structural complex and some of the residues around it to obtain a masked protein sequence; The construction complex and the mask protein sequence are input into the trained protein sequence design model to obtain a plurality of protein enzyme sequences output by the protein sequence design model, and the target protein enzyme sequence is determined from the plurality of protein enzyme sequences.
[0073] Based on any of the above embodiments, the transformation unit is specifically used for: Inputting the structural complex into the encoder in the protein sequence design model to encode the structural information of the structural complex to obtain structural encoding information; The structural encoding information and the masked protein sequence are input into a decoder in the protein sequence design model to perform protein enzyme sequence prediction to obtain the multiple protein enzyme sequences.
[0074] Based on any of the above embodiments, the device further includes an optimization unit, configured to: Based on the protein sequence design model, other residues in the target protein enzyme sequence are designed to obtain an optimized protein enzyme sequence, wherein the other residues are residues in the target protein enzyme sequence except the pocket and some residues around it.
[0075] Based on any of the above embodiments, the optimization unit is specifically used for: Determining multiple sequence alignment information of the target protein enzyme sequence; Based on the protein sequence design model and the multiple sequence alignment information, other residues in the target protein enzyme sequence are designed to obtain the optimized protein enzyme sequence.
[0076] Based on any of the above embodiments, the optimization unit is specifically used for: Based on the protein sequence design model and the multiple sequence alignment information, other residues in the target protein enzyme sequence are designed to obtain multiple candidate protein enzyme sequences; Based on at least one of the similarity between the candidate protein enzyme sequence and the original sequence, the similarity between the candidate protein enzyme sequences, and the score of the candidate protein enzyme sequence, the candidate protein enzyme sequence is screened to obtain the optimized protein enzyme sequence.
[0077] Based on any of the above embodiments, the composite determination unit is specifically used for: docking the known substrate with the protein enzyme in the reference complex to obtain configurations of multiple known substrates; The structured complex is determined based on the energies of the complexes respectively corresponding to the configurations of the plurality of known substrates.
[0078] Figure 6 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 6 As shown, the electronic device may include: a processor 610, a communications interface 620, a memory 630 and a communications bus 640, wherein the processor 610, the communications interface 620 and the memory 630 communicate with each other through the communications bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the protein enzyme design method, which includes: determining a known substrate to be catalyzed; based on the known substrate or the related enzyme catalytic reaction mechanism, retrieving a reference complex from a pre-constructed protein enzyme database, and docking the known substrate with the protein enzyme in the reference complex to obtain a constructed complex; based on the trained protein sequence design model, modifying the protein pocket in the constructed complex and some residues around it to obtain a target protein enzyme sequence.
[0079] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.
[0080] On the other hand, the present invention also provides a computer program product, which includes a computer program, and the computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the protein enzyme design method provided by the above-mentioned methods, which method includes: determining a known substrate to be catalyzed; based on the known substrate or the related enzyme catalytic reaction mechanism, retrieving a reference complex from a pre-constructed protein enzyme database, and docking the known substrate with the protein enzyme in the reference complex to obtain a constructed complex; based on a trained protein sequence design model, modifying the protein pocket and some of the residues around it in the constructed complex to obtain a target protein enzyme sequence.
[0081] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the protein enzyme design method provided by the above methods, the method comprising: determining a known substrate to be catalyzed; based on the known substrate or a related enzyme catalytic reaction mechanism, retrieving a reference complex from a pre-constructed protein enzyme database, and docking the known substrate with the protein enzyme in the reference complex to obtain a constructed complex; based on a trained protein sequence design model, modifying the protein pocket and some of the residues around it in the constructed complex to obtain a target protein enzyme sequence.
[0082] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0083] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for designing a protein enzyme, characterized in that: include: Identify known substrates to be catalyzed; Based on the known substrate or the related enzyme catalytic reaction mechanism, a reference complex is retrieved from a pre-constructed protein enzyme database, and the known substrate is docked with the protein enzyme in the reference complex to obtain a constructed complex; Based on the trained protein sequence design model, the protein pocket in the structural complex and some of the residues around it are modified to obtain the target protein enzyme sequence.
2. The method for designing a protein enzyme according to claim 1, characterized in that: The protein pocket in the structural complex and some of the residues around it are modified based on the trained protein sequence design model to obtain the target protein enzyme sequence, including: Masking the protein pocket in the structural complex and some of the residues around it to obtain a masked protein sequence; The construction complex and the mask protein sequence are input into the trained protein sequence design model to obtain a plurality of protein enzyme sequences output by the protein sequence design model, and the target protein enzyme sequence is determined from the plurality of protein enzyme sequences.
3. The method for designing a protein enzyme according to claim 2, characterized in that: The step of inputting the construction complex and the mask protein sequence into the trained protein sequence design model to obtain a plurality of protein enzyme sequences output by the protein sequence design model comprises: Inputting the structural complex into the encoder in the protein sequence design model to encode the structural information of the structural complex to obtain structural encoding information; The structural encoding information and the masked protein sequence are input into a decoder in the protein sequence design model to perform protein enzyme sequence prediction to obtain the multiple protein enzyme sequences.
4. The method for designing a protein enzyme according to claim 1, characterized in that: After the protein pocket in the structural complex and some of the residues around it are modified based on the trained protein sequence design model to obtain the target protein enzyme sequence, the method further comprises: Based on the protein sequence design model, other residues in the target protein enzyme sequence are designed to obtain an optimized protein enzyme sequence, wherein the other residues are residues in the target protein enzyme sequence except the pocket and some residues around it.
5. The method for designing a protein enzyme according to claim 4, characterized in that: The method of designing other residues in the target protein enzyme sequence based on the protein sequence design model to obtain an optimized protein enzyme sequence includes: Determining multiple sequence alignment information of the target protein enzyme sequence; Based on the protein sequence design model and the multiple sequence alignment information, other residues in the target protein enzyme sequence are designed to obtain the optimized protein enzyme sequence.
6. The method for designing a protein enzyme according to claim 5, characterized in that: The step of designing other residues in the target protein enzyme sequence based on the protein sequence design model and the multiple sequence alignment information to obtain the optimized protein enzyme sequence comprises: Based on the protein sequence design model and the multiple sequence alignment information, other residues in the target protein enzyme sequence are designed to obtain multiple candidate protein enzyme sequences; Based on at least one of the similarity between the candidate protein enzyme sequence and the original sequence, the similarity between the candidate protein enzyme sequences, and the score of the candidate protein enzyme sequence, the candidate protein enzyme sequence is screened to obtain the optimized protein enzyme sequence.
7. The method for designing a protein enzyme according to any one of claims 1 to 6, characterized in that: The step of docking the known substrate with the protein enzyme in the reference complex to obtain a constructed complex comprises: docking the known substrate with the protein enzyme in the reference complex to obtain configurations of multiple known substrates; The structured complex is determined based on the energies of the complexes respectively corresponding to the configurations of the plurality of known substrates.
8. A protein enzyme design device, characterized in that: include: a substrate determination unit, for determining a known substrate to be catalyzed; A complex determination unit, for retrieving a reference complex from a pre-constructed protein enzyme database based on the known substrate or a related enzyme catalytic reaction mechanism, and docking the known substrate with the protein enzyme in the reference complex to obtain a constructed complex; The transformation unit is used to design a model based on the trained protein sequence, transform the protein pocket in the structural complex and some residues around it to obtain the target protein enzyme sequence.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for designing a protein enzyme according to any one of claims 1 to 7 is implemented.
10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for designing a protein enzyme according to any one of claims 1 to 7 is implemented.