Codon optimization method and device

Abstract

The invention discloses a codon optimization method and device.According to the codon optimization method, after an initial codon sequence corresponding to an input protein sequence is obtained, a naturalness score and a protein high expression probability are jointly used as optimization targets, iterative optimization is conducted on the initial codon sequence, and then the initial codon sequence is obtained; and obtaining a target codon sequence matched with the protein sequence. According to the method, the naturalness score and the high expression probability of the codon are jointly used as optimization targets, iterative optimization is carried out on the codon sequence, and the codon sequence which not only conforms to the context compatibility of the host codon but also has the high expression characteristic can be obtained.

Description

Technical Field

[0001] This application relates to the field of gene design technology in synthetic biology, specifically to codon optimization methods. This application also relates to codon optimization devices, electronic devices, and computer-readable storage media. Background Technology

[0002] Under natural conditions, exogenous genes (such as genes cloned from plants and introduced into E. coli, or genes cloned from humans and introduced into yeast) often exhibit low protein expression levels or even no expression due to a mismatch between their codon usage habits (or preferences) and the host. Codons are the core bridge between gene sequences and protein products. Codon optimization is a key technology in molecular biology and synthetic biology. Its core principle is to improve the expression efficiency of a target gene in a specific host organism by adjusting the gene's codon sequence without altering the amino acid sequence of the encoded protein. Essentially, it achieves molecular adaptation between exogenous genes and the host organism by adjusting the gene's codon sequence.

[0003] Existing technologies struggle to balance host codon context compatibility and high protein expression characteristics when optimizing codons. Summary of the Invention

[0004] This invention provides a codon optimization method, a codon optimization device, an electronic device, and a computer-readable storage medium to solve the problem that existing codon optimization methods have difficulty in simultaneously taking into account host codon context compatibility and high protein expression characteristics.

[0005] To solve or improve upon the aforementioned technical problems to some extent, according to one aspect of the present invention, a codon optimization method is provided, comprising: Obtain the initial codon sequence corresponding to the input protein sequence; By using both the naturalness score and the probability of high protein expression as optimization objectives, the initial codon sequence is iteratively optimized to obtain a target codon sequence that matches the protein sequence.

[0006] In one implementation, the step of iteratively optimizing the initial codon sequence by using both the naturalness score and the probability of high protein expression as optimization objectives includes: A: The first prediction module is used to predict and obtain a first naturalness score to characterize the degree of matching between the initial codon sequence and the host, and the second prediction module is used to predict and obtain a first high expression probability value to characterize the high expression effect of the initial codon sequence. B: Based on the first naturalness score, the first high expression probability value, and the intermediate results generated during prediction, the initial codon sequence is optimized to obtain the first intermediate codon sequence; C: The first prediction module is used to predict and obtain a second naturalness score to characterize the degree of matching between the first intermediate codon sequence and the host, and the second prediction module is used to predict and obtain a second high expression probability value to characterize the high expression effect of the first intermediate codon sequence; D: Based on the second naturalness score, the second highest expression probability value, and the intermediate results generated during prediction, the first intermediate codon sequence is optimized to obtain the second intermediate codon sequence; E: The first prediction module is used to predict and obtain a third naturalness score to characterize the degree of matching between the second intermediate codon sequence and the host, and the second prediction module is used to predict and obtain a third high expression probability value to characterize the high expression effect of the second intermediate codon sequence. F: Based on the third naturalness score, the third high expression probability value, and the intermediate results generated during prediction, the second intermediate codon sequence is optimized to obtain the third intermediate codon sequence; Repeat step BF above until the target codon sequence is obtained, which meets the predetermined naturalness requirements and protein expression level.

[0007] In one implementation, optimizing the initial codon sequence based on a first naturalness score, a first high expression probability value, and intermediate results generated during prediction includes: using the product of the first naturalness score and the first high expression probability value as a fitness score to optimize the initial codon sequence; or... The optimization of the first intermediate codon sequence based on the second naturalness score, the second high expression probability value, and the intermediate results generated during prediction includes: using the product of the second naturalness score and the second high expression probability value as a fitness score to optimize the first intermediate codon sequence; or, Based on the third naturalness score, the third high expression probability value, and the intermediate results generated during prediction, the second intermediate codon sequence is optimized, including: using the product between the third naturalness score and the third high expression probability value as a fitness score to optimize the second intermediate codon sequence.

[0008] In one implementation, optimizing the initial codon sequence includes: maximizing the fitness score using a genetic algorithm or a hallucination design method to optimize the initial codon sequence; or, Optimizing the first intermediate codon sequence includes: maximizing the fitness score using a genetic algorithm or illusion design method to optimize the first intermediate codon sequence; or, The second intermediate codon sequence is optimized by maximizing the fitness score using a genetic algorithm or illusion design method.

[0009] In one implementation, the first prediction module includes a prediction model trained using masked language modeling to evaluate the naturalness score of a codon sequence.

[0010] In one embodiment, obtaining the initial codon sequence corresponding to the input protein sequence includes: using the first prediction module to obtain the initial codon sequence by mask prediction.

[0011] In one implementation, the second prediction module is trained as follows: Construct a protein expression dataset, which includes associated protein abundance data and corresponding host species codon sequence information; The protein expression dataset is used as a training sample to train the model and obtain the second prediction module. The second prediction module is used to predict the high expression probability of the corresponding protein based on the codon sequence.

[0012] In one embodiment, the first prediction module includes a first protein language model and a first RNA language model; and / or, the second prediction module includes a second protein language model and a second RNA language model.

[0013] According to another aspect of the present invention, a codon optimization apparatus is provided, comprising: The initial codon sequence acquisition unit is used to obtain the initial codon sequence corresponding to the input protein sequence; The codon optimization unit is used to iteratively optimize the initial codon sequence by taking the naturalness score and the high protein expression probability as optimization objectives, so as to obtain a target codon sequence that matches the protein sequence.

[0014] According to another aspect of the present invention, an electronic device is provided, including a processor and a memory; wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the above-described codon optimization method.

[0015] According to another aspect of the present invention, a computer-readable storage medium is provided, on which one or more computer instructions are stored, which are executed by a processor to implement the above-described codon optimization method.

[0016] Compared with the prior art, the present invention has the following advantages: In the codon optimization method provided by this invention, after obtaining the initial codon sequence corresponding to the input protein sequence, the naturalness score and the high expression probability of the protein are used as optimization objectives to iteratively optimize the initial codon sequence and obtain a target codon sequence that matches the protein sequence. This method, by using both the naturalness score and the high expression probability of the codons as optimization objectives and iteratively optimizing the codon sequence, can obtain a codon sequence that both conforms to host codon context compatibility and possesses high expression characteristics. Attached Figure Description

[0017] Figure 1 This is a flowchart of the codon optimization method provided in the embodiments of this application; Figure 2 This is a unit block diagram of the cryptography optimization device provided in the embodiments of this application; Figure 3 This is a schematic diagram of the logical structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0018] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0019] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a,” “described,” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0020] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0021] In this application, the term "multiple" may refer to two or more, and "at least one" may refer to one, two or more.

[0022] The term "and / or" in this application is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this application generally indicates that the preceding and following related objects have an "or" relationship.

[0023] The codon is the basic unit of the genetic code. The codon sequence determines the type and order of amino acids during protein synthesis, thus guiding the translation of genetic information. Due to codon degeneracy, multiple codons can encode the same amino acid; these codons are called synonymous codons. Theoretically, for a given protein sequence, there can be many different coding sequences (CDS) encoding the same amino acid chain. However, the use of synonymous codons is not uniform; they exhibit different preferences that vary across different species, genes, and even different regions of the same gene. Influenced by factors such as the concentration of tRNA corresponding to different codons, mRNA secondary structure, and restriction endonuclease sites, the selection of synonymous codons has a complex impact on protein expression. Therefore, when designing DNA sequences for heterologous protein expression, it is essential to match the host's codon usage preferences (CUPs); this process is called codon optimization.

[0024] Existing technologies struggle to balance host codon context compatibility and high expressiveness. Furthermore, existing technologies typically rely on unsupervised training based on massive amounts of data. For host species lacking sufficient training samples, such as those with small sample sizes and weak research foundations (e.g., some new plant and microbial species), traditional optimization methods that depend on massive host gene data training are difficult to apply and cannot perform effective codon optimization. Therefore, the application scope of existing codon optimization methods is relatively limited and lacks flexibility.

[0025] To address the limitation in the application scope of existing codon optimization methods, this application provides a codon optimization method, a corresponding codon optimization apparatus, an electronic device, and a computer-readable storage medium. The following embodiments provide a detailed description of the aforementioned codon optimization method, apparatus, electronic device, and computer-readable storage medium.

[0026] This application provides a cryptographic optimization method, which can be applied to computing devices used to implement cryptographic optimization. Figure 1 The flowchart of the codon optimization method provided in the embodiments of this application is as follows. Figure 1 The codon optimization method provided in this embodiment will be described in detail. The embodiments described below are used to explain the principle of the codon optimization method and are not intended to limit actual use.

[0027] like Figure 1 As shown, the codon optimization method provided in this embodiment includes the following steps: S101, obtain the initial codon sequence corresponding to the input protein sequence.

[0028] This step is used to obtain the initial codon sequence corresponding to the input protein sequence. The codon sequence (CDS) is the core region of a gene sequence that can be transcribed into mRNA and further translated into protein. The range of the CDS sequence strictly corresponds to the amino acid sequence of the protein, starting with the start codon and ending with the stop codon. The input protein sequence (i.e., the amino acid sequence) is the target benchmark for codon optimization to ensure that the optimized codon sequence still encodes the protein. The initial codon sequence is the object of optimization, and the purpose of optimization is to improve expression efficiency by adjusting its encoding method.

[0029] S102 uses both the naturalness score and the probability of high protein expression as optimization objectives to iteratively optimize the initial codon sequence and obtain the target codon sequence that matches the protein sequence.

[0030] After obtaining the initial codon sequence corresponding to the input protein sequence in the above steps, this step is used to iteratively optimize the initial codon sequence by taking the naturalness score and the high protein expression probability as optimization objectives, so as to obtain the target codon sequence that matches the protein sequence.

[0031] In this embodiment, each iteration of optimization includes a prediction phase and an optimization phase. The naturalness score and the probability of high protein expression are used as optimization objectives to iteratively optimize the initial codon sequence. This includes the following: In the prediction phase, a first prediction module predicts the naturalness score of the codon sequence, and a second prediction module predicts the probability of high protein expression. In the optimization phase, the naturalness score predicted by the first prediction module and the high expression probability predicted by the second prediction module are used as optimization objectives, and this process is iteratively repeated to ensure that the final target codon sequence is both compatible with host codon context and has a high probability of high protein expression. In other words, the target codon sequence obtained using this method maximizes the probability of high protein expression while maintaining naturalness.

[0032] In this embodiment, the first prediction module includes a first protein language model and a first RNA language model. The first protein language model is used to extract features from the protein sequence to obtain protein sequence features, and the first RNA language model is used to extract features from the codon sequence to be optimized (in iteration) to obtain coding sequence features. The terms in the RNA language model are in units of codons, and correspond one-to-one with the terms in the protein language model in units of amino acids. These two sets of features can be integrated by weighted addition (to comprehensively represent coding sequence and protein information) and serve as the basis data for predicting the naturalness score of the codon sequence. By using a pre-trained first protein language model (e.g., the ESM2 model) and a first RNA language model (e.g., the mRNA-FM model), since the above models have already learned and captured the rich semantic and structural information of biological sequences, this embodiment does not require training the model from scratch, greatly reducing the dependence on the scale of the target species data and achieving effective learning with a smaller scale of data.

[0033] The first prediction module also includes a prediction model trained using masked language modeling to evaluate the naturalness score of codon sequences. For example, it can be trained using a lexical masking modeling task to learn the codon usage patterns of endogenous proteins in a specific species' genome and the codon usage rules under the context of the host species. Based on amino acids and the context of the codon sequence, it can predict synonymous codons at a certain position. The naturalness score predicted by this model can be used to measure the compatibility between the codon sequence and the host species, i.e., whether it approximates the codon usage patterns of endogenous proteins in the host species. Correspondingly, obtaining the initial codon sequence corresponding to the input protein sequence in step S101 can specifically refer to obtaining the initial codon sequence using the first prediction module through masked prediction. For example, constructing a codon sequence of the same length as the input protein sequence, where all codons are padded with... <mask>Using the CDS as the initial input sequence, the prediction model of the first prediction module predicts the synonymous codon with the highest probability corresponding to the amino acid at each position, and the predicted codons form the initial codon sequence mentioned above.

[0034] The second prediction module includes a second protein language model and a second RNA language model. The second protein language model is used to extract features from the protein sequence to obtain protein sequence features, and the second RNA language model is used to extract features from the codon sequence to be optimized (in iteration) to obtain coding sequence features. These two sets of features are integrated by weighted addition (to comprehensively represent coding sequence and protein information) and used as the basis data for predicting the high expression probability value of the codon sequence.

[0035] In this embodiment, the second prediction module is trained and obtained in the following manner: First, a protein expression dataset is constructed, which includes associated protein abundance data and codon sequence information of the corresponding host species. For example, non-redundant codon sequences and their translated protein sequences are collected from the reference genomes of various plant species (host species). Then, protein sequences (or other protein abundance data) in the protein abundance database PaxDb are mapped to the aforementioned non-redundant codon sequence dataset. During mapping, if a protein sequence that is completely identical to the non-redundant codon sequence is found, it is directly used as the matching result. If no completely matching entry is found, a similarity comparison is performed. For example, in the comparison results, the sequence with the smallest E value and the similarity of the comparison region not less than a predetermined proportion (e.g., 90%), the coverage of the comparison region not less than a predetermined proportion (e.g., 50%), is selected as the matching result. This obtains the correspondence between the codon sequence of each plant species and whether the protein encoded by the codon sequence exists in high abundance in vivo.

[0036] Secondly, the protein expression dataset is used as training samples for model training to obtain a second prediction module. This second prediction module is used to predict the probability of high protein expression based on codon sequences. Unlike traditional unsupervised models, the above model training method enables the second prediction module to directly correlate codon sequence features with protein expression levels, effectively capturing codon factors that regulate protein production.

[0037] In this embodiment, during the prediction phase, the naturalness score of the codon sequence is predicted by the first prediction module, and the high expression probability of the protein in the codon sequence is predicted by the second prediction module. During the optimization phase, the naturalness score predicted by the first prediction module and the high expression probability value predicted by the second prediction module are used together as the optimization target, and the process is iteratively repeated based on this target. Specifically, this may include the following process: A: The first prediction module is used to predict and obtain the first naturalness score, which characterizes the degree of matching between the initial codon sequence and the host, and the second prediction module is used to predict and obtain the first high expression probability value, which characterizes the high expression effect of the initial codon sequence. B: Based on the first naturalness score, the first high expression probability value, and the intermediate results generated during prediction, the initial codon sequence is optimized to obtain the first intermediate codon sequence. The intermediate results generated during prediction refer to gradient information. That is, while predicting the high expression probability of a codon sequence, the gradient information of the high expression probability value relative to the encoding vector of the codon at each position can be calculated. This gradient information can guide the codon at that position to change to other codons (i.e., the optimization process), thereby increasing the high expression probability value. C: Use the first prediction module to predict and obtain a second naturalness score to characterize the degree of matching between the first intermediate codon sequence and the host, and use the second prediction module to predict and obtain a second high expression probability value to characterize the high expression effect of the first intermediate codon sequence; D: Based on the second naturalness score, the second highest expression probability value, and the intermediate results generated during prediction, the first intermediate codon sequence is optimized to obtain the second intermediate codon sequence; E: The first prediction module is used to predict and obtain a third naturalness score to characterize the degree of matching between the second intermediate codon sequence and the host, and the second prediction module is used to predict and obtain a third high expression probability value to characterize the high expression effect of the second intermediate codon sequence. F: Based on the third naturalness score, the third high expression probability value, and the intermediate results generated during prediction, the second intermediate codon sequence is optimized to obtain the third intermediate codon sequence; Repeat step BF above until the target codon sequence is obtained. The target codon sequence meets the predetermined naturalness requirements and protein expression level, for example, when the high expression probability value (E) > 0.9 and the naturalness score (N) > 0.6, or when the fitness score is the highest and has not improved for several consecutive rounds, stop the iteration.

[0038] In this embodiment, step B above optimizes the initial codon sequence based on the first naturalness score, the first high expression probability value, and the intermediate results generated during prediction. Specifically, this can mean: using the product of the first naturalness score (N) and the first high expression probability value (E) as the fitness score (F) to optimize the initial codon sequence; wherein, the specific method for optimizing the initial codon sequence can be: using a genetic algorithm (GA) or hallucination design method to maximize the fitness score to optimize the initial codon sequence.

[0039] Similarly, in step D above, the first intermediate codon sequence is optimized based on the second naturalness score, the second high expression probability value, and the intermediate results generated during prediction. Specifically, this can mean: using the product of the second naturalness score and the second high expression probability value as the fitness score to optimize the first intermediate codon sequence; the specific method for optimizing the first intermediate codon sequence can also be: using a genetic algorithm or illusion design method to maximize the fitness score to optimize the first intermediate codon sequence. Similarly, in step F above, the second intermediate codon sequence is optimized based on the third naturalness score, the third high expression probability value, and the intermediate results generated during prediction. Specifically, this can mean using the product of the third naturalness score and the third high expression probability value as the fitness score to optimize the second intermediate codon sequence. The specific method for optimizing the second intermediate codon sequence can also be to maximize the fitness score using a genetic algorithm or illusion design method to optimize the second intermediate codon sequence.

[0040] For example, when using a genetic algorithm to maximize fitness scores, an initial population is first generated based on the input codon sequence at a 5% codon synonym substitution rate, with each individual representing a synonymous codon combination. During population evaluation, the algorithm uses multiple independent models (e.g., 5 independent models) from the second prediction module to calculate a high expression probability value (E) for each candidate sequence. in, For the first The model predicts a high expression probability. Simultaneously, multiple samplings are performed, with each codon sampled with a certain probability (e.g., 15%). These codons have a high probability (e.g., 80%) of being replaced with unknown codons. <mask>A token has a certain probability (e.g., 10%) of being replaced by a random codon, and a certain probability (e.g., 10%) of retaining the original codon.

[0041] The naturalness score of the codon sequence is calculated using the first prediction module: in, The number of times the codon is randomly sampled (e.g.) =10), The number of codons sampled in each time. The output of the first prediction module is in the 1st... In the second sampling Each codon represents the probability of the original codon at that position in the candidate sequence. Then, the sequence fitness is calculated. In subsequent iterations, each iteration first selects the sequence with the highest fitness from the previous generation to enter the next generation without mutation. Simultaneously, the top-ranking individuals (e.g., the top 20%) are selected as parents for subsequent crossover and mutation. The crossover operation randomly selects a codon cut point between two parents, generating offspring with the first half from one parent and the second half from the other. The mutation operation synonymously replaces a predetermined percentage (5%) of the individual's codons. After each generation forms a new population, the high expression probability of all sequences in that generation is recalculated. Naturalness and suitability The iteration stops when the maximum fitness value remains unchanged for a predetermined number of consecutive iterations, or when the number of iterations exceeds the preset number.

[0042] For example, when using Hallucination Design to maximize fitness scores, gradient-based iterative optimization of the codon sequence is performed. The entire process involves multiple iterations. The model achieves generational optimization by balancing the high expression probability boost brought by synonymous substitutions at each codon position with the preservation of sequence naturalness.

[0043] The codon optimization method provided in this embodiment splits codon optimization into two tasks: matching host codon context compatibility (ensuring the naturalness of the codon sequence) and maximizing the probability of high protein expression. A model framework is constructed to execute these two tasks respectively. For example, during the iteration process, the first prediction module predicts the naturalness of the codon and the second prediction module predicts the probability of high expression. Then, a genetic algorithm or illusion design is used to integrate the above two modules, and the dual objectives are iteratively optimized to generate a target codon sequence that both meets the host codon context compatibility and has high expression characteristics.

[0044] Furthermore, this method no longer relies on massive amounts of host data. It can train models with a small amount of data to predict the naturalness scores of codons and the probability of high protein expression, making it particularly suitable for special hosts such as plants and new species. For example, the first prediction module includes a first protein language model and a first RNA language model, and the second prediction module includes a second protein language model and a second RNA language model. These pre-trained language models have already learned and captured rich semantic and structural information of biological sequences. Therefore, in this embodiment, there is no need to train the model from scratch, which greatly reduces the dependence on the scale of target species data and achieves effective learning with a smaller scale of data.

[0045] It is particularly noteworthy that this embodiment incorporates protein expression abundance annotation information (i.e., the protein expression dataset includes associated protein abundance data and corresponding host species codon sequence information) when training the second prediction module. Therefore, through supervised learning, the model can directly correlate codon sequence features with protein expression levels, effectively capturing codon factors that regulate protein yield. Furthermore, by utilizing the aforementioned protein expression abundance annotation information, the information density of the entire training dataset is increased, and the required data volume is significantly reduced, making the method provided in this embodiment particularly suitable for special hosts such as plants and new species.

[0046] The above embodiments provide a codon optimization method. Correspondingly, another embodiment of this application also provides a codon optimization device. Since the codon optimization device embodiment is basically similar to the codon optimization method embodiment, it is described in a relatively simple way. For details of the relevant technical features, please refer to the corresponding description of the codon optimization method embodiment provided above. The following description of the codon optimization device embodiment is merely illustrative.

[0047] Please refer to Figure 2 Understanding this embodiment, Figure 2 This is a unit block diagram of the codon optimization device provided in this embodiment, such as... Figure 2 As shown, the codon optimization apparatus provided in this embodiment includes: Initial codon sequence acquisition unit 201 is used to obtain the initial codon sequence corresponding to the input protein sequence; The codon optimization unit 202 is used to iteratively optimize the initial codon sequence by taking the naturalness score and the high protein expression probability as optimization objectives, so as to obtain the target codon sequence that matches the protein sequence.

[0048] The step of iteratively optimizing the initial codon sequence by using both the naturalness score and the probability of high protein expression as optimization objectives includes: A: The first prediction module is used to predict and obtain a first naturalness score to characterize the degree of matching between the initial codon sequence and the host, and the second prediction module is used to predict and obtain a first high expression probability value to characterize the high expression effect of the initial codon sequence. B: Based on the first naturalness score, the first high expression probability value, and the intermediate results generated during prediction, the initial codon sequence is optimized to obtain the first intermediate codon sequence; C: The first prediction module is used to predict and obtain a second naturalness score to characterize the degree of matching between the first intermediate codon sequence and the host, and the second prediction module is used to predict and obtain a second high expression probability value to characterize the high expression effect of the first intermediate codon sequence; D: Based on the second naturalness score, the second highest expression probability value, and the intermediate results generated during prediction, the first intermediate codon sequence is optimized to obtain the second intermediate codon sequence; E: The first prediction module is used to predict and obtain a third naturalness score to characterize the degree of matching between the second intermediate codon sequence and the host, and the second prediction module is used to predict and obtain a third high expression probability value to characterize the high expression effect of the second intermediate codon sequence. F: Based on the third naturalness score, the third high expression probability value, and the intermediate results generated during prediction, the second intermediate codon sequence is optimized to obtain the third intermediate codon sequence; Repeat step BF above until the target codon sequence is obtained, which meets the predetermined naturalness requirements and protein expression level.

[0049] The optimization of the initial codon sequence based on the first naturalness score, the first high expression probability value, and intermediate results generated during prediction includes: using the product of the first naturalness score and the first high expression probability value as a fitness score to optimize the initial codon sequence; or, The optimization of the first intermediate codon sequence based on the second naturalness score, the second high expression probability value, and the intermediate results generated during prediction includes: using the product of the second naturalness score and the second high expression probability value as a fitness score to optimize the first intermediate codon sequence; or, Based on the third naturalness score, the third high expression probability value, and the intermediate results generated during prediction, the second intermediate codon sequence is optimized, including: using the product between the third naturalness score and the third high expression probability value as a fitness score to optimize the second intermediate codon sequence.

[0050] Optimizing the initial codon sequence includes: maximizing the fitness score using a genetic algorithm or illusion design method to optimize the initial codon sequence; or, Optimizing the first intermediate codon sequence includes: maximizing the fitness score using a genetic algorithm or illusion design method to optimize the first intermediate codon sequence; or, The second intermediate codon sequence is optimized by maximizing the fitness score using a genetic algorithm or illusion design method.

[0051] The first prediction module includes a prediction model trained using masked language modeling to evaluate the naturalness score of a codon sequence.

[0052] Obtaining the initial codon sequence corresponding to the input protein sequence includes: using the first prediction module to obtain the initial codon sequence through mask prediction.

[0053] The second prediction module is trained in the following way: Construct a protein expression dataset, which includes associated protein abundance data and corresponding host species codon sequence information; The protein expression dataset is used as a training sample to train the model and obtain the second prediction module. The second prediction module is used to predict the high expression probability of the corresponding protein based on the codon sequence.

[0054] The first prediction module includes a first protein language model and a first RNA language model; and / or, the second prediction module includes a second protein language model and a second RNA language model.

[0055] This invention also provides an electronic device that can be configured with the aforementioned cryptographic optimization device in program form to execute the cryptographic optimization method provided in this invention. This electronic device can be a desktop computer, laptop computer, smartphone, or IoT (Internet of Things) device, or a server-side device such as a conventional server, cloud server, or server array. Optionally, one possible hardware structure of this electronic device may be as follows: Figure 3 As shown, it includes: at least one processor 301, at least one communication interface 302, at least one memory 303 and at least one communication bus 304; Optionally, the communication interface 302 can be an interface of a communication module, such as the interface of a GSM module; The processor 301 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

[0056] Memory 303 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0057] The memory 303 stores a program, and the processor 01 calls the program stored in the memory 303 to execute the codon optimization method provided in this embodiment of the invention.

[0058] For a detailed description of the implementation process of each action by the processor, please refer to the relevant descriptions in the aforementioned embodiments of the codon optimization method and the codon optimization device, which will not be repeated here.

[0059] In the above embodiments, a codon optimization method, a codon optimization apparatus, and an electronic device are provided. Furthermore, another embodiment of this application provides a computer-readable storage medium for implementing the above-described codon optimization method. The computer-readable storage medium embodiments provided in this application are described relatively simply; relevant parts can be found in the corresponding descriptions of the codon optimization method embodiments described above. The embodiments described below are merely illustrative. The computer-readable storage medium provided in this embodiment stores computer instructions, which, when executed by a processor, perform the following steps: Obtain the initial codon sequence corresponding to the input protein sequence; By using both the naturalness score and the probability of high protein expression as optimization objectives, the initial codon sequence is iteratively optimized to obtain a target codon sequence that matches the protein sequence.

[0060] The step of iteratively optimizing the initial codon sequence by using both the naturalness score and the probability of high protein expression as optimization objectives includes: A: The first prediction module is used to predict and obtain a first naturalness score to characterize the degree of matching between the initial codon sequence and the host, and the second prediction module is used to predict and obtain a first high expression probability value to characterize the high expression effect of the initial codon sequence. B: Based on the first naturalness score, the first high expression probability value, and the intermediate results generated during prediction, the initial codon sequence is optimized to obtain the first intermediate codon sequence; C: The first prediction module is used to predict and obtain a second naturalness score to characterize the degree of matching between the first intermediate codon sequence and the host, and the second prediction module is used to predict and obtain a second high expression probability value to characterize the high expression effect of the first intermediate codon sequence; D: Based on the second naturalness score, the second highest expression probability value, and the intermediate results generated during prediction, the first intermediate codon sequence is optimized to obtain the second intermediate codon sequence; E: The first prediction module is used to predict and obtain a third naturalness score to characterize the degree of matching between the second intermediate codon sequence and the host, and the second prediction module is used to predict and obtain a third high expression probability value to characterize the high expression effect of the second intermediate codon sequence. F: Based on the third naturalness score, the third high expression probability value, and the intermediate results generated during prediction, the second intermediate codon sequence is optimized to obtain the third intermediate codon sequence; Repeat step BF above until the target codon sequence is obtained, which meets the predetermined naturalness requirements and protein expression level.

[0061] The optimization of the initial codon sequence based on the first naturalness score, the first high expression probability value, and intermediate results generated during prediction includes: using the product of the first naturalness score and the first high expression probability value as a fitness score to optimize the initial codon sequence; or, The optimization of the first intermediate codon sequence based on the second naturalness score, the second high expression probability value, and the intermediate results generated during prediction includes: using the product of the second naturalness score and the second high expression probability value as a fitness score to optimize the first intermediate codon sequence; or, Based on the third naturalness score, the third high expression probability value, and the intermediate results generated during prediction, the second intermediate codon sequence is optimized, including: using the product between the third naturalness score and the third high expression probability value as a fitness score to optimize the second intermediate codon sequence.

[0062] Optimizing the initial codon sequence includes: maximizing the fitness score using a genetic algorithm or illusion design method to optimize the initial codon sequence; or, Optimizing the first intermediate codon sequence includes: maximizing the fitness score using a genetic algorithm or illusion design method to optimize the first intermediate codon sequence; or, The second intermediate codon sequence is optimized by maximizing the fitness score using a genetic algorithm or illusion design method.

[0063] The first prediction module includes a prediction model trained using masked language modeling to evaluate the naturalness score of a codon sequence.

[0064] Obtaining the initial codon sequence corresponding to the input protein sequence includes: using the first prediction module to obtain the initial codon sequence through mask prediction.

[0065] The second prediction module is trained in the following way: Construct a protein expression dataset, which includes associated protein abundance data and corresponding host species codon sequence information; The protein expression dataset is used as a training sample to train the model and obtain the second prediction module. The second prediction module is used to predict the high expression probability of the corresponding protein based on the codon sequence.

[0066] The first prediction module includes a first protein language model and a first RNA language model; and / or, the second prediction module includes a second protein language model and a second RNA language model.

[0067] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0068] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0069] 1. Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any cryptographic optimization method or technique. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include non-transitory computer-readable media, such as modulated data signals and carrier waves.

[0070] 2. Those skilled in the art will understand that embodiments of this application can be provided as codon optimization methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0071] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.< / mask> < / mask>

Claims

1. A codon optimization method, characterized in that, include: Obtain the initial codon sequence corresponding to the input protein sequence; By using both the naturalness score and the probability of high protein expression as optimization objectives, the initial codon sequence is iteratively optimized to obtain a target codon sequence that matches the protein sequence.

2. The codon optimization method according to claim 1, characterized in that, The step of iteratively optimizing the initial codon sequence by using both the naturalness score and the probability of high protein expression as optimization objectives includes: A: The first prediction module is used to predict and obtain a first naturalness score to characterize the degree of matching between the initial codon sequence and the host, and the second prediction module is used to predict and obtain a first high expression probability value to characterize the high expression effect of the initial codon sequence. B: Based on the first naturalness score, the first high expression probability value, and the intermediate results generated during prediction, the initial codon sequence is optimized to obtain the first intermediate codon sequence; C: The first prediction module is used to predict and obtain a second naturalness score to characterize the degree of matching between the first intermediate codon sequence and the host, and the second prediction module is used to predict and obtain a second high expression probability value to characterize the high expression effect of the first intermediate codon sequence; D: Based on the second naturalness score, the second highest expression probability value, and the intermediate results generated during prediction, the first intermediate codon sequence is optimized to obtain the second intermediate codon sequence; E: The first prediction module is used to predict and obtain a third naturalness score to characterize the degree of matching between the second intermediate codon sequence and the host, and the second prediction module is used to predict and obtain a third high expression probability value to characterize the high expression effect of the second intermediate codon sequence. F: Based on the third naturalness score, the third high expression probability value, and the intermediate results generated during prediction, the second intermediate codon sequence is optimized to obtain the third intermediate codon sequence; Repeat step BF above until the target codon sequence is obtained, which meets the predetermined naturalness requirements and protein expression level.

3. The codon optimization method according to claim 2, characterized in that, The optimization of the initial codon sequence based on the first naturalness score, the first high expression probability value, and intermediate results generated during prediction includes: using the product of the first naturalness score and the first high expression probability value as a fitness score to optimize the initial codon sequence; or, The optimization of the first intermediate codon sequence based on the second naturalness score, the second high expression probability value, and the intermediate results generated during prediction includes: using the product of the second naturalness score and the second high expression probability value as a fitness score to optimize the first intermediate codon sequence; or, Based on the third naturalness score, the third high expression probability value, and the intermediate results generated during prediction, the second intermediate codon sequence is optimized, including: using the product between the third naturalness score and the third high expression probability value as a fitness score to optimize the second intermediate codon sequence.

4. The codon optimization method according to claim 3, characterized in that, Optimizing the initial codon sequence includes: maximizing the fitness score using a genetic algorithm or illusion design method to optimize the initial codon sequence; or, Optimizing the first intermediate codon sequence includes: maximizing the fitness score using a genetic algorithm or illusion design method to optimize the first intermediate codon sequence; or, The second intermediate codon sequence is optimized by maximizing the fitness score using a genetic algorithm or illusion design method.

5. The codon optimization method according to claim 2, characterized in that, The first prediction module includes a prediction model trained using masked language modeling to evaluate the naturalness score of a codon sequence.

6. The codon optimization method according to claim 5, characterized in that, Obtaining the initial codon sequence corresponding to the input protein sequence includes: using the first prediction module to obtain the initial codon sequence through mask prediction.

7. The codon optimization method according to claim 2, characterized in that, The second prediction module is trained in the following way: Construct a protein expression dataset, which includes associated protein abundance data and codon sequence information of the corresponding host species; The protein expression dataset is used as a training sample to train the model and obtain the second prediction module. The second prediction module is used to predict the high expression probability of the corresponding protein based on the codon sequence.

8. The codon optimization method according to claim 2, characterized in that, The first prediction module includes a first protein language model and a first RNA language model; and / or, the second prediction module includes a second protein language model and a second RNA language model.

9. A codon optimization device, characterized in that, include: The initial codon sequence acquisition unit is used to obtain the initial codon sequence corresponding to the input protein sequence; The codon optimization unit is used to iteratively optimize the initial codon sequence by taking the naturalness score and the high protein expression probability as optimization objectives, so as to obtain a target codon sequence that matches the protein sequence.

10. An electronic device, characterized in that, Includes processor and memory; among which, The memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the codon optimization method as described in any one of claims 1-8.

11. A computer-readable storage medium storing one or more computer instructions thereon, characterized in that, The instruction is executed by the processor to implement the codon optimization method as described in any one of claims 1-8.

Images