Codon optimization method and apparatus, device, and storage medium

Abstract

Provided are a codon optimization method and apparatus, a device, and a storage medium. The method comprises: acquiring a target protein sequence; and inputting the target protein sequence into a pre-trained target codon optimization model to obtain a codon optimization result of the target protein sequence, wherein the target codon optimization model comprises at least one neural network module, and each neural network module comprises a recurrent neural network and a gated recurrent unit network.

Description

A codon optimization method, device, equipment and storage medium

[0001] The present application claims priority to the Chinese patent application No. 202410415052.9, filed on April 8, 2024, with the Chinese Patent Office, the content of which is incorporated herein by reference in its entirety. TECHNICAL FIELD

[0002] The present application relates to the technical field of bioengineering, for example, to a codon optimization method, device, equipment and storage medium. BACKGROUND

[0003] The demand for improving protein expression in the industry and medical field is widespread, and the pursuit of expression improvement generally has no upper limit. Related technologies can perform codon optimization based on a trained neural network model, wherein the model is composed of a 128-layer recurrent neural network (RNN) and a 16-layer bidirectional recurrent neural network (Bi-RNN), but the accuracy of the neural network model on the test set is generally lower than 0.53, and the improvement effect of protein expression is limited.

[0004] SUMMARY

[0005] The present application provides a codon optimization method, device, equipment and storage medium, which can perform codon optimization on protein sequences based on a model composed of a recurrent neural network and a gated recurrent unit network to improve protein expression effect.

[0006] In a first aspect, the present application provides a codon optimization method, which comprises:

[0007] obtaining a target protein sequence;

[0008] inputting the target protein sequence into a pre-trained target codon optimization model to obtain a codon optimization result of the target protein sequence;

[0009] wherein the target codon optimization model comprises at least one neural network module, and each neural network module comprises a recurrent neural network and a gated recurrent unit network.

[0010] In a second aspect, the present application provides a codon optimization device, which comprises:

[0011] a protein sequence acquisition module configured to acquire a target protein sequence;

[0012] a codon optimization module configured to input the target protein sequence into a pre-trained target codon optimization model to obtain a codon optimization result of the target protein sequence.

[0013] The target codon optimization model comprises at least one neural network module, and each neural network module comprises a recurrent neural network and a gated recurrent unit network.

[0014] In a third aspect, the present application provides a computer device, comprising:

[0015] at least one processor;

[0016] a memory configured to store at least one program;

[0017] When the at least one program is executed by the at least one processor, the at least one processor implements the codon optimization method of the first aspect.

[0018] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the codon optimization method of the first aspect. BRIEF DESCRIPTION OF DRAWINGS

[0019] FIG. 1 is a flowchart of a codon optimization method provided by the present application;

[0020] FIG. 2 is a flowchart of another codon optimization method provided by the present application;

[0021] FIG. 3 is a schematic diagram of the association between the accuracy of a model and a network module provided by the present application;

[0022] FIG. 4 is a schematic diagram of the results of a neural network module provided by the present application;

[0023] FIG. 5 is a schematic diagram of the structure of an initial codon optimization model provided by the present application;

[0024] FIG. 6 is a schematic diagram of the structure of a codon optimization model in the related art provided by the present application;

[0025] FIG. 7 is a schematic diagram of the comparison of the accuracy of the model of the present application and the model in the related art;

[0026] FIG. 8 is a schematic diagram of the comparison of the fluorescence intensity of the output sequence of the model of the present application and the output sequence of the model in the related art provided by the present application;

[0027] FIG. 9 is a schematic diagram of the structure of a codon optimization device provided by the present application;

[0028] FIG. 10 is a schematic diagram of the structure of a computer device provided by the present application. DETAILED DESCRIPTION

[0029] The technical solutions and advantages of the embodiments of the present application will be more clearly understood from the following description of the embodiments of the present application with reference to the accompanying drawings. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0030] FIG. 1 is a flowchart of a codon optimization method provided by the present application. The present application can be applied to the scenario of codon optimization of protein sequences. The method can be executed by a codon optimization device, which can be implemented in the form of software and / or hardware.

[0031] As shown in FIG. 1, the codon optimization method includes S110-S120.

[0032] S110, obtaining a target protein sequence.

[0033] The target protein sequence can be a protein sequence that needs to be codon optimized. For example, the target protein sequence is composed of at least one combination of amino acids. For example, a codon is a triplet nucleotide sequence on DNA. The arrangement of 4 nucleotides (A, T, C, G) in three different positions forms 64 codons, and 61 codons can be translated into amino acids except for 3 stop codons. The number of codon types is much larger than the number of amino acid types (20), and the same amino acid can be encoded by different codons, that is, codons are degenerate. Related research has found that changing the codon corresponding to the amino acid can affect the expression amount of the protein. Therefore, the technical solution of the present application wants to find the most suitable codon for the amino acid in the protein sequence, so as to improve the expression amount of the protein.

[0034] S120, inputting the target protein sequence into a pre-trained target codon optimization model to obtain a codon optimization result of the target protein sequence.

[0035] The target codon optimization model can be a neural network model used for codon optimization. The target codon optimization model includes at least one neural network module, and each neural network module includes a recurrent neural network and a gated recurrent unit network. The recurrent neural network (RNN) is good at processing the proximity relationship between the protein and the nucleic acid sequence, such as local high-frequency codon selection, codon fitness configuration, and adverse motif avoidance. The gated recurrent unit network (GRU) is good at processing the long-distance relationship between the protein and the nucleic acid sequence, such as splice site avoidance. By constructing the target codon optimization model based on the recurrent neural network and the gated recurrent unit network, the target codon optimization model can improve the ability of local feature processing and global processing, and improve the accuracy of codon optimization based on the model.

[0036] Since an amino acid can correspond to multiple codon expressions, the target codon optimization model can learn the selection rules between the amino acid and the adapted codon expression through training, and then obtain a codon optimization result. The codon optimization result can be an output result after codon optimization. For example, the codon optimization result can be composed of codons corresponding to each amino acid, that is, the codon optimization result can be a nucleic acid sequence corresponding to the target protein sequence.

[0037] For example, based on the target codon optimization model, codon optimization is performed on the target protein sequence, including: determining at least one amino acid to be optimized in the target protein sequence, determining at least one candidate codon according to the type of the amino acid to be optimized, for each amino acid to be optimized, determining a target codon corresponding to the amino acid to be optimized from the candidate codon according to the correspondence relationship between the type of the amino acid and the adapted codon obtained through pre-training, and finally determining a codon optimization result according to the target codons corresponding to all amino acids to be optimized.

[0038] The technical solution provided in the present application includes the following steps: obtaining a target protein sequence; inputting the target protein sequence into a target codon optimization model that has been pre-trained to obtain a codon optimization result of the target protein sequence; and the target codon optimization model includes at least one neural network module, and each neural network module includes a recurrent neural network and a gated recurrent unit network. The technical solution provided in the present application solves the problem of insufficient protein expression effect in the related art codon optimization technology, and can perform codon optimization on the protein sequence based on the model composed of the recurrent neural network and the gated recurrent unit network, thereby improving the protein expression effect.

[0039] FIG. 2 is a flowchart of another method for codon optimization provided by the present application, which can be applied to the scenario of codon optimization of a protein sequence. This embodiment is based on the above and illustrates how to input the target protein sequence into a pre-trained target codon optimization model to obtain the codon optimization result of the target protein sequence. The device can be realized by software and / or hardware and integrated into a computer device with application development function.

[0040] As shown in FIG. 2, the codon optimization method includes S210-S230.

[0041] S210, obtaining a target protein sequence.

[0042] The target protein sequence can be a protein sequence that needs to be codon optimized. For example, the target protein sequence is composed of at least one combination of amino acids. For example, a codon is a triplet nucleotide sequence on DNA. The arrangement of 4 nucleotides (A, T, C, G) in three different positions forms 64 codons, and 61 codons can be translated into amino acids except for 3 stop codons. The number of codons is much larger than the number of amino acids (20), and the same amino acid can be encoded by different codons, i.e., codons are degenerate. Related research has found that changing the codon corresponding to the amino acid can affect the expression amount of the protein. Therefore, the technical solution of the present application wants to find the most suitable codon for the amino acid in the protein sequence, thereby improving the expression amount of the protein.

[0043] S220, performing matrix processing on the target protein sequence to obtain a protein sequence matrix.

[0044] The protein sequence matrix can be a matrix composed of at least one vector. The amino acids in the protein sequence matrix are represented in vector form. Since the number of types of amino acids is 20, and considering the stop codon, the vector representing the amino acid has 21 dimensions, so 21 vectors representing amino acids can be set, and one vector corresponds to one amino acid. For example, the vector of alanine can be represented as: [1, 0, 0, …, 0].

[0045] S230, inputting the protein sequence matrix into the target codon optimization model to obtain a target nucleic acid sequence matrix, and determining a codon optimization result according to the target nucleic acid sequence matrix.

[0046] The target codon optimization model can be a neural network model for codon optimization. The target codon optimization model includes at least one neural network module, which includes a recurrent neural network and a gated recurrent unit network. The recurrent neural network (RNN) is a type of artificial neural network specially designed for processing sequence information. It is characterized by the fact that the hidden layer performs the same task on each element of the input information in a time sequence, and links the elements of the input information in a time sequence, thereby realizing the storage and processing of sequence information. RNN network is good at processing the proximity relationship between protein and nucleic acid sequence, such as local high-frequency codon selection, codon fitness configuration, and adverse motif avoidance. The gated recurrent unit network (GRU) is a special recurrent neural network. It is characterized by using reset gate and update gate to control information transmission, which is used to process long-term dependencies in time series. The GRU network is good at processing the long-distance relationship between protein and nucleic acid sequence, such as splice site avoidance.

[0047] The target nucleic acid sequence matrix can be the codon-optimized nucleic acid sequence matrix output by the target codon optimization model. Each amino acid in the protein sequence matrix has a corresponding codon, and the codon is obtained by arranging and combining three nucleic acids, so it can also be understood that each amino acid in the protein sequence matrix has a plurality of corresponding nucleic acid sequences. Since there are multiple amino acids in the protein sequence matrix, there will be multiple sets of nucleic acid sequences, each set of nucleic acid sequences consisting of multiple nucleic acid sequences. The target codon optimization model selects a set of nucleic acid sequences from the multiple sets of nucleic acid sequences as the target nucleic acid sequence matrix. The codon optimization result can be the output result after codon optimization. The target nucleic acid sequence matrix can be directly used as the codon optimization result, or the target nucleic acid sequence matrix can be pre-processed to obtain the codon optimization result. This is not limited.

[0048] Optionally, the number of the at least one neural network module is greater than or equal to one and less than or equal to nine; the recurrent neural network and the gated recurrent unit network in each of the neural network modules are connected in series. The connection order of the recurrent neural network and the gated recurrent unit network is not limited here. That is, the output of the recurrent neural network can be the input of the gated recurrent unit network, or the input of the recurrent neural network can be the output of the gated recurrent unit network. According to a large amount of experimental data, when the number of neural network modules in the target codon optimization model is in the interval [1, 9], the model output result can maintain a good protein expression effect. The number of neural network modules in the target codon optimization model can be any value in the interval [1, 9]. For example, FIG. 3 is a schematic diagram of the association between the model accuracy and the number of network modules provided by the present application. The number of “stacked structure units” is the number of “neural network modules”. As shown in FIG. 3, when the number of neural network modules in the target codon optimization model is in the interval [1, 9], the target codon optimization model can maintain high accuracy.

[0049] Optionally, the recurrent neural network is a unidirectional recurrent neural network or a bidirectional recurrent neural network; and the gated recurrent unit network is a unidirectional gated recurrent unit network or a bidirectional gated recurrent unit network.

[0050] The unidirectional recurrent neural network is an RNN network, and the bidirectional recurrent neural network is a Bi-RNN network. That is, the recurrent neural network can be unidirectional or bidirectional. Similarly, the unidirectional gated recurrent unit network is a GRU network, and the bidirectional gated recurrent unit network is a Bi-GRU network. That is, the gated recurrent unit network can be unidirectional or bidirectional. For example, FIG. 4 is a structural schematic diagram of a neural network module provided by the present application. As shown in FIG. 4, the neural network module includes a recurrent neural network and a gated recurrent unit network. The recurrent neural network and the gated recurrent unit network are connected in series, and the output of the recurrent neural network can be the input of the gated recurrent unit network, or the input of the recurrent neural network can be the output of the gated recurrent unit network.

[0051] Optionally, the target codon optimization model further includes at least one feedforward neural network, the at least one feedforward neural network and the at least one neural network module are connected in series, and the input of the feedforward neural network is the output of the neural network module.

[0052] The feedforward neural network (FNN) is a mathematical model based on the working principle of biological neurons, composed of multiple layers of neurons, and divided into an input layer, a hidden layer (optional), and an output layer. Each neuron performs weighted summation on the input through weights and an activation function, and outputs to the next layer. The feedforward neural network has the characteristics of simple structure, strong non-linear mapping capability, and high training reliability, and by adding a feedforward neural network in the target codon optimization model, the data analysis capability of the model can be effectively improved.

[0053] Optionally, the training process of the target codon optimization model comprises: inputting the preset sample protein sequence into an initial codon optimization model to obtain a sample codon optimization result; determining the value of the loss function according to the sample codon optimization result and the preset codon optimization label of the preset sample protein sequence; adjusting the initial codon optimization model according to the value of the loss function to obtain the target codon optimization model.

[0054] The initial codon optimization model can be an original codon optimization model without pre-training. The preset sample protein sequence can be a preset protein sample sequence. The preset sample protein sequence can contain a corresponding preset codon optimization label, which is used to evaluate the identification result of the initial codon optimization model and adjust the model according to the evaluation result. For example, the natural coding sequence can be used as the preset codon optimization label, and the matching degree of the model output sequence and the natural sequence in codon selection can be compared, and the matching degree can be used as the model accuracy. The model accuracy can reflect the effect of the model in codon optimization, and the model accuracy is positively correlated with the quality of the model in codon optimization.

[0055] For example, the sample codon optimization result can be the sequence output by the initial codon optimization model. The loss function can be an error function between the sequence output by the initial codon optimization model and the natural coding sequence. By substituting the sample codon optimization result and the preset codon optimization label into the preset formula for solving the loss function, the value of the loss function is obtained. By determining the value of the loss function, the initial codon optimization model can be adjusted based on the derivative callback method to obtain the target codon optimization model.

[0056] Optionally, adjusting the initial codon optimization model according to the value of the loss function to obtain the target codon optimization model comprises: obtaining the function value of the loss function corresponding to each preset sample protein sequence respectively, and determining the mean value of the loss function according to all function values; based on the mean value of the loss function, the model parameters of the initial codon optimization model are adjusted by gradually reducing the learning rate to obtain the target codon optimization model.

[0057] The loss function mean value can be the average of the function values of all loss functions. For example, the function value of each preset sample protein sequence corresponding loss function can be calculated respectively, and the average of the function values of all loss functions can be calculated to obtain the loss function mean value. Based on the loss function mean value, the model parameters of the initial codon optimization model are derived by using the strategy of gradually reducing the learning rate to derive the target codon optimization model.

[0058] For example, in the case of Escherichia coli, the protein sequence and the nucleic acid sequence encoding the protein in the Escherichia coli genome correspond to the input and output of the neural network, respectively. The source of the protein sequence and the nucleic acid sequence includes but is not limited to genomic sequencing data, transcriptome sequencing data, and translation sequencing data. The data set is divided into a training set and a test set in a ratio of 4:1.

[0059] The initial codon optimization model of the technical solution of the present application includes two (Bi-RNN+Bi-GRU) structure units, and two layers of feedforward neural network (FNN) are added as the whole neural network to be trained. For example, FIG. 5 is a structural schematic diagram of an initial codon optimization model provided by the present application. FIG. 6 is a structural schematic diagram of a codon optimization model in the related art provided by the present application.

[0060] In FIG. 5, the activation function of the first layer FNN is the hyperbolic tangent function (Tanh): Tanh(x) = (e x –e -x ) / (e x +e -x )

[0061] The activation function of the second layer FNN is the S-shaped function (Sigmoid): Sigmoid(x) = 1 / (1+e -x )

[0062] The matrixed protein sequence and nucleic acid sequence are respectively taken as the input and output of the neural network. Taking the nucleic acid sequence encoding the endogenous protein of Escherichia coli as the standard sequence, for the same protein sequence input, the matching degree between the output sequence of the neural network and the standard sequence can be taken as the accuracy of the neural network. The stacking number of the (Bi-RNN+Bi-GRU) structure unit is adjusted and the model is trained respectively. When the stacking number of the structure unit is 2, the accuracy of the trained neural network is the highest. In the technical solution of the present application, the neural network model shown in FIG. 5 is the optimal one, and the model accuracy is 0.578.

[0063] The codon optimization model of the related art (FIG. 6) is composed of a 128-layer recurrent neural network (RNN) and a 16-layer bidirectional recurrent neural network (Bi-RNN). The RNN and the Bi-RNN are good at processing local information, but lack the ability to process long-distance information. For example, FIG. 7 is a diagram showing the accuracy of the model of the present application and the model of the related art. The left side is the diagram showing the accuracy of the model of the present application, and the right side is the diagram showing the accuracy of the model of the related art. As shown in FIG. 7, the accuracy of the model of the present application is 0.578, and the highest accuracy reached by the neural network constructed in the related art after training is 0.527.

[0064] The neural network is constructed and trained according to the technical solution in FIG. 5 of the present application, and the trained model is set to codon optimize the fluorescent fusion proteins CD72-GFP, M-Ras-GFP, Pim1-GFP, 7E2FV-GFP, and GluA3-GFP. The wild-type and optimized coding sequences are constructed into expression vectors on Pet24a through gene synthesis and molecular cloning. After the plasmid is transformed into BL21 and cultured on a plate for 12-16 hours, a single colony is picked and cultured in LB culture solution for 12-16 hours, and then inoculated into fresh culture solution at a ratio of 1:100 to ensure consistent initial conditions. Green fluorescence detection is performed after 12-16 hours (excitation light 485 / 20 nm, emission light 528 / 20 nm). It should be noted that excitation light 485 / 20 nm means that the center wavelength of the excitation light is 485 nm, and the bandwidth is 20 nm, i.e., the wavelength range of the excitation light is 485±10 nm. Emission light 528 / 20 nm means that the center wavelength of the emission light is 528 nm, and the bandwidth is 20 nm, i.e., the wavelength range of the emission light is 528±10 nm.

[0065] For example, FIG. 8 is a diagram showing the comparison of the fluorescence intensity of the sequence output by the model of the present application and the sequence output by the model of the related art. The fluorescence intensity of the fluorescent fusion protein is positively correlated with the protein expression amount. In the verification of the expression amount of each group of proteins, the optimization effect of each codon optimization tool can be compared by the difference (Bias) between the sample fluorescence intensity and the highest fluorescence intensity (i.e., the fluorescence intensity of the sample with the highest fluorescence intensity in the verification of the expression amount of each group of proteins):

[0066] Bias=(highest fluorescence intensity-sample fluorescence intensity) / highest fluorescence intensity

[0067] For each codon optimization tool, the average value of the Bias of the five groups of proteins in FIG. 8 is calculated, and the optimization effect ranking is as follows:

[0068] The present application (Bias=0.138)>A company (Bias=0.357)>B company (Bias=0.537)>C company (Bias=0.866)

[0069] Therefore, the protein expression amount improving effect of the present application is obviously higher than that of various codon optimization technologies in the related art.

[0070] The technical scheme provided by the present application comprises the following steps: obtaining a target protein sequence; performing matrix processing on the target protein sequence to obtain a protein sequence matrix; inputting the protein sequence matrix into a target codon optimization model to obtain a target nucleic acid sequence matrix, and determining the codon optimization result according to the target nucleic acid sequence matrix. The technical scheme of the present application solves the problem of insufficient protein expression effect in the codon optimization technology in the related art, and can perform codon optimization on a protein sequence based on a model composed of a recurrent neural network and a gated recurrent unit network to improve the protein expression effect.

[0071] FIG. 9 is a structural schematic diagram of a codon optimization device provided by the present application. The present application can be applied to the scenario of performing codon optimization on a protein sequence. The device can be realized by software and / or hardware, and integrated into a computer device with application development function.

[0072] As shown in FIG. 9, the codon optimization device comprises a protein sequence acquisition module 310 and a codon optimization module 320.

[0073] The protein sequence acquisition module 310 is configured to acquire a target protein sequence. The codon optimization module 320 is configured to input the target protein sequence into a target codon optimization model that has been pre-trained to obtain a codon optimization result of the target protein sequence. The target codon optimization model comprises at least one neural network module, and each neural network module comprises a recurrent neural network and a gated recurrent unit network.

[0074] The technical scheme provided by the present application comprises the following steps: obtaining a target protein sequence; inputting the target protein sequence into a target codon optimization model that has been pre-trained to obtain a codon optimization result of the target protein sequence. The target codon optimization model comprises at least one neural network module, and each neural network module comprises a recurrent neural network and a gated recurrent unit network. The technical scheme of the present application solves the problem of insufficient protein expression effect in the codon optimization technology in the related art, and can perform codon optimization on a protein sequence based on a model composed of a recurrent neural network and a gated recurrent unit network to improve the protein expression effect.

[0075] In an optional implementation, the codon optimization module 320 is specifically configured to: perform matrix processing on the target protein sequence to obtain a protein sequence matrix; wherein the amino acids in the protein sequence matrix are represented in vector form; input the protein sequence matrix into the target codon optimization model to obtain a target nucleic acid sequence matrix, and determine the codon optimization result according to the target nucleic acid sequence matrix.

[0076] In an optional implementation, the codon optimization device further includes a codon optimization model training module configured to: input the preset sample protein sequences into an initial codon optimization model to obtain sample codon optimization results; determine the numerical value of a loss function according to the sample codon optimization results and preset codon optimization labels of the preset sample protein sequences; and adjust the initial codon optimization model according to the numerical value of the loss function to obtain the target codon optimization model.

[0077] In an optional implementation, the codon optimization model training module includes a model parameter adjustment unit configured to: respectively acquire the function values of the loss functions corresponding to each preset sample protein sequence, and determine the mean value of the loss functions according to all the function values; and based on the mean value of the loss functions, gradually decay the model parameters of the initial codon optimization model in a learning rate to obtain the target codon optimization model.

[0078] In an optional implementation, the number of the at least one neural network module is greater than or equal to one and less than or equal to nine; the recurrent neural network and the gated recurrent unit network in each of the neural network modules are connected in series.

[0079] In an optional implementation, the recurrent neural network is a unidirectional recurrent neural network or a bidirectional recurrent neural network; and the gated recurrent unit network is a unidirectional gated recurrent unit network or a bidirectional gated recurrent unit network.

[0080] In an optional implementation, the target codon optimization model further includes at least one feedforward neural network, the at least one feedforward neural network and the at least one neural network module are connected in series, and the input of the feedforward neural network is the output of the neural network module.

[0081] The codon optimization device provided in the present application can execute any codon optimization method provided in the present application, and has the corresponding function modules of the execution method.

[0082] FIG. 10 is a structural schematic diagram of a computer device provided in the present application. FIG. 10 shows a block diagram of an exemplary computer device 12 suitable for implementing embodiments of the present application. The computer device 12 shown in FIG. 10 is merely one example and should not be taken as limiting the scope of the present application. The computer device 12 can be any terminal device with computing capability, which can be configured in a codon-optimized device.

[0083] As shown in FIG. 10, the computer device 12 is in the form of a general- purpose computing device. The components of the computer device 12 can include, but are not limited to, at least one processor or processing unit 16, a system memory 28, and a bus 18 that couples various system components including the system memory 28 to the processing unit 16.

[0084] The bus 18 can be one or more of several types of bus structures including a memory bus or memory controller, a peripheral bus, a graphics bus, a processor or local bus using any of a variety of bus architectures including Industry Standard Architecture (ISA), Micro Channel Architecture (MCA), Enhanced ISA (EISA), Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus.

[0085] The computer device 12 typically includes a variety of computer system readable media. Such media can be any available media that is accessible by the computer device 12 and includes both volatile and non-volatile media, removable and non-removable media.

[0086] System memory 28 can include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache 32. Computer device 12 can further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 can be provided for reading from and writing to non-removable, non-volatile magnetic media (not shown in FIG. 10 and typically called a "hard drive"). Although not specifically shown, a magnetic disk drive can also be used for reading from and writing to a removable, non-volatile magnetic disk (e.g., a "floppy disk"), and an optical disk drive can be used for reading from or writing to a removable non-volatile optical disk (such as a CD-ROM, DVD-ROM or other optical media). In such instances, each can be connected to bus 18 by one or more data media interfaces. As will be further depicted and described below, memory can include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the application.

[0087] Program / utility 40 having a set (at least one) of program modules 42 can be stored in system memory 28 by way of example, such programs / modules include, but are not limited to, an operating system, one or more application programs, other program modules, and program data, each of or some combination of which can include implementation of a networking environment. Program modules 42 generally carry out the functions and / or methodologies of embodiments of the application as described herein.

[0088] Computer device 12 can also communicate with one or more external devices 14 such as a keyboard, a pointing device, a display 24, etc.; one or more devices that enable a user to interact with computer device 12; and / or any devices (e.g., network card, modem, etc.) that enable computer device 12 to communicate with one or more other computing devices. Such communication can occur via Input / Output (I / O) interface(s) 22. Still yet, computer device 12 can communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or the Internet) through network adapter 20. As depicted, network adapter 20 communicates with the other components of computer device 12 via bus 18. It should be appreciated that although not shown, other hardware and / or software modules could be used in conjunction with computer device 12. Such as, but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data archival storage systems, etc.

[0089] Processing unit 16 performs various function applications and data processing by running programs stored in system memory 28, such as implementing the codon optimization method provided in the present application, which includes:

[0090] obtaining a target protein sequence;

[0091] inputting the target protein sequence into a pre-trained target codon optimization model to obtain a codon optimization result of the target protein sequence;

[0092] The target codon optimization model includes at least one neural network module, and each neural network module includes a recurrent neural network and a gated recurrent unit network.

[0093] The present application provides a computer readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the codon optimization method provided in any of the present application, including:

[0094] obtaining a target protein sequence;

[0095] inputting the target protein sequence into a pre-trained target codon optimization model to obtain a codon optimization result of the target protein sequence;

[0096] The target codon optimization model comprises at least one neural network module, and each neural network module comprises a recurrent neural network and a gated recurrent unit network.

[0097] The computer storage medium of the present application can adopt any combination of one or more computer readable media. The computer readable medium can be a computer readable signal medium or a computer readable storage medium. The computer readable storage medium may, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or apparatus, or any combination thereof. More specific examples (non-exhaustive list) of the computer readable storage medium include: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, the computer readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device or apparatus.

[0098] The computer readable signal medium can include a data signal propagating in a baseband or as part of a carrier wave propagating through a transmission medium, in which the computer readable program code is carried. Such a propagating data signal can take many forms, including but not limited to electro-magnetic, optical or any suitable combination thereof. The computer readable signal medium can also be any computer readable medium that is not a computer readable storage medium and that can transmit, propagate or transport a program for use by or in connection with an instruction execution system, apparatus or device.

[0099] The program code contained on the computer readable medium can be transmitted by any suitable medium, including but not limited to: wireless, wire, optical cable, radio frequency (RF), etc., or any suitable combination thereof.

[0100] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0101] Those skilled in the art will appreciate that the modules or steps of the present application described above can be implemented with general computing devices, which can be centralized on a single computing device or distributed on a network composed of multiple computing devices, and optionally, they can be implemented with program codes executable by the computing devices, so that they can be stored in storage devices and executed by the computing devices, or they can be respectively manufactured as individual integrated circuit modules, or multiple modules or steps among them can be manufactured as a single integrated circuit module. Thus, the present application is not limited to any specific combination of hardware and software.

[0102] It is noted that the above are only optional embodiments of the present application and the principles of the applied technology. Those skilled in the art will understand that the present application is not limited to the specific embodiments herein, and various obvious changes, re-adjustments and substitutions can be made by those skilled in the art without departing from the scope of the present application. Therefore, although the present application has been described in detail through the above embodiments, the present application is not limited to the above embodiments, and can include more other equivalent embodiments without departing from the concept of the present application, and the scope of the present application is determined by the scope of the appended claims.

Claims

1. A codon optimization method comprising: Obtain target protein sequence; Inputting the target protein sequence into a pre-trained target codon optimization model to obtain a codon optimization result of the target protein sequence; The target codon optimization model includes at least one neural network module, and each neural network module includes a recurrent neural network and a gated recurrent unit network.

2. The method according to claim 1, wherein Inputting the target protein sequence into a pre-trained target codon optimization model to obtain the codon optimization result of the target protein sequence, including: Performing matrix processing on the target protein sequence to obtain a protein sequence matrix; wherein the amino acids in the protein sequence matrix are represented in the form of vectors; The protein sequence matrix is ​​input into the target codon optimization model to obtain a target nucleic acid sequence matrix, and the codon optimization result is determined according to the target nucleic acid sequence matrix.

3. The method according to claim 1, wherein The training process of the target codon optimization model includes: Inputting the preset sample protein sequence into the initial codon optimization model to obtain a sample codon optimization result; Determining a value of a loss function according to the sample codon optimization result and a preset codon optimization label of the preset sample protein sequence; The initial codon optimization model is adjusted according to the value of the loss function to obtain the target codon optimization model.

4. The method according to claim 3, wherein: Adjusting the initial codon optimization model according to the value of the loss function to obtain the target codon optimization model includes: Obtain the function value of the loss function corresponding to each preset sample protein sequence respectively, and determine the mean value of the loss function based on all function values; Based on the mean of the loss function, the initial codon is recalled by gradually decaying the learning rate. The model parameters of the model are optimized to obtain the target codon optimization model.

5. The method according to claim 1, wherein The number of the at least one neural network module is greater than or equal to one and less than or equal to nine; the recurrent neural network and the gated recurrent unit network in each of the neural network modules are connected in series.

6. The method according to claim 5, wherein: The recurrent neural network is a unidirectional recurrent neural network or a bidirectional recurrent neural network; the gated recurrent unit network is a unidirectional gated recurrent unit network or a bidirectional gated recurrent unit network.

7. The method according to claim 1, wherein The target codon optimization model also includes at least one feedforward neural network, the at least one feedforward neural network and the at least one neural network module are connected in series, and the input of the at least one feedforward neural network is the output of the at least one neural network module.

8. A codon optimization device comprising: A protein sequence acquisition module is configured to acquire the target protein sequence; a codon optimization module, configured to input the target protein sequence into a pre-trained target codon optimization model to obtain a codon optimization result of the target protein sequence; The target codon optimization model includes at least one neural network module, and each neural network module includes a recurrent neural network and a gated recurrent unit network.

9. A computer device comprising: at least one processor; a memory configured to store at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the codon optimization method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the codon optimization method according to any one of claims 1 to 7 is implemented.

Images