mRNA codon analysis system based on deep learning network
Through the mRNA codon analysis system based on deep learning network, the problem that the existing technology is difficult to understand the expression rules and variant characteristics of pathogen mRNA sequences is solved, and in-depth analysis and optimization of mRNA sequences are achieved, and strategic support is provided.
Patent Information
- Application Number
- CN202411262942.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-10
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-09-10
AI Technical Summary
It is difficult for the prior art to deeply understand and analyze the expression patterns and variant characteristics of pathogen mRNA sequences, especially subtle topological semantic correlation characteristics, which affect the strategic support of public health institutions.
Using the mRNA codon analysis system based on deep learning networks, the semantic coding layer of sequence fragments, the topological correlation feature extraction layer between fragments and the global information coding layer are obtained and optimized mRNA sequences to capture their global topological semantic correlation features. pattern and generate optimized mRNA sequences.
It has achieved in-depth understanding and analysis of mRNA sequences, captured its expression patterns and variant characteristics, generated optimized mRNA sequences, provided strategic support to public health institutions, and improved the depth and accuracy of the analysis.
Smart Images

Figure CN118782158B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of data processing and analysis, and specifically to an mRNA codon analysis system based on a deep learning network. Background Art
[0002] Slight changes in mRNA codons can give the virus new characteristics. Therefore, it is necessary to deeply understand and learn the expression rules of mRNA. With the development of deep learning and its research and application in the field of biomedicine, a new idea for the analysis of mRNA codons has been provided. Summary of the invention
[0003] This application is made in view of the above problems. One purpose of this application is to provide an mRNA codon analysis system based on a deep learning network.
[0004] The embodiment of the present application provides an mRNA codon analysis system based on a deep learning network, which includes:
[0005] The original sequence acquisition layer is used to obtain pathogen mRNA sequences;
[0006] A sequence fragment semantic coding layer, used for extracting fragment semantic features of the pathogen mRNA sequence to obtain a sequence of context pathogen mRNA sequence fragment semantic coding vectors;
[0007] The inter-fragment topological association feature extraction layer is used to construct and extract the association features between any two context pathogen mRNA sequence fragment semantic coding vectors in the sequence of context pathogen mRNA sequence fragment semantic coding vectors to obtain an inter-sequence fragment association topological feature matrix;
[0008] A global information encoding layer, used for performing global information encoding based on a graph encoding formula on the sequence of semantic encoding vectors of the context pathogen mRNA sequence fragments and the correlation topological feature matrix between the sequence fragments to obtain a sequence of semantic feature vectors of the pathogen mRNA sequence fragments under the correlation topology;
[0009] The optimization layer is used to generate an optimized mRNA sequence based on the sequence of the semantic feature vector of the pathogen mRNA sequence fragment under the association topology.
[0010] For example, according to the mRNA codon analysis system based on deep learning network in an embodiment of the present application, the sequence fragment semantic encoding layer includes:
[0011] A sequence segmentation module, used for segmenting the pathogen mRNA sequence to obtain a set of pathogen mRNA sequence fragments;
[0012] An embedding coding module, used for embedding coding each pathogen mRNA sequence fragment in the set of pathogen mRNA sequence fragments to obtain a set of pathogen mRNA sequence fragment embedding coding vectors;
[0013] The semantic association encoding module is used to embed the pathogen mRNA sequence fragments into a set of encoding vectors and input them into an inter-sequence fragment association encoder to obtain a sequence of semantic encoding vectors of the context pathogen mRNA sequence fragments.
[0014] For example, according to the mRNA codon analysis system based on deep learning network in an embodiment of the present application, the embedded coding module is used to:
[0015] The SeqVec, ProtTrans and / or UniRep models are used to embed code each pathogen mRNA sequence fragment in the set of pathogen mRNA sequence fragments to obtain a set of embedded coding vectors of the pathogen mRNA sequence fragments.
[0016] For example, according to the mRNA codon analysis system based on deep learning network in an embodiment of the present application, the semantic association encoding module is used to:
[0017] The set of pathogen mRNA sequence fragments embedded in the coding vector is input into the sequence fragment association encoder based on the Transformer layer to obtain the sequence of the context pathogen mRNA sequence fragment semantic coding vector.
[0018] For example, according to the mRNA codon analysis system based on a deep learning network in an embodiment of the present application, the inter-fragment topological association feature extraction layer includes:
[0019] A semantic association quantification module is used to calculate the semantic association between any two context pathogen mRNA sequence fragment semantic encoding vectors in the sequence of context pathogen mRNA sequence fragment semantic encoding vectors to obtain an inter-sequence fragment association topology matrix composed of multiple semantic associations;
[0020] The topological correlation feature extraction module is used to input the sequence segment correlation topological matrix composed of the multiple semantic correlation degrees into the topological feature extractor to obtain the sequence segment correlation topological feature matrix.
[0021] For example, according to the mRNA codon analysis system based on deep learning network in an embodiment of the present application, the semantic association quantification module is used to:
[0022] Calculate the positional dot product of the semantic coding vector of the previous contextual pathogen mRNA sequence fragment and the semantic coding vector of the next contextual pathogen mRNA sequence fragment in the arbitrary two contextual pathogen mRNA sequence fragment semantic coding vectors, and accumulate the vectors obtained by the positional dot product element by element to obtain the inter-fragment semantic accumulation value;
[0023] Calculate the element-by-element sum of the squares of the semantic coding vector of the previous context pathogen mRNA sequence fragment and the element-by-element sum of the squares of the semantic coding vector of the next context pathogen mRNA sequence fragment, and calculate the square root of the sum of the two numerical values obtained by the element-by-element sum of the squares, divide the semantic cumulative value between the fragments by the numerical value obtained by the square root to obtain the semantic association between the semantic coding vector of the previous context pathogen mRNA sequence fragment and the semantic coding vector of the next context pathogen mRNA sequence fragment.
[0024] For example, according to the mRNA codon analysis system based on deep learning network in an embodiment of the present application, the topological association feature extraction module is used to:
[0025] The inter-sequence segment association topological matrix composed of the multiple semantic association degrees is input into a topological feature extractor based on a hole convolutional neural network model to obtain the inter-sequence segment association topological feature matrix.
[0026] For example, according to the mRNA codon analysis system based on deep learning network in an embodiment of the present application, the global information encoding layer is used to:
[0027] The sequence of the semantic encoding vector of the context pathogen mRNA sequence fragment and the association topology feature matrix between the sequence fragments are input into the mRNA sequence global encoder based on the graph neural network model to obtain the sequence of the semantic feature vector of the pathogen mRNA sequence fragment under the association topology.
[0028] For example, according to the mRNA codon analysis system based on deep learning network in an embodiment of the present application, the optimization layer is used to:
[0029] The sequence of semantic feature vectors of pathogen mRNA sequence fragments under the association topology is input into a decoder to obtain the optimized mRNA sequence.
[0030] The embodiments of the present application also provide a method for analyzing mRNA codons based on a deep learning network, which comprises:
[0031] Obtain pathogen mRNA sequences;
[0032] Extracting fragment semantic features of the pathogen mRNA sequence to obtain a sequence of context pathogen mRNA sequence fragment semantic encoding vectors;
[0033] Constructing and extracting correlation features between any two context pathogen mRNA sequence fragment semantic coding vectors in the sequence of context pathogen mRNA sequence fragment semantic coding vectors to obtain a correlation topological feature matrix between sequence fragments;
[0034] Performing global information encoding based on graph encoding on the sequence of semantic encoding vectors of the context pathogen mRNA sequence fragments and the correlation topological feature matrix between the sequence fragments to obtain a sequence of semantic feature vectors of the pathogen mRNA sequence fragments under the correlation topology;
[0035] Based on the sequence of the semantic feature vector of the pathogen mRNA sequence fragment under the association topology, an optimized mRNA sequence is generated.
[0036] According to the deep learning network-based mRNA codon analysis system of the embodiment of the present application, it uses the powerful representation and sequence analysis capabilities of the deep learning network to deeply understand and analyze the pathogen mRNA sequence, capture the global topological semantic association feature patterns of the mRNA sequence, grasp the expression rules and variation characteristics of the mRNA sequence, and further use the decoder to intelligently generate optimized mRNA sequences to provide strategic support for public health agencies. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application are briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present application, and are not intended to limit the present application.
[0038] Figure 1 A schematic diagram of the application architecture of the mRNA codon analysis system based on the deep learning network in the embodiment of the present application is shown;
[0039] Figure 2 A schematic diagram of the structure of the mRNA codon analysis system based on the deep learning network in the embodiment of the present application is shown;
[0040] Figure 3 A schematic diagram of the structure of the sequence segment semantic encoding layer of the mRNA codon analysis system based on the deep learning network in the embodiment of the present application is shown;
[0041] Figure 4 A schematic diagram of the structure of the inter-fragment topological association feature extraction layer of the mRNA codon analysis system based on the deep learning network in the embodiment of the present application is shown;
[0042] Figure 5 A flow chart of the mRNA codon analysis method based on a deep learning network in an embodiment of the present application is shown;
[0043] Figure 6 An application scenario diagram of the mRNA codon analysis system based on a deep learning network in an embodiment of the present application is shown;
[0044] Figure 7 A schematic diagram of a storage medium according to an embodiment of the present application is shown.
[0045] Among them, 100, server; 110, processor; 120, memory; 130, input device; 140, output device; 200, terminal device; 800, mRNA codon analysis system based on deep learning network; 810, original sequence acquisition layer; 820, sequence fragment semantic encoding layer; 830, topological association feature extraction layer between fragments; 840, global information encoding layer; 850, optimization layer; 821, sequence segmentation module; 822, embedded encoding module; 823, semantic association encoding module; 831, semantic association quantification module; 832, topological association feature extraction module; D, pathogen mRNA sequence; S, server deployed with mRNA codon analysis algorithm based on deep learning network; 1000, computer readable storage medium; 1001, computer executable instructions. DETAILED DESCRIPTION
[0046] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.
[0047] The terms used in this specification are those common terms currently widely used in the art in consideration of the functions of the present application, but these terms may vary according to the intention of a person of ordinary skill in the art, precedents, or new technologies in the art. In addition, specific terms may be selected by the applicant, and in this case, their detailed meanings will be described in the detailed description of the present application. Therefore, the terms used in the specification should not be understood as simple names, but rather as a general description based on the meaning of the terms and the present application.
[0048] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.
[0049] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. At the same time, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.
[0050] Figure 1 A schematic diagram of the application architecture of an mRNA codon analysis system based on a deep learning network in an embodiment of the present application is shown, including a server 100 and a terminal device 200.
[0051] The terminal device 200 and the server 100 can be connected via the Internet to achieve mutual communication. Optionally, the above-mentioned Internet uses standard communication technology and / or protocol. The Internet is usually the Internet, but it can also be any network, including but not limited to any combination of a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), a mobile, wired or wireless network, a private network or a virtual private network. In some embodiments, the data exchanged through the network can be represented by technologies and / or formats including Hyper Text Markup Language (HTML), Extensible Markup Language (XML), etc. In addition, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), Internet Protocol Security (IPsec), etc. can be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technologies can also be used to replace or supplement the above-mentioned data communication technologies.
[0052] The server 100 can provide various network services for the terminal device 200, wherein the server 100 can be a single server, a server cluster consisting of several servers, or a cloud computing center. Specifically, the server 100 may include a processor 110 (Center Processing Unit, CPU), a memory 120, an input device 130, and an output device 140, etc. The input device 130 may include a keyboard, a mouse, a touch screen, etc., and the output device 140 may include a display device, such as a liquid crystal display (Liquid Crystal Display, LCD), a cathode ray tube (Cathode Ray Tube, CRT), etc.
[0053] The memory 120 may include a read-only memory (ROM) and a random access memory (RAM), and provides the processor 110 with program instructions and data stored in the memory 120. In an embodiment of the present application, the memory 120 may be used to store the corresponding method of the mRNA codon analysis system based on the deep learning network in the embodiment of the present application.
[0054] The processor 110 calls the program instructions stored in the memory 120, and the processor 110 is used to execute the steps of any one of the mRNA codon analysis methods based on the deep learning network in the embodiments of the present application according to the obtained program instructions.
[0055] In addition, the application architecture diagram in the embodiment of the present application is intended to more clearly illustrate the technical solution in the embodiment of the present application, and does not constitute a limitation on the technical solution provided in the embodiment of the present application. Of course, for other application architectures and business applications, the technical solution provided in the embodiment of the present application is also applicable to similar problems.
[0056] The following is a non-restrictive description of the mRNA codon analysis system based on a deep learning network provided according to at least one embodiment of the present application through several examples or embodiments. As described below, different features in these specific examples or embodiments can be combined with each other without conflicting with each other to obtain new examples or embodiments, and these new examples or embodiments also fall within the scope of protection of the present application.
[0057] In response to the above technical problems, the technical concept of this application is to use the powerful representation and sequence analysis capabilities of deep learning networks to deeply understand and analyze pathogen mRNA sequences, capture the global topological semantic association feature patterns of mRNA sequences, grasp the expression laws and variation characteristics of mRNA sequences, and further use the decoder to intelligently generate optimized mRNA sequences to provide strategic support for public health agencies.
[0058] Based on this, Figure 2The structural schematic diagram of the mRNA codon analysis system 800 based on the deep learning network in the embodiment of the present application is shown. The mRNA codon analysis system 800 based on the deep learning network includes: an original sequence acquisition layer 810, which is used to obtain the pathogen mRNA sequence; a sequence fragment semantic encoding layer 820, which is used to extract the fragment semantic features of the pathogen mRNA sequence to obtain the sequence of the context pathogen mRNA sequence fragment semantic encoding vector; an inter-fragment topological association feature extraction layer 830, which is used to construct and extract the association features between any two context pathogen mRNA sequence fragment semantic encoding vectors in the sequence of the context pathogen mRNA sequence fragment semantic encoding vector to obtain the inter-sequence fragment association topological feature matrix; a global information encoding layer 840, which is used to perform global information encoding based on the graph encoding formula on the sequence of the context pathogen mRNA sequence fragment semantic encoding vector and the inter-sequence fragment association topological feature matrix to obtain the sequence of the pathogen mRNA sequence fragment semantic feature vector under the association topology; an optimization layer 850, which is used to generate an optimized mRNA sequence based on the sequence of the pathogen mRNA sequence fragment semantic feature vector under the association topology.
[0059] Specifically, in the technical solution of the present application, first, the pathogen mRNA sequence is obtained. Wherein, the pathogen mRNA sequence is a key intermediate product of viral replication and gene expression. In an embodiment of the present application, obtaining the pathogen mRNA sequence includes steps such as RNA extraction, reverse transcription, amplification, sequencing and verification.
[0060] Since mRNA sequences are usually very long and contain a lot of information. In the technical solution of the present application, the pathogen mRNA sequence is segmented to reduce the complexity of the analysis, thereby obtaining a set of pathogen mRNA sequence fragments; and each pathogen mRNA sequence fragment in the set of pathogen mRNA sequence fragments is embedded and encoded to obtain a set of pathogen mRNA sequence fragment embedded coding vectors. Here, the original mRNA sequence is a string composed of nucleotides (A, U, C, G), and deep learning models usually require numerical input. In an embodiment of the present application, the use of SeqVec, ProtTrans and / or UniRep models to embed and encode each pathogen mRNA sequence fragment in the set of pathogen mRNA sequence fragments can convert each sequence fragment into a numerical vector for subsequent identification and analysis.
[0061] Considering that each sequence fragment of the mRNA sequence does not exist independently, there are complex nonlinear correlations between each sequence fragment and other sequence fragments. For example, in the mRNA sequence, even nucleotides that are far apart may have functional connections, such as the impact on protein structure and function. Therefore, in the technical solution of the present application, it is expected that the set of encoding vectors embedded in the pathogen mRNA sequence fragments is input into the sequence fragment association encoder based on the Transformer layer to capture the long-distance dependencies between the mRNA sequence fragments, thereby obtaining a sequence of contextual pathogen mRNA sequence fragment semantic encoding vectors.
[0062] Accordingly, if Figure 3 As shown, the sequence fragment semantic coding layer 820 includes: a sequence segmentation module 821, which is used to perform sequence segmentation on the pathogen mRNA sequence to obtain a set of pathogen mRNA sequence fragments; an embedding coding module 822, which is used to perform embedding coding on each pathogen mRNA sequence fragment in the set of pathogen mRNA sequence fragments to obtain a set of pathogen mRNA sequence fragment embedding coding vectors; and a semantic association coding module 823, which is used to input the set of pathogen mRNA sequence fragment embedding coding vectors into an inter-sequence fragment association encoder to obtain a sequence of context pathogen mRNA sequence fragment semantic coding vectors.
[0063] The embedded coding module 822 is used to: use SeqVec, ProtTrans and / or UniRep models to embed code each pathogen mRNA sequence fragment in the set of pathogen mRNA sequence fragments to obtain a set of embedded coding vectors of the pathogen mRNA sequence fragments.
[0064] Among them, the semantic association encoding module 823 is used to: input the set of pathogen mRNA sequence fragments embedded in the encoding vector into the sequence fragment association encoder based on the Transformer layer to obtain a sequence of context pathogen mRNA sequence fragment semantic encoding vectors.
[0065] Although the Transformer-layer-based sequence segment association encoder can integrate the long-distance contextual semantic association relationship of the pathogen mRNA sequence segments, the topological structure between the individual sequence segments cannot be fully characterized and expressed. For example, although the self-attention mechanism of the Transformer layer is computationally based on the global, it may tend to capture regional features with frequent interactions (subtle associations may be ignored in the process of feature expression), and is not enough to fully express the subtle topological connections between all sequence segments. However, in the technical background of the present application, subtle changes in mRNA codons can give the virus new characteristics. Therefore, in order to enable the subtle topological connections between mRNA sequence fragments to be paid attention to, in the technical solution of the present application, it is expected to use the global topological association learning ability of the graph neural network model to gain a deeper understanding of the pathogen mRNA sequence.
[0066] More specifically, the semantic association between any two context pathogen mRNA sequence fragment semantic coding vectors in the sequence of the context pathogen mRNA sequence fragment semantic coding vectors is first calculated to quantify the semantic relevance between every two context pathogen mRNA sequence fragment semantic coding vectors, thereby obtaining a sequence fragment association topological matrix composed of multiple semantic associations; then, the sequence fragment association topological matrix composed of the multiple semantic associations is input into a topological feature extractor based on a void convolutional neural network model to learn the topological association feature patterns of local mRNA sequence fragments, thereby obtaining a sequence fragment association topological feature matrix; subsequently, the sequence of the context pathogen mRNA sequence fragment semantic coding vectors and the sequence fragment association topological feature matrix are input into an mRNA sequence global encoder based on a graph neural network model to capture the global information of the entire mRNA sequence, thereby obtaining a sequence of pathogen mRNA sequence fragment semantic feature vectors under the association topology.
[0067] Accordingly, if Figure 4 As shown, the inter-fragment topological association feature extraction layer 830 includes: a semantic association quantization module 831, which is used to calculate the semantic association between any two context pathogen mRNA sequence fragment semantic coding vectors in the sequence of context pathogen mRNA sequence fragment semantic coding vectors to obtain an inter-sequence fragment association topological matrix composed of multiple semantic associations; a topological association feature extraction module 832, which is used to input the inter-sequence fragment association topological matrix composed of the multiple semantic associations into a topological feature extractor to obtain the inter-sequence fragment association topological feature matrix.
[0068] Among them, the semantic association quantification module 831 is used to: calculate the positional dot product of the semantic coding vector of the previous context pathogen mRNA sequence fragment and the semantic coding vector of the next context pathogen mRNA sequence fragment in the semantic coding vectors of any two context pathogen mRNA sequence fragments, and accumulate the vectors obtained by the positional dot product element by element to obtain the inter-fragment semantic accumulation value; calculate the element-by-element sum of the squares of the semantic coding vector of the previous context pathogen mRNA sequence fragment and the element-by-element sum of the squares of the semantic coding vector of the next context pathogen mRNA sequence fragment, and calculate the square root of the sum of the two numerical values obtained by the element-by-element sum of the squares, divide the inter-fragment semantic accumulation value by the numerical value obtained by the square root to obtain the semantic association between the semantic coding vector of the previous context pathogen mRNA sequence fragment and the semantic coding vector of the next context pathogen mRNA sequence fragment.
[0069] In a specific example, the semantic association quantification module 831 is used to calculate the semantic association between any two context pathogen mRNA sequence fragment semantic encoding vectors in the sequence of context pathogen mRNA sequence fragment semantic encoding vectors using the following semantic association formula to obtain the inter-sequence fragment association topology matrix composed of multiple semantic associations; wherein the semantic association formula is:
[0070]
[0071] in, and are respectively the first semantic encoding vector of the previous context pathogen mRNA sequence fragment and the second semantic encoding vector of the next context pathogen mRNA sequence fragment in any two context pathogen mRNA sequence fragment semantic encoding vectors. eigenvalues, The value of is the length of the semantic encoding vector of each context pathogen mRNA sequence fragment, represents dot product, Represents the semantic association between the semantic coding vector of the previous context pathogen mRNA sequence fragment and the semantic coding vector of the next context pathogen mRNA sequence fragment.
[0072] Among them, the topological association feature extraction module 832 is used to: input the inter-sequence segment association topological matrix composed of the multiple semantic association degrees into the topological feature extractor based on the void convolutional neural network model to obtain the inter-sequence segment association topological feature matrix.
[0073] Correspondingly, the global information encoding layer 840 is used to: input the sequence of the context pathogen mRNA sequence fragment semantic encoding vector and the association topology feature matrix between the sequence fragments into the mRNA sequence global encoder based on the graph neural network model to obtain the sequence of the pathogen mRNA sequence fragment semantic feature vector under the association topology.
[0074] Furthermore, the sequence of semantic feature vectors of pathogen mRNA sequence fragments under the association topology is input into a decoder to obtain an optimized mRNA sequence. Accordingly, the optimization layer 850 is used to: input the sequence of semantic feature vectors of pathogen mRNA sequence fragments under the association topology into a decoder to obtain the optimized mRNA sequence.
[0075] In a preferred embodiment, the optimization layer 850 includes: a vector optimization module, used to optimize the sequence of semantic feature vectors of pathogen mRNA sequence fragments under the association topology to obtain an optimized sequence of semantic feature vectors of pathogen mRNA sequence fragments under the association topology; a decoding module, used to input the sequence of semantic feature vectors of pathogen mRNA sequence fragments under the optimized association topology into the decoder to obtain the optimized mRNA sequence.
[0076] In particular, considering that the sequence of the contextual pathogen mRNA sequence fragment semantic coding vector and the topological feature matrix of the association between the sequence fragments respectively express the local embedded coding semantic space domain contextual association coding semantic features and semantic association degree topological association features of each pathogen mRNA sequence fragment, thus, after the sequence of the contextual pathogen mRNA sequence fragment semantic coding vector and the topological feature matrix of the association between the sequence fragments are input into the mRNA sequence global encoder based on the graph neural network model, the sequence of the pathogen mRNA sequence fragment semantic feature vector under the association topology obtained will also be due to the relative independence of the coding semantic distribution of the pathogen mRNA sequence fragment semantic feature vector under each association topology under the topological association representation, and the difference in coding semantic distribution will lead to the lack of interpretability of decoding regression attributed to complex semantic fusion, thereby affecting the accuracy of the decoding result. Based on this, the present application considers optimizing the sequence of the semantic feature vector of the pathogen mRNA sequence fragment under the association topology.
[0077] In a preferred embodiment, the sequence of the semantic feature vector of the pathogen mRNA sequence fragment under the association topology is optimized to obtain the sequence of the semantic feature vector of the pathogen mRNA sequence fragment under the optimized association topology, comprising the steps of:
[0078] Cascading the sequences of the semantic feature vectors of the pathogen mRNA sequence fragments under the association topology into a semantic cascade feature vector of the pathogen mRNA sequence fragments under the association topology;
[0079] The self-association matrix of the semantic cascade feature vector of the pathogen mRNA sequence fragment under the association topology is calculated, that is:
[0080]
[0081] in, represents the semantic cascade feature vector of pathogen mRNA sequence fragments under the association topology, represents the transpose of a vector, represents matrix multiplication, represents the autocorrelation matrix;
[0082] The semantic concatenation feature vector of the pathogen mRNA sequence fragment under the association topology is multiplied by the self-association matrix in a query manner to obtain an intermediate feature vector, that is:
[0083]
[0084] in, represents the semantic cascade feature vector of pathogen mRNA sequence fragments under the association topology, represents the autocorrelation matrix, represents matrix multiplication, represents the intermediate eigenvector;
[0085] Calculate the first Row vector and Inner product sum of row vectors distance, to obtain the first weight matrix and the second weight matrix , expressed as:
[0086]
[0087]
[0088] in, and The autocorrelation matrix is represented by Row vector and row vector, represents the transpose of a vector, represents matrix multiplication, represents the autocorrelation matrix, represents the first weight matrix, represents the first weight matrix The eigenvalues of the positions, represents the second weight matrix, represents the second weight matrix The eigenvalues of the positions, Represents the calculation vector distance;
[0089] The first weight matrix and the second weight matrix After matrix multiplication, further matrix multiplication is performed with the intermediate feature vector to obtain the semantic cascade feature vector of the pathogen mRNA sequence fragment under the optimized association topology, that is:
[0090]
[0091] in, represents the first weight matrix, represents the second weight matrix, represents the intermediate eigenvector, represents matrix multiplication, Representing the semantic cascade feature vector of pathogen mRNA sequence fragments under the optimized association topology;
[0092] The semantic cascade feature vectors of pathogen mRNA sequence fragments under the optimized association topology are split into sequences of semantic feature vectors of pathogen mRNA sequence fragments under the optimized association topology.
[0093] That is, for the eigenvalues of the sequence of semantic feature vectors of pathogen mRNA sequence fragments under the association topology, the associated response space is scattered bit by bit, and the deep embedding structure in the high-order association space is obtained through the joint embedding space of the vector inner product embedding space and the vector independent embedding space, and the decomposable associated response sub-dimensions of the eigenvalues of the sequence of semantic feature vectors of pathogen mRNA sequence fragments under the association topology are represented as a set of spatial sub-dimensions to further perform dependent space mapping of the decomposable associated response sub-dimensions of the deep embedding structure of the sequence of semantic feature vectors of pathogen mRNA sequence fragments under the association topology, so as to improve the interpretability of the solution space of the decoding regression solution of the sequence of semantic feature vectors of pathogen mRNA sequence fragments under the association topology based on the association distribution through the complex dimensional representation of the associated response, thereby improving the accuracy of the optimized mRNA sequence obtained by inputting the sequence of semantic feature vectors of pathogen mRNA sequence fragments under the association topology into the decoder.
[0094] Based on the above embodiments, see Figure 5 As shown in FIG. 1 , a flow chart of a method for analyzing mRNA codons based on a deep learning network in an embodiment of the present application is shown. For example, the method for analyzing mRNA codons based on a deep learning network can be executed by a server, which can be Figure 1 The server 100 shown in FIG. Figure 5As shown, according to the mRNA codon analysis method based on deep learning network according to the embodiment of the present application, the steps include: S510, obtaining the pathogen mRNA sequence; S520, performing fragment semantic feature extraction on the pathogen mRNA sequence to obtain the sequence of context pathogen mRNA sequence fragment semantic coding vectors; S530, constructing and extracting the association features between any two context pathogen mRNA sequence fragment semantic coding vectors in the sequence of context pathogen mRNA sequence fragment semantic coding vectors to obtain the inter-sequence fragment association topological feature matrix; S540, performing graph-based global information encoding on the sequence of context pathogen mRNA sequence fragment semantic coding vectors and the inter-sequence fragment association topological feature matrix to obtain the sequence of pathogen mRNA sequence fragment semantic feature vectors under the association topology; S550, generating an optimized mRNA sequence based on the sequence of pathogen mRNA sequence fragment semantic feature vectors under the association topology.
[0095] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned mRNA codon analysis method based on deep learning network have been described in the above reference. Figures 2 to 4 The invention has been described in detail in the description of the deep learning network-based mRNA codon analysis system 800, and therefore, its repeated description will be omitted.
[0096] Furthermore, in this application, an mRNA sequence-protein expression related database is established, specifically including: (1) Data collection objectives: A Python crawler is written through the database website to collect mRNA sequences and their corresponding protein expression related databases following the principle of representational state transfer through the application programming interface (RESTful API) of the database website, and the transcript sequences and corresponding proteins of the human genome are collected from NCBI RefSeq, the coding regions of each sequence of the human genome are collected from NCBI CCDS Database, the corresponding abundance of various proteins in the human proteome are collected from PaxDB, and databases such as GEO and GTEx collect expression data based on RNA-seq, and databases such as RBFdb collect ribosome map analysis related data. The literature retrieval method is used to collect data from relevant literature reports not included in the above databases. At the same time, the database built by the Vaccine Evaluation Center of the Centers for Disease Control and Prevention is used to collect data such as the spatial structure, sequence, and effectiveness analysis of relevant vaccines. (2) Data preprocessing: Based on the protein alias-number comparison table, Python's Pandas package was used to associate mRNA sequences with protein expression data, and an automated data checking tool was written to perform formatting checks. Sequence-protein expression data tables were classified and established, mainly including multiple IDs, mRNA full sequences, CDS sequences, and data related to protein expression such as protein abundance and gene expression values. Missing data were manually searched, classified, saved, or discarded as appropriate. (3) Database construction: The above data were managed using the PostgreSQL database system, indexed, and divided into libraries and tables; Python-based add, delete, modify, and query APIs were written to facilitate application calls. (4) Codon index-related literature reports and program implementation: Through literature retrieval, reports related to the codon index were collected. According to their different definitions, the literature was sorted by category, and their program implementation was sought: a. For reports with attached program implementation, they were collected, integrated, and tested based on their licenses; b. For literature that did not provide program implementation, the codon index calculation program was implemented by itself based on its definition formula. From the numerous codon indices reported, a clearly defined and programmable codon index is established and integrated into a computational tool for subsequent research.
[0097] Furthermore, an mRNA codon optimization model was constructed.
[0098] Among them, regarding the feature engineering based on codon index, the distribution of various codon indexes in the above-collected data sets is examined, and the data with very high inclination is logarithmized and standardized. By calculating the Pearson and Spearman correlation coefficients between the codon indexes, the codon indexes that are close and far away are screened, and the ones with correlation coefficients close to -1 are negated. Due to the large number of codon indexes and the limited training data sets, the manifold learning method is used to extract features and compress feature vectors based on the correlation of codon indexes to avoid the dimensionality disaster during training.
[0099] Among them, manifold learning is mainly based on the assumption that high-dimensional data is embedded into high dimensions by low-dimensional manifolds, and high-dimensional data is mapped back to low dimensions to achieve dimensionality reduction. The main methods include isometric feature mapping (Isomap), Locally Linear Embedding (LLE), Hessian Eigenmapping, Spectral Embedding, Local Tangent Space Alignment (LTSA), t-distributed Stochastic Neighbor Embedding (t-SNE), etc. The above methods are used to reduce the dimensionality of codon index features respectively, and combined with hierarchical clustering, principal component analysis and other methods to comprehensively consider their classification results, and establish a new sequence feature vector based on codon index.
[0100] Regarding the Monte Carlo tree search-based reinforcement learning model, reinforcement learning is a learning paradigm based on the interaction between the learning agent (agent) and the environment through continuous trial and error. The learning agent takes corresponding actions in a certain state and receives feedback from the environment. The feedback includes entering a new state and obtaining corresponding rewards. The ultimate goal is to learn a strategy, that is, what actions to take in what state, in order to achieve the goal of maximizing long-term cumulative rewards. This application adopts the Q-learning method in reinforcement learning, and learns a strategy that maximizes the overall reward by iteratively updating the Q function:
[0101]
[0102] in is in state Next select action The expected reward, Is to perform an action The short-term immediate reward immediately after is the discount factor, For the next state The maximum expected reward for all possible actions, is the learning rate.
[0103] In this application, the model will traverse the sequence amino acids one by one from the beginning and select the optimal codon for each amino acid, mainly including the following steps:
[0104] a. Take the sequence that has been currently codon selected as the current state , in this state, the next action , that is, selecting the corresponding codon for the next amino acid and checking the current state and all possible actions for the next step.
[0105] b. For each current state and the corresponding next action selection, evaluate its total score reward through the Q network In this step, the current state is encoded using a Transformer encoder model, concatenated with the action matrix into a corresponding vector, and then outputted as a reward score after passing through a multi-layer perceptron (MLP).
[0106] This deep neural network uses mean square error loss:
[0107]
[0108] in Indicates that the current parameter is Network status Execute an action The prediction score reward is The final training target parameter is Under the network, the status Execute an action The maximum prediction score reward is obtained. The short-term immediate reward obtained by the action is evaluated by a decreasing function of the difference between the feature vector of the current sequence based on the codon index and the set mean of the feature vector of the highly expressed sequence. , is the expected value of the expression, which is the average value of all prediction samples.
[0109] c. The Monte Carlo Tree Search method is used to explore the next action. This method will explore the possible actions for the next step and the subsequent steps after this step, evaluate the node selection through the Upper Confidence Bounds (UCT) method, balance the search breadth and depth, and return the best action choice :
[0110]
[0111] in Status The set of all possible action choices, Status Next Action Rating rewards, is the state in the Monte Carlo random search process The number of visits, For the status Next action The number of times sampled, constant Used to control the balance between exploration and exploitation.
[0112] For each amino acid in the sequence, repeat the above steps until the codon selection corresponding to all amino acids is completed.
[0113] Regarding the performance evaluation and optimization upgrade of the model, 10% of the data in the dataset is first randomly selected as the test set, and 5-fold cross validation is used in the remaining training set, with the training set and validation set divided at a ratio of 4:1. The validation set is used for rapid parameter adjustment and hyperparameter selection, and the optimal hyperparameters are determined and tested in the test set. The final result is the average of the cross-validation. Common evaluation indicators for NLP tasks are selected to evaluate the test results: Jaccard distance, Levenshtein distance, bilingual evaluation alternative (BLEU), recall-oriented summary evaluation (ROUGE), etc.
[0114] Jaccard distance is an indicator for measuring the difference between two sets. In this application, it is used to measure the difference in feature vectors between the output sequence and the high protein expression sequence.
[0115]
[0116] Levenshtein distance, also known as edit distance, is an indicator to measure the difference between two strings, and is the minimum number of operations to convert one string into another. In this application, it is defined as the minimum number of operations between the sequence of the same protein after the model-optimized codon usage is converted into the sequence of the actual human genome.
[0117]
[0118] BLEU (Bilingual Evaluation Understudy) is a metric for evaluating the quality of machine translation, which works by comparing the similarity between the output of machine translation and one or more reference translations. In this application, it is used to compare the similarity between the optimized sequence and the high protein expression sequence.
[0119]
[0120] ROUGE (Recall-Oriented Understudy for Gisting Evaluation) is a method for comparing the similarity between generated text and reference text. One of the most commonly used ROUGE metrics is ROUGE-L based on the longest common subsequence (LCS). In this application, it is used to compare the common length part of the sequence corresponding to the same protein optimized sequence and the actual human genome sequence.
[0121]
[0122]
[0123]
[0124] Further, Figure 6 FIG. 1 is an application scenario diagram of an mRNA codon analysis system based on a deep learning network according to an embodiment of the present application. Figure 6 As shown, in this application scenario, first, the pathogen mRNA sequence is obtained (for example, Figure 6 D), then, the pathogen mRNA sequence is input into a server (e.g., Figure 6 In S) as shown in , the server is capable of using the mRNA codon analysis algorithm based on the deep learning network to process the pathogen mRNA sequence to obtain an optimized mRNA sequence.
[0125] Based on the above embodiments, another exemplary embodiment of an electronic device is also provided in the embodiments of the present application. In some possible embodiments, the electronic device in the embodiments of the present application may include a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor can implement the steps of the mRNA codon analysis method based on the deep learning network in the above embodiments when executing the program.
[0126] For example, in the case of electronic equipment Figure 1 Taking the server 100 in the example as an example, the processor in the electronic device is the processor 110 in the server 100, and the memory in the electronic device is the memory 120 in the server 100.
[0127] An embodiment of the present application also provides a computer-readable storage medium. Figure 7 1 shows a schematic diagram of a computer-readable storage medium 1000 according to an embodiment of the present application. Figure 7As shown, the computer-readable storage medium 1000 stores computer-executable instructions 1001. When the computer-executable instructions 1001 are executed by the processor, the mRNA codon analysis method based on the deep learning network according to the embodiment of the present application described with reference to the above figures can be executed. The computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may, for example, include a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may, for example, include a read-only memory (ROM), a hard disk, a flash memory, etc.
[0128] The embodiment of the present application also provides a computer program product or computer program, which includes computer executable instructions, which are stored in a computer-readable storage medium. The processor of the computer device reads the computer executable instructions from the computer-readable storage medium, and the processor executes the computer executable instructions, so that the computer device performs the mRNA codon analysis method based on the deep learning network according to the embodiment of the present application.
[0129] Those skilled in the art will appreciate that the contents disclosed in this application may be subject to various variations and improvements. For example, the various devices or components described above may be implemented by hardware, or by software, firmware, or a combination of some or all of the three.
[0130] In addition, although the present application makes various references to certain units in the system according to embodiments of the present application, any number of different units can be used and run on the client and / or server. The units are only illustrative, and different aspects of the system and method can use different units.
[0131] Those skilled in the art will appreciate that all or part of the steps in the above method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a disk or an optical disk. Optionally, all or part of the steps in the above embodiment can also be implemented using one or more integrated circuits. Accordingly, each module / unit in the above embodiment can be implemented in the form of hardware or in the form of software function modules. The present application is not limited to any particular form of combination of hardware and software.
[0132] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which this application belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an idealized or extremely formal sense, unless explicitly defined as such herein.
[0133] The above is an explanation of the present application and should not be considered as a limitation thereof. Although several exemplary embodiments of the present application have been described, it will be readily understood by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present application. Therefore, all of these modifications are intended to be included within the scope of the present application as defined by the claims. It should be understood that the above is an explanation of the present application and should not be considered to be limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present application is defined by the claims and their equivalents.
Claims
1. An mRNA codon analysis system based on a deep learning network, characterized in that: include: The original sequence acquisition layer is used to obtain pathogen mRNA sequences; A sequence fragment semantic coding layer, used for extracting fragment semantic features of the pathogen mRNA sequence to obtain a sequence of context pathogen mRNA sequence fragment semantic coding vectors; The inter-fragment topological association feature extraction layer is used to construct and extract the association features between any two context pathogen mRNA sequence fragment semantic coding vectors in the sequence of context pathogen mRNA sequence fragment semantic coding vectors to obtain an inter-sequence fragment association topological feature matrix; A global information encoding layer, used for performing global information encoding based on a graph encoding formula on the sequence of semantic encoding vectors of the context pathogen mRNA sequence fragments and the correlation topological feature matrix between the sequence fragments to obtain a sequence of semantic feature vectors of the pathogen mRNA sequence fragments under the correlation topology; The optimization layer is used to generate an optimized mRNA sequence based on the sequence of the semantic feature vector of the pathogen mRNA sequence fragment under the association topology.
2. The mRNA codon analysis system based on deep learning network according to claim 1, characterized in that: The sequence segment semantic encoding layer includes: A sequence segmentation module, used for segmenting the pathogen mRNA sequence to obtain a set of pathogen mRNA sequence fragments; An embedding coding module, used for embedding coding each pathogen mRNA sequence fragment in the set of pathogen mRNA sequence fragments to obtain a set of pathogen mRNA sequence fragment embedding coding vectors; The semantic association encoding module is used to embed the pathogen mRNA sequence fragments into a set of encoding vectors and input them into an inter-sequence fragment association encoder to obtain a sequence of semantic encoding vectors of the context pathogen mRNA sequence fragments.
3. The mRNA codon analysis system based on deep learning network according to claim 2, characterized in that: The embedded coding module is used for: The SeqVec, ProtTrans and / or UniRep models are used to embed code each pathogen mRNA sequence fragment in the set of pathogen mRNA sequence fragments to obtain a set of embedded coding vectors of the pathogen mRNA sequence fragments.
4. The mRNA codon analysis system based on deep learning network according to claim 3, characterized in that: The semantic association encoding module is used to: The set of pathogen mRNA sequence fragments embedded in the coding vector is input into the sequence fragment association encoder based on the Transformer layer to obtain the sequence of the context pathogen mRNA sequence fragment semantic coding vector.
5. The mRNA codon analysis system based on deep learning network according to claim 4, characterized in that: The inter-fragment topological association feature extraction layer includes: A semantic association quantification module is used to calculate the semantic association between any two context pathogen mRNA sequence fragment semantic encoding vectors in the sequence of context pathogen mRNA sequence fragment semantic encoding vectors to obtain an inter-sequence fragment association topology matrix composed of multiple semantic associations; The topological correlation feature extraction module is used to input the sequence segment correlation topological matrix composed of the multiple semantic correlation degrees into the topological feature extractor to obtain the sequence segment correlation topological feature matrix.
6. The mRNA codon analysis system based on deep learning network according to claim 5, characterized in that: The semantic association quantification module is used to: Calculate the positional dot product of the semantic coding vector of the previous contextual pathogen mRNA sequence fragment and the semantic coding vector of the next contextual pathogen mRNA sequence fragment in the arbitrary two contextual pathogen mRNA sequence fragment semantic coding vectors, and accumulate the vectors obtained by the positional dot product element by element to obtain the inter-fragment semantic accumulation value; Calculate the element-by-element sum of the squares of the semantic coding vector of the previous context pathogen mRNA sequence fragment and the element-by-element sum of the squares of the semantic coding vector of the next context pathogen mRNA sequence fragment, and calculate the square root of the sum of the two numerical values obtained by the element-by-element sum of the squares, divide the semantic cumulative value between the fragments by the numerical value obtained by the square root to obtain the semantic association between the semantic coding vector of the previous context pathogen mRNA sequence fragment and the semantic coding vector of the next context pathogen mRNA sequence fragment.
7. The mRNA codon analysis system based on deep learning network according to claim 6, characterized in that: The topological association feature extraction module is used to: The inter-sequence segment association topological matrix composed of the multiple semantic association degrees is input into a topological feature extractor based on a hole convolutional neural network model to obtain the inter-sequence segment association topological feature matrix.
8. The mRNA codon analysis system based on deep learning network according to claim 7, characterized in that: The global information encoding layer is used to: The sequence of the semantic encoding vector of the context pathogen mRNA sequence fragment and the association topology feature matrix between the sequence fragments are input into the mRNA sequence global encoder based on the graph neural network model to obtain the sequence of the semantic feature vector of the pathogen mRNA sequence fragment under the association topology.
9. The mRNA codon analysis system based on deep learning network according to claim 8, characterized in that: The optimization layer is used to: The sequence of semantic feature vectors of pathogen mRNA sequence fragments under the association topology is input into a decoder to obtain the optimized mRNA sequence.
10. A method for mRNA codon analysis based on a deep learning network, characterized in that: include: Obtain pathogen mRNA sequences; Extracting fragment semantic features of the pathogen mRNA sequence to obtain a sequence of context pathogen mRNA sequence fragment semantic encoding vectors; Constructing and extracting correlation features between any two context pathogen mRNA sequence fragment semantic coding vectors in the sequence of context pathogen mRNA sequence fragment semantic coding vectors to obtain a correlation topological feature matrix between sequence fragments; Performing global information encoding based on graph encoding on the sequence of semantic encoding vectors of the context pathogen mRNA sequence fragments and the correlation topological feature matrix between the sequence fragments to obtain a sequence of semantic feature vectors of the pathogen mRNA sequence fragments under the correlation topology; Based on the sequence of the semantic feature vector of the pathogen mRNA sequence fragment under the association topology, an optimized mRNA sequence is generated.
