A method for designing mRNA vaccine sequences based on sequence and structure information

Abstract

The present invention relates to the field of medical molecular biology technology, and in particular to an mRNA vaccine sequence design method based on sequence and structure information. The method predicts important epitope sites of a target protein through an artificial intelligence prediction model in combination with an immune epitope database, a human proteome, and features such as residue flexibility, sequence conservation, and immunogenicity. Then, the immune source sequences with a wide range of allele coverage are screened out through the MHC type coverage of the Chinese population. The method can be widely applied to RNA viruses, realize intelligent planning and automatic execution of commonly used epitope analysis schemes, and provide efficient methods and technical support for various basic and clinical combined research.

Description

Technical Field

[0001] The present invention relates to the field of medical molecular biology technology, and in particular to a method for designing mRNA vaccine sequences based on sequence and structure information. Background Art

[0002] Infectious diseases such as COVID-19, monkeypox, and seasonal influenza are all caused by RNA viruses. In the field of infectious diseases, the development of high-quality vaccines is a long way to go.

[0003] Mechanism of action of mRNA vaccine: Using lipid nanoparticles, the mRNA sequence encoding the spike protein (S protein) of the new coronavirus is encapsulated in liposomes. After the mRNA new coronavirus vaccine enters the human body, it can use its own cells to translate the mRNA fragments it carries into antigenic proteins on the surface of the new coronavirus. Although this antigen protein is produced by its own cells, because its sequence is exogenous, it will still stimulate B cells and T cells to produce specific immune responses and establish memory cells, which can prevent infectious diseases. Compared with traditional vaccines, mRNA vaccines have excellent immune activation capabilities. By endogenously expressing antigen proteins, they can induce more extensive and effective cellular and humoral immune responses, resulting in a higher protection rate. At the same time, mRNA vaccines have a short preparation cycle, controllable costs, and high protection efficiency, and have the potential to quickly respond to sudden outbreaks of infectious diseases such as the new coronavirus.

[0004] At present, my country started late in the field of nucleic acid vaccines represented by mRNA vaccines. There is a big gap between China and the world's leading mRNA vaccine companies (such as BioNTech and Moderna) in key aspects such as sequence design, delivery system and expansion of production. Currently, no mature mRNA products have been launched. How to integrate DNA genome data, existing protein structure data and structure prediction data design to assist in the design of mRNA vaccine sequences for the corresponding viruses is of great significance. Summary of the invention

[0005] In view of the deficiencies of the above-mentioned background technology, the present invention provides an mRNA vaccine sequence design method based on sequence and structure information.

[0006] The technical solution adopted by the present invention is as follows: a method for designing mRNA vaccine sequences based on sequence and structure information, which mainly comprises the following steps:

[0007] S1. Construct the protein structure of the target protein;

[0008] S2. Perform structural feature analysis, sequence conservation analysis, peptide feature analysis and peptide affinity analysis on the target protein to obtain corresponding feature data;

[0009] S3. Input the structural features, sequence conservation features, peptide features and peptide affinity features into the XGBoost learning model, and build an epitope prediction model based on the weak classification algorithm to screen important epitope sites of the target protein;

[0010] S4. The amino acid sequences and DNA sequences corresponding to the important epitope sites screened out are used as candidate antigen sequences, and the best immunogenic sequences are screened according to their affinity with MHC.

[0011] Preferably, the structural feature analysis in S2 includes: amino acid residue solvent analysis, surface area analysis and flexibility analysis.

[0012] Preferably, the peptide feature analysis in S1 includes: peptide toxicity analysis, induced interferon-γ analysis, interleukin-4 analysis and interleukin-10 analysis.

[0013] Preferably, in S3, the objective function is used as the training target of the epitope prediction model to achieve the minimum objective function, the base classification algorithm is used as the generation strategy of the epitope prediction model, the minimum objective function value is selected as the root node, and so on, until the tree building is completed, and the epitope prediction model is created.

[0014] Preferably, in S3, Formula I is used to train the epitope prediction model to achieve the minimum objective function

[0015]

[0016] Among them, l(*) is the loss function, which is used to express the prediction score and the true score y i The gap between is the i-th sample x i The predicted value at the tth iteration is the predicted value at the t-1th iteration plus the predicted value f at the tth iteration. t ; Ω(*) is a regular term, which is used to prevent overfitting caused by the tree being too complex. It combines the first-order and second-order derivatives of the loss function to make the objective function converge faster.

[0017] Preferably, in S3, Formula II is used as the generation strategy of the epitope prediction model, the minimum objective function value is selected as the root node, and so on until the tree is built to create the epitope prediction model

[0018]

[0019] Among them, D k The characteristic value of each sample representing the kth feature (x nk ) and the corresponding h i The collection composed of h represents all samples of the kth feature i sum, h represents all samples whose eigenvalues ​​are less than z i Sum; weighted quantile method is used in the process of threshold division of features, with the second-order derivative value h i The sample is divided according to its weight, and the node with the minimum objective function is found as the best splitting point.

[0020] Preferably, sequence conservation analysis is performed using Formula III for sequence comparison analysis:

[0021] cons(i)=1+∑ k p k (i)logp k (i) / log20 Formula III

[0022] p k (i) represents the probability of the kth observed amino acid in the entire sequence.

[0023] Preferably, the affinity between the peptides and HLA is analyzed using NetMHCpan4.0 software.

[0024] Preferably, the screening criteria in S5 are: based on the affinity between the candidate antigen sequence and MHC, the antigen sequence with the largest coverage ratio is selected as the best immunogenic sequence.

[0025] Beneficial effects: Compared with the prior art, the mRNA vaccine sequence design method provided by the present invention, which is based on sequence and structure information, predicts the important epitope sites of the target protein through an artificial intelligence prediction model, combined with the immune epitope database, the human proteome, and the characteristics of residue flexibility, sequence conservation, and immunogenicity, and then screens out the immune source sequences with a wide range of allele coverage through the MHC type coverage of the Chinese population; and this method can be widely applied to RNA viruses, realize the intelligent planning and automatic execution of commonly used epitope analysis schemes, and provide efficient methods and technical support for various basic and clinical research. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a flow chart of the present invention;

[0027] Figure 2 This is the three-dimensional structure of the spike protein;

[0028] Figure 3 is the amino acid residue characteristic analysis diagram;

[0029] Figure 4 Schematic diagram of predicted epitopes for the spike protein. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be interpreted as being limited to the embodiments described herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0031] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0032] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0033] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0034] It is understandable that before using the technical solutions disclosed in the embodiments of the present disclosure, the types, scope of use, usage scenarios, etc. of the personal information involved in the present disclosure should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with relevant laws and regulations.

[0035] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the invention belongs. Although any methods and materials similar or equivalent to the methods and materials described herein can be used to practice or detect the present invention, preferred methods and materials are described herein. Generally, terms related to cell and molecular biology and chemistry are used that are well known and commonly used in the art. Certain experimental techniques that are not specifically limited are generally performed according to conventional methods that are already known in the art and are described in a variety of comprehensive and more specific references cited and discussed in this specification.

[0036] Example 1 Predicting the complete structure of the novel coronavirus spike protein

[0037] The target protein structure can be a protein tertiary structure designed by technicians based on their business function requirements using existing tools, using the homologous structures of each part and the protein skeleton coordinates as the target protein structure. The spike protein of the new coronavirus is a trimer structure. Currently, the spike protein has not yet solved its complete crystal structure. Using the homologous structures of each part, HHpred and modeller and other tools to construct a complete target protein structure (see Figure 2 ); In addition, TMHMM Server v.2.0 was used to predict the transmembrane region of the spike protein, see Figure 2 The red α-helical part is marked in the box.

[0038] Example 2 Structural feature analysis based on target protein structure

[0039] When proteins interact, the residues on the surface of the proteins generally participate in the contact surface. In addition, the side chains of the contacting residues may also undergo conformational changes. Therefore, it is necessary to analyze the solvent accessible surface area and flexibility of the amino acid residues based on the structure. This method is mainly based on the analysis of the characteristics of amino acid residues by tools such as GETAREA and MEDUSA. Some of the results are as follows: Figure 3 shown.

[0040] Example 3 Sequence Conservation Analysis

[0041] The spike protein sequences of the new coronavirus were collected from databases such as GSAID and NCBI, and 3,111,691 Omicron branch sequences, 1,157,956 Alpha branch sequences, 41,847 Beta branch sequences, and 122,402 Gamma branch sequences were obtained. Formula III was used to analyze the conservation of the sequences.

[0042] cons(i)=1+∑kpk(i)logpk(i) / log 20 Formula III

[0043] Pk(i) represents the probability of the kth observed amino acid in the entire sequence

[0044] Example 4 Peptide feature analysis

[0045] Most peptides are harmless to the human body, but there are many toxic peptides in nature, such as most snake venoms or tetrodotoxin. When synthesizing peptide vaccines for immunotherapy, since the preparation is also a peptide, it is necessary to determine whether the synthesized peptide is toxic before it is injected into the human body to avoid causing serious medical accidents.

[0046] Interferon-γ (IFN-γ) is a soluble dimeric cytokine that is secreted by CD4Th1 and CD8 cytotoxic T cells during antigen-specific immunity. In addition to directly inhibiting viral replication, the importance of IFN-γ to the immune system is also reflected in its immunostimulatory and immunoregulatory functions.

[0047] IL4 stimulates the proliferation of activated B cells and T cells, as well as the differentiation of B cells into plasma cells. It is a key regulator of humoral and adaptive immunity.

[0048] IL-10 inhibits the expression of many proinflammatory cytokines, chemokines, and chemokine receptors and mediates allergen tolerance in allergen-specific immunotherapy.

[0049] The toxicity, induction of interferon-γ, interleukin-4 (IL4) and interleukin-10 (IL10) and other characteristics of the peptides were analyzed in turn using tools such as TOXINPRED, IFNepitope, IL4Pred and IL-10Pred. Some of the results are shown in Table 1:

[0050] Table 1

[0051]

[0052] Example 5 Peptide affinity analysis

[0053] The Chinese population resource library (such as NyuWa Genome resource) was used to obtain the HLA type coverage of the Chinese population, and HLA typing with a larger coverage ratio was selected to increase the vaccine coverage ratio. The affinity between peptides and HLA was analyzed using NetMHCpan4.0, and some results are shown in Table 2.

[0054] Table 2

[0055] HLA type Peptide Score_EL Rank_EL, % BindLevel HLA-A*26:01 EFVFKNIDGY 0.75517 0.054 <=SB HLA-A*26:01 NVYADSFVIR 0.019316 3.222 HLA-A*26:01 GIYQTSNFRV 0.000251 25.923 HLA-A*26:01 KYFKNHTSPD 0.000003 85 HLA-A*26:01 VNNTVYDPLQ 0.000003 85 HLA-A*26:01 FKNHTSPDVD 0.000001 95

[0056] Example 6 Constructing an epitope prediction model and predicting important epitope sites

[0057] The structural features (solvent accessible surface area, flexibility and rigidity), conservative features of protein sequences, and peptide features (toxicity, induction of interferon-γ, interleukin-4 (IL4) and interleukin-10 (IL10)) were reasonably encoded, selected, reduced in dimension, and normalized, and then input into the ensemble learning XGBoost model framework.

[0058] The objective function of formula I is used as the training target of the epitope prediction model to minimize the objective function.

[0059]

[0060] Among them, l(*) is the loss function, which is used to express the prediction score and the true score y i The gap between is the i-th sample x i The predicted value at the tth iteration is the predicted value at the t-1th iteration plus the predicted value f at the tth iteration. t ; Ω(*) is a regularization term, which is used to prevent overfitting caused by the tree being too complex. It combines the first-order and second-order derivatives of the loss function to make the objective function converge faster;

[0061] Formula II was used as a strategy for generating epitope prediction models.

[0062]

[0063] Among them, D k The characteristic value of each sample representing the kth feature (x nk ) and the corresponding h i The collection composed of h represents all samples of the kth feature i sum, h represents all samples whose eigenvalues ​​are less than z i Sum; weighted quantile method is used in the process of threshold division of features, with the second-order derivative value h i The weight of the sample is used to divide the sample, and the node with the minimum objective function is found as the best split point, and so on, until the tree is built, a series of CART regression trees are generated as the epitope prediction model of the base classifier;

[0064] Then the new sample is sequentially fed into all base classifiers, and the result of adding up the scores of each epitope residue is the score of the corresponding epitope residue. The important epitope sites are selected from the sites with scores significantly higher than other sites. Figure 4 As shown, the important amino acid positions are 206, 336, 493, 495, 498, 502, 481, 490, 676 and 678.

[0065] Example 7 Screening of original vaccine sequences

[0066] Based on the predicted important epitope sites of the spike protein, the corresponding amino acid sequence and DNA sequence were traced back as candidate design schemes for the antigen sequence. According to the affinity between the candidate antigen sequence and MHC, the antigen sequence with the largest coverage ratio was selected as the optimal immunogen sequence. The optimal immunogen sequence was subjected to vaccine production steps such as recombinant plasmid construction, in vitro transcription and modification to produce an mRNA vaccine.

[0067] Finally, it should be noted that the above description is only a preferred embodiment of the present invention. Under the guidance of the present invention, those skilled in the art can make a variety of similar expressions without violating the purpose and claims of the present invention, and such changes all fall within the protection scope of the present invention.

Claims

1. A method for designing mRNA vaccine sequences based on sequence and structure information, characterized in that The following steps are involved: S1. Construct the protein structure of the target protein; S2. Perform structural feature analysis, sequence conservation analysis, peptide feature analysis and peptide affinity analysis on the target protein to obtain corresponding feature data; S3. Input the structural features, sequence conservation features, peptide features and peptide affinity features into the XGBoost learning model, and build an epitope prediction model based on the weak classification algorithm to screen important epitope sites of the target protein; Formula I is used as the training target of the epitope prediction model to minimize the objective function Among them, l(*) is the loss function, which is used to express the prediction score and the true score y i The gap between is the i-th sample x i The predicted value at the tth iteration is the predicted value at the t-1th iteration plus the predicted value f at the tth iteration. t ; Ω(*) is a regularization term, which is used to prevent overfitting caused by the tree being too complex. It combines the first-order and second-order derivatives of the loss function to make the objective function converge faster; Formula II is used as the generation strategy of the epitope prediction model, the minimum objective function value is selected as the root node, and so on until the tree is built to create the epitope prediction model Among them, D k The characteristic value of each sample representing the kth feature (x nk ) and the corresponding h i The collection composed of h represents all samples of the kth feature i sum, h represents all samples whose eigenvalues ​​are less than z i Sum; weighted quantile method is used in the process of threshold division of features, with the second-order derivative value h i The sample is divided according to its weight, and the node with the minimum objective function is found as the best splitting point. S4. The amino acid sequences and DNA sequences corresponding to the important epitope sites screened out are used as candidate antigen sequences, and the best immunogenic sequences are screened according to their affinity with MHC.

2. A method for designing mRNA vaccine sequences based on sequence and structure information according to claim 1, characterized in that The structural feature analysis in S2 includes: amino acid residue solvent analysis, surface area analysis and flexibility analysis.

3. The mRNA vaccine sequence design method based on sequence and structure information according to claim 1, characterized in that The peptide feature analysis in S1 includes: peptide toxicity analysis, induced interferon-γ analysis, interleukin-4 analysis, and interleukin-10 analysis.

4. The mRNA vaccine sequence design method based on sequence and structure information according to claim 1, characterized in that: Sequence conservation analysis uses formula III for sequence comparison analysis cons(i) = 1 + ∑ k p k (i) log p k (i) / log 20 Equation III p k (i) represents the probability of the kth observed amino acid in the entire sequence.

5. The mRNA vaccine sequence design method based on sequence and structure information according to claim 1, characterized in that: NetMHCpan4.0 software was used to analyze the affinity between peptides and HLA.

6. The method for designing mRNA vaccine sequences based on sequence and structure information according to claim 1, characterized in that The screening criteria in S5 are: based on the affinity between the candidate antigen sequence and MHC, the antigen sequence with the largest coverage ratio is selected as the best immunogenic sequence.

Images