Method for guiding viral vaccine component update and immune strategy based on immune response kinetic model

Through the DynaVac model, the normal differential equation model is used to quantitatively describe and predict the immune response dynamics after vaccination, which solves the problem of optimizing the vaccination strategy of the novel coronavirus and achieves the improvement of the immune response effect to the mutant strain.

CN119068970APending Publication Date: 2024-12-03PEKING UNIV +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411227694.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-03
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The antigen evolution of the novel coronavirus has led to limited vaccine-induced antibody responses in preventing infection, and the immunoblotting effect is complex, making it difficult for existing research methods to effectively optimize vaccination strategies.

Method used

The DynaVac model is proposed, and the immune response dynamics after viral vaccination is quantitatively described and predicted by simulating different vaccination strategies, and the vaccine component update and immune strategies are optimized.

Benefits of technology

A comprehensive quantitative prediction of the dynamics of immune response after vaccination is achieved, providing a scientific basis for optimizing vaccine components and vaccination strategies, and improving the immune response effect against mutant strains.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119068970A_ABST
    Figure CN119068970A_ABST
Patent Text Reader

Abstract

The invention discloses a method for guiding viral vaccine component update and immune strategy based on an immune response kinetic model. According to the model, an ordinary differential equation based on an immunological molecular mechanism is used for quantitatively describing and predicting an immune response process generated by a body after viral vaccine inoculation, key cells and molecular processes of antibody immune response are abstracted, and activation, proliferation, competition and differentiation of memory B cells generated by primary immunity in repeated immunity are also considered. By fitting antibody-pseudovirus neutralization titration experimental data under various inoculation combinations, the model can comprehensively and quantitatively describe and predict humoral immune response kinetics induced by vaccines in primary and booster immunization, and has universality and flexibility. By deeply describing immune response kinetics, the model can reveal the inherent law of immunogenicity evolution, and a powerful quantitative tool is provided for optimizing vaccine component design and formulating an immunization strategy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the interdisciplinary field of immunology and computational systems biology, and specifically relates to an innovative computational method that utilizes an ordinary differential equation model based on immunological molecular mechanisms to quantitatively describe and predict the immune response process produced by the body after viral vaccination. Background Art

[0002] The frequent antigenic evolution of the new coronavirus (SARS-CoV-2) has brought unprecedented challenges to the ability of vaccine-induced antibody responses to prevent infection. In response to this problem, the World Health Organization (WHO) recommends updating vaccine components, and some variant booster vaccines have been approved. However, due to the existence of the immunological imprinting effect, the antibody response induced by variant vaccines does not always meet expectations. Immunological imprinting refers to the immune memory produced when the body is first exposed to a pathogen, which will be preferentially activated when similar antigens are encountered again, and may inhibit new antibody responses against new antigen epitopes. Therefore, reducing the inhibitory effect of immunological imprinting-mediated antibody responses against new variants is crucial to optimizing vaccination strategies.

[0003] Recent studies on immunization strategies have shown that increasing antigenic diversity and performing multiple booster immunizations can mitigate the immunological imprinting effect. These studies, together with other work exploring immunological imprinting caused by influenza virus vaccination or infection, provide important empirical insights into the evolution of immunogenicity. However, due to the diversity of population immune history, the pattern of immunological imprinting can be very complex. For example, some studies have reported that strong initial immune intensity may lead to the inability to effectively induce variant-specific antibody responses after repeated Omicron booster immunization. This phenomenon highlights the complexity of the immunological imprinting phenomenon and exposes the limitations of current empirical research methods, because it is unrealistic to exhaustively experiment to verify every possible immunological imprinting situation. Therefore, it is urgent to invent systems biology methods to quantitatively model and analyze the cellular mechanisms behind the evolution of antibody immunity. This is of great significance for understanding the complex immune response dynamics during repeated vaccination with homologous or heterologous new crown variants, and for accurately optimizing vaccine component updates and immune strategy formulation accordingly. Summary of the invention

[0004] The present invention proposes a general mathematical model framework named DynaVac, which is used to quantitatively describe and predict the dynamics of humoral immune responses induced by vaccination, especially to capture the dynamic evolution of various immune components during primary and booster immunization.

[0005] The DynaVac model first abstracts the cellular and molecular processes of antibody immune response into several key components, including mRNA vaccine, antigen, naive B cells, antigen-specific memory B cells and antibodies. We first used ordinary differential equations (ODEs) to characterize the dynamic changes of these components over time during the primary immunization process, covering a series of processes such as mRNA translation to produce antigens, affinity maturation of naive B cells, differentiation into memory B cells and plasma cells that secrete antibodies, and antigen-antibody neutralization.

[0006] To further characterize the secondary immune response induced by homologous or variant antigens, we extended the model to include the activation, proliferation, and differentiation of memory B cells generated by the primary immunization against the original antigen in repeated immunization. The model takes into account the cross-neutralization effect of antibodies induced by the primary immunization on the new antigen, as well as the dynamic characteristics of memory B cells with different reactivities when competing for antigen signals. We introduced the cross-neutralization coefficient c between different variant antigens in the ODE framework. i,j , in order to explore the interaction between the primary immunization antigen and the booster immunization antigen. These coefficients can be directly measured from the neutralization experiment, which determines the strength of the cross-reaction between the antigen and the antibody, and thus affects the proliferation of cross-reactive memory B cells during the heterologous booster immunization process. The proliferation of memory B cells depends on the proliferation signal provided by the antigen presenting cells, and the coefficient c i,j This reflects the characteristics of each population competing for antigens when there are multiple memory B cell populations targeting different antigens.

[0007] Specifically, the present invention uses a group of ordinary differential equations to quantitatively describe and predict the dynamics of the immune response of the virus vaccine, and establishes a vaccination strategy simulation algorithm. On this basis, the parameters are fitted according to the experimental data to realize the construction of the immune response dynamics model, which includes the following steps:

[0008] (1) Establish a set of ordinary differential equations to describe the dynamics of the immune response to viral vaccines, characterizing the relationship between antigen (Ag), initial B cells (N), and memory B cells (M). on , M off ) and antibody (Ab) and other key components, the ordinary differential equations are:

[0009]

[0010] Among them, i, j, and k represent the numbers of the virus variants, and n represents the total number of virus variants involved;

[0011] Ag stands for antigen, then Ag i 、Ag j 、Ag k Represent the specific antigen levels of virus variants i, j, and k, respectively;

[0012] Ab stands for antibody, then Ab i ,Ab j Represent the levels of specific antibodies against virus variants i and j, respectively;

[0013] N represents the initial B cells, then N i represents the level of the initial B cell population specific for the viral variant i antigen;

[0014] M stands for memory B cells, M off Indicates the non-activated memory B cell population, M on indicates the activated memory B cell population, then Respectively represent the size of the non-activated memory B cell population specific to the antigen of virus variants i and j, They represent the size of activated memory B cell populations specific for the antigens of virus variants i and j, respectively;

[0015] γ neu represents the antigen-antibody neutralization rate, then They represent the neutralization rates of specific antigens and antibodies against virus variants i and k, respectively;

[0016] γ Ag represents the rate of antigen degradation, γ Ab Represents the antibody degradation rate; s N represents the maximum rate of initial B cell affinity maturation, s M represents the maximum expansion rate of memory B cells; d N represents the initial B cell decay rate, d M represents the decay rate of memory B cells; k N2M represents the maximum rate of differentiation of initial B cells into memory B cells; p N represents the maximum initial antibody production rate, p M represents the maximum memory antibody production rate; m0 represents the cross-reaction affinity threshold for memory B cell expansion; c0 represents the cross-reaction affinity threshold for memory B cells to differentiate into antibody-producing plasma cells;

[0017] c i,j 、c j,i 、c i,k 、c j,k represents the cross-neutralization coefficient, expressed as c i,j For example, it is defined as the affinity constant Ka of the specific antibody of virus variant j to the specific antigen of virus variant i. i,j The affinity constant Ka of the specific antibody of the virus variant j to its own antigen j Ratio:

[0018] c i,j =Ka i,j / Ka j

[0019] By definition, c i,i =1, and for i≠j,c i,j ∈(0,1);

[0020] Time window a N (t) Describes the duration of the initial immune response:

[0021]

[0022] Among them, t0 represents the vaccination time point, t ini is the time interval before the onset of the germinal center reaction, t min and t max are the shortest and longest durations of the immune response, respectively. The response lasts at least t min , at t min to max The duration between the two depends on the amount of residual antigen, and the max , the reaction stops;

[0023] The specific antigen of virus variant i is expressed as function f(Ag i , K) reaches saturation:

[0024]

[0025] f(Ag i , K)∈[0,1) represents the surface fraction occupied by antigen relative to the maximum loading capacity of antigen-presenting cells, where K represents the amount of antigen required to occupy half of the maximum loading capacity of antigen-presenting cells;

[0026] Time window a M (t) Describe the duration of the memory immune response:

[0027]

[0028] Among them, t0 represents the vaccination time point, t ini is the time interval before the onset of the germinal center reaction, t min and t max are the shortest and longest durations of immune response, respectively; memory immune response occurs rapidly at the vaccination time point t0;

[0029] (2) Based on the ODE model constructed above, develop a computational method to simulate any vaccination strategy:

[0030] For any vaccination strategy that only includes monovalent vaccines, it is first encoded into four vectors T = (t 01 , t 02 ,…,t 0m), V=(v1,v2,...,v m ), D=(d1, d2, ..., d m ), P=(p1, p2, ..., p m ). This strategy involves n virus variants and m consecutive vaccinations. Let k be an integer from 1 to m, t 0k is the time of the kth vaccination, vk is the variant number of the vaccine used for the kth vaccination, d k is the vaccination dose of the kth vaccination; p k ∈{1, 2, 3} represents the type of vaccine used in the kth vaccination (1 represents protein vaccine, 2 represents mRNA vaccine, and 3 represents inactivated vaccine); then the initial condition of the immune response kinetic model after the first vaccination (k=1) is:

[0031]

[0032] The initial condition of the immune response kinetic model after the kth vaccination is:

[0033]

[0034] N k (t 0k )=N k-1 (t 0k ), Ab k (t 0k )=Ab k-1 (t 0k )

[0035] Among them, i represents the virus variant number, P s , P r , P v They are the antigen content per unit dose of recombinant protein vaccine, mRNA vaccine and inactivated vaccine; Ag k-1,i (t 0k ) represents the antigen level of variant i before the kth vaccination; Ag k,i (t 0k ) represents the antigen level of variant i after the kth vaccination; N k-1 (t 0k ), Ab k-1 (t 0k ), N k (t 0k ), Ab k (t 0k ) are all n-dimensional vectors. N k-1 (t 0k ), Ab k-1 (t0k ) represent the levels of n variant - specific naïve memory B cells, activated memory B cells, naive memory B cells, and antibodies before the k - th vaccination, respectively; N k (t 0k ) and Ab k (t 0k ) represent the levels of n variant - specific naïve memory B cells, activated memory B cells, naive memory B cells, and antibodies after the k - th vaccination, respectively.

[0036] For the vaccination strategy involving multivalent vaccines, only need to replace the m - dimensional vector V with a q×m matrix:

[0037]

[0038] where q represents the highest vaccine valence in the multivalent vaccine vaccination strategy, and v j,k ∈{0, 1, 2,..., n} represents the variant number of the j - th antigen component in the vaccine used for the k - th vaccination. If the valence of the vaccine used for the k - th vaccination (denoted as q k ) is lower than the highest vaccine valence q in the strategy, then fill the missing antigen components with 0, that is, for the k - th column (the k - th vaccination) of the matrix V, if q k <j≤q, then v j,k = 0. The initial conditions of the antigen levels for the first vaccination (k = 1) and subsequent vaccinations (k>1) are adjusted respectively as:

[0039]

[0040] where q k represents the valence of the vaccine used for the k - th vaccination, v.k represents the k - th column of the matrix V, and the initial conditions of the remaining components are the same as those of the monovalent vaccine vaccination strategy.

[0041] According to the initial conditions of the above equations and the ODE system established in step (1), the dynamics of the humoral immune response after each vaccination can be numerically simulated in the time period [t 0k , t 0k+1 .

[0042] t∈[t 0k , t 0k+2 is the simulation result of the k - th vaccination. Ag k (t), N k (t), Ab k (t) are all n - dimensional vector functions of time t, representing respectively in the time period [t 0k , t 0k+1] Dynamic changes of specific antigens, initial memory B cells, non-activated memory B cells, activated memory B cells and antibody levels of n variants over time t. Connecting the simulation results of each vaccination, we can get the complete simulation results of the vaccination strategy:

[0043] [Ag(t), N(t), M off (t), M on (t),Ab(t)],t∈[t 01 , t end ]

[0044] Among them, t end is the simulation end time, t end >t 0m .

[0045] (3) Based on the above simulation algorithm, the antibody-pseudovirus neutralization titer data measured in the experiment were fitted, and the parameters in the ordinary differential equations were estimated using an intelligent optimization algorithm to obtain a parameterized immune response kinetic model.

[0046] In some embodiments of the present invention, the parameters of the immune response kinetic model are estimated by fitting the serum antibody-pseudovirus neutralization titration experimental data of various vaccination schemes, and the specific method is as follows:

[0047] ⅠData Collection

[0048] Design a variety of vaccination schemes in experimental animals (such as mice, etc.), and collect serum antibody-pseudovirus neutralization titer data under each scheme experiment. These schemes cover homologous and heterologous immunization strategies.

[0049] Ⅱ Construction of Neutralization Titer Matrix

[0050] Using the serum antibody-pseudovirus neutralization titer data of homologous immunization, the antigen-antibody neutralization titer matrix X = (x i,j ) n×n . Where x i,j represents the neutralization titer of variant j-specific antibodies against variant i-specific antigens, and n represents the total number of virus variants (vaccines) involved in the experiment. The vector x composed of the diagonal elements of X = (x 11, x 22 , …, x nn ), representing the neutralization titer of each antibody against its own antigen, and these data can be further used to estimate the relative proportion of antibody-specific neutralization rates.

[0051] Ⅲ Calculation of relative ratio of self-antigen-antibody neutralization rate

[0052] A variant strain is selected as the benchmark, which is numbered as δ, and its autoantigen-antibody neutralization rate is defined as γneu For other variants, the relative value of the autoantigen-antibody self-neutralization rate of each variant and the variant δ is calculated based on the experimentally measured autoantigen-antibody neutralization titer data x Then construct the relative ratio vector of self-antigen-antibody neutralization rate:

[0053]

[0054] Then we can further get the autoantigen-antibody neutralization rate vector Γ neu :

[0055]

[0056] IV. Construction of antigen-antibody cross-neutralization matrix

[0057] The neutralization efficiency of each variant-specific antibody against non-self antigens was calculated through the antigen-antibody neutralization titer matrix X, thereby constructing the antigen-antibody cross-neutralization matrix C. The matrix elements were calculated using the following formula:

[0058] c i,j =x i,j / x j,j

[0059] Among them, x i,j represents the neutralization titer of antibody j against antigen i, x j,j is the neutralization titer of antibody j against its own antigen.

[0060] Ⅴ Construction of pseudovirus-antibody cross-neutralization matrix

[0061] To estimate other parameters, a pseudovirus-antibody neutralization titer matrix was constructed based on experimental data. in represents the neutralization titer of variant j-specific antibody against pseudovirus i, n′ represents the total number of pseudoviruses involved in the experiment, and n represents the total number of virus variants (vaccines) involved in the experiment. The pseudovirus-antibody cross-neutralization matrix is ​​constructed by the following formula

[0062]

[0063] Defining antibody-pseudovirus cross-neutralization coefficient Neutralization titer of antibody against any pseudovirus Neutralization titer against self-antigens (x i,i ) ratio.

[0064] VI. Parameter optimization using all neutralization titer data

[0065] Assume that the experiment involves L vaccination strategies, T l , Vl , D l , P l They respectively represent the vaccination time, vaccine variant number, vaccination dose and vaccine type of the first vaccination strategy encoded in the manner of step (2).

[0066] Let the model parameter set to be determined be θ, then the specific antibody level vector Ab of each variant corresponding to the first vaccination strategy can be obtained by simulation according to the algorithm in step (2): l (n-dimensional), written as time t, vaccination strategy T l , V l , D l , P l And the parameter set θ as a function of the independent variable:

[0067] Ab l (t,T l , V l , D l , P l ,θ)

[0068] Ab lm (θ) represents the serum antibody measurement time t corresponding to the lth immunization strategy when the parameter set is θ lm The following vectors show the specific antibody levels for each variant:

[0069] Ab lm (θ)=Ab l (t lm , T l , V l , D l , P l ,θ)

[0070] make Represents the model prediction value of the neutralization titer matrix of serum antibodies against n' pseudoviruses corresponding to all L immunization strategies; Ab m (θ)=(Ab 1m (θ), Ab 2m (θ), ..., Ab Lm (θ)) is the model prediction value of the antibody level matrix at the serum antibody measurement time corresponding to all L immunization strategies. Then:

[0071]

[0072] in, represents the pseudovirus-antibody cross-neutralization matrix, Γ neu represents the antigen-antibody self-neutralization rate vector, J L =(1, 1, ..., 1) 1×Lis a row vector of all 1s, and ° represents the matrix Hadamard product. N=(N i,l ) n′×L Represents the experimental observation values ​​of the neutralization titer matrix of serum antibodies against n′ pseudoviruses corresponding to all L immunization strategies.

[0073]

[0074] Among them, rN and These are the log2 fold change matrices of the neutralization titer matrix of the actual experiment and the model prediction, respectively.

[0075] The objective function Loss, which measures the goodness of fit of the model, is defined as the model-predicted value of the neutralization titer The deviation measure from the experimental observation value rN is:

[0076]

[0077] Where θ is the model parameter vector and the subscript F represents the Frobenius norm of the matrix.

[0078] In the specific implementation, according to the defined objective function Loss, a genetic algorithm is used to iteratively optimize the parameters. In some embodiments of the present invention, the genetic algorithm is implemented using the 'ga' function in the Global Optimization Toolbox of MATLAB, each parameter set is encoded as a chromosome, and the fitness function is set to the inverse of the objective function L to minimize the objective function value.

[0079] Based on the immune response kinetic model established by the above method, the present invention provides a method for optimizing viral vaccine components and immunization strategies, namely, using the model parameterized by step (3) to simulate the multi-dose vaccination process of viral variant vaccine through the algorithm in step (2), predict the corresponding immune response dynamics, and compare and evaluate the neutralization level of the immunogenicity induced by the candidate immunization strategy on the target variant, so as to obtain the optimal vaccine components, dosage and schedule.

[0080] Furthermore, the present invention also provides a readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the method for optimizing virus vaccine components and immunization strategies based on an immune response kinetic model is implemented. An electronic device including a memory storing the computer program and a processor is also within the scope of protection of the present invention.

[0081] In summary, the DynaVac model proposed in the present invention consists of a set of nonlinear ordinary differential equations with a dimension of 6×n (n represents the total number of variants). By fitting the antibody-pseudovirus titration experimental data under various vaccination combinations, the model can comprehensively and quantitatively characterize and predict the dynamics of humoral immune responses induced by the new crown vaccine in primary and booster immunizations, providing a powerful tool for in-depth understanding of the evolution of immunogenicity, optimizing vaccine component design and immunization strategies. The advantage of the DynaVac model lies in its universality and flexibility. It is not only applicable to new crown vaccines of different technical platforms, but can also be expanded to analyze vaccine-induced immunity of other pathogens. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 .DynaVac model concept diagram and equations, where:

[0083] a. Simplified cellular and molecular processes of humoral immune responses after primary and booster vaccination. The left figure outlines the key humoral responses during primary vaccination, including translation of mRNA (vaccine) into antigen, affinity maturation of naive B cells and differentiation into memory B cells, antibody-secreting plasma cells, and antibody-antigen neutralization. The right figure describes the key processes in heterologous booster vaccination, including the recall of existing memory B cells and cross-neutralization of variant antigens by antibodies induced by the prototype vaccine.

[0084] b. Conceptual model of cellular and molecular processes used to describe immunogenicity evolution. The left figure abstractly shows the key variable interactions between mRNA (R), antigen (Ag), naive B cells (N), memory B cells (M), and antibodies (Ab) in primary vaccination, using colored arrows to represent specific biological processes. The right figure shows the interaction between multiple variant antigens (Ag i ), initial B cells (N i ), resting and active memory B cells (Moff, Mon), and antigen-specific antibodies (Ab i ). Solid arrows represent interactions between antigen-specific variables, while dashed arrows represent cross-reactive immune interactions including cross-neutralization between different antigens and antibodies (red dashed arrows) and cross-reactive memory B cell expansion (purple dashed arrows).

[0085] c. Development of ordinary differential equations (ODEs) to quantitatively model the dynamic evolution of immunogenicity of variant vaccines during sequential immunizations. The left figure shows the ordinary differential equations used to model the temporal dynamics of cellular and molecular variables in the primary immune response. The right figure shows the extended ODE framework used to capture the dynamics of naïve and memory B cell interactions in heterologous booster vaccination, including the cross-neutralization coefficients (c i,j). Each term in the equation is colored according to the corresponding biological process in b.

[0086] Figure 2 .Comparison of experimentally measured and DynaVac model-predicted neutralizing titers. The scatter plot shows the log2 fold change in neutralizing titers against 11 SARS-CoV-2 variants (rows) for 28 immunization regimens (columns), with the reference being the neutralizing titers against the prototype strain after two doses of the inactivated vaccine (first row, first data point). Experimental data (orange) were used to parameterize the model, and model predictions are shown in purple. Asterisks indicate data points that were corrected by the model due to experimental measurement limits. The panels below describe the detailed immunization schedules, with gray circles representing inactivated vaccines, colored squares representing monovalent mRNA vaccines, and mixed color squares representing bivalent vaccines.

[0087] Figure 3 The DynaVac model was parameterized and validated using the data from this study. The scatter plot shows the comparison between the log2 neutralization titer fold change predicted by the model and the experimental observations in the data set used for parameter estimation. The coefficient of determination (R 2 ) represents the proportion of the observed data variation that can be explained by the model prediction results.

[0088] Figure 4 Computer simulations using the DynaVac model revealed an interplay between antigenic distance and immunohistochemistry, where:

[0089] a. Schematic diagram of the simulated vaccination scheme: the primary immunization includes two doses of the prototype strain-specific vaccine on days 0 and 30, respectively, followed by two doses of the variant strain-specific vaccine for booster immunization on days 390 and 420, respectively.

[0090] b. Dynamic changes of key components of humoral immunity during the evolution of immunogenicity at different antigenic distances. Each row shows the changing trends of antigen levels, initial B cell maturity, memory B cell levels and antibody levels over time. The gray area represents the prototype strain-specific quantity, and the red area represents the variant strain-specific quantity. The first column represents the simulation results of the primary immunization (0-60 days). The second, third and fourth columns respectively represent the simulation results of heterologous booster immunization (360-434 days) using variant vaccines with antigenic distances of 0.5, 0.95 and 0.99. Among them, the antigenic distance is defined as 1 minus the cross-neutralization rate between the prototype strain and the variant antigen. DETAILED DESCRIPTION

[0091] The embodiments of the present invention will be described in detail below by way of examples in conjunction with the accompanying drawings. However, the following examples are only used to illustrate the present invention and should not be construed as limiting the scope of the present invention.

[0092] 1 Immunokinetic modeling method

[0093] 1.1 Overview of humoral immune response

[0094] After the first infection or vaccination, the precursor naive B cells that have undergone VDJ gene recombination in the bone marrow enter the lymph nodes. These cells have not yet encountered the antigen, so their B cell receptors (BCR) have a low affinity for a given antigen. If an antigen such as a viral surface protein binds to the BCR of a naive B cell with sufficient affinity, the naive B cell will be activated and initiate a germinal center reaction in the lymph nodes.

[0095] In the germinal center, under the action of activation-induced cytidine deaminase, B cell progeny will undergo high-frequency somatic mutations. After mutation, the affinity of BCR for the present antigen may increase or decrease. If the binding ability to antigens presented on the surface of follicular dendritic cells is enhanced, B cells can internalize antigens and present peptide-MHC-II complexes to helper T cells, obtaining positive selection signals (mainly cytokines). B cells that obtain positive selection further proliferate, while unselected B cells undergo apoptosis. This Darwinian process, namely affinity maturation, gradually increases the average antibody affinity of the B cell population in the germinal center.

[0096] Among affinity-matured B cells, some differentiate into memory B cells, and others become plasma cells that secrete antibodies. Memory B cells are a key component of acquired immunity and can last for months to years. They maintain the expression of high-affinity BCRs. When exposed to the same or similar antigens again, memory B cells can be rapidly activated, proliferate, and differentiate to produce large amounts of high-affinity antibodies ( Figure 1 a and b, first column).

[0097] 1.2 Mathematical modeling of primary immunization

[0098] Based on the above overview of humoral immune response, a set of minimal ordinary differential equation (ODE) models can be constructed to describe the dynamics of key components in the primary immune response of mRNA vaccines ( Figure 1 c, first column). The state variables in the ODE model include:

[0099] R: mRNA level

[0100] Ag: Antigen level

[0101] N: Total affinity level of initial B cells

[0102] M: memory B cell level

[0103] Ab: Antibody level

[0104]

[0105]

[0106] Formula (1) describes the dynamics of mRNA levels. After the mRNA vaccine is injected into the body, mRNA is expressed in the form of γ R degradation rate.

[0107] Formula (2) describes the dynamics of antigen levels. Antigenic proteins (such as the SARS-CoV-2 spike protein) are produced from mRNA at a translation rate of k. Antigens bound to antibodies are cleared by the immune system. It is assumed that the binding and dissociation rates of antigen and antibody are much faster than the degradation rate of the antigen-antibody complex. This allows a quasi-steady-state approximation of the antigen-antibody complex concentration:

[0108]

[0109] Where K a represents the affinity constant between antigen and antibody, [Ag:Ab] represents the antigen-antibody complex. The degradation rate of the antigen-antibody complex is set to γ c .

[0110] γ c [Ag:Ab]=K a γ c [Ag][Ab]

[0111] In addition, the antigen itself is γ Ag degradation rate.

[0112] Formula (3) describes the maturation dynamics of the naive B cell population. In fact, most naive B cells in the germinal center express different B cell receptors with different affinities for antigens. Here, we use the mean field approximation to model the overall maturation process - using a variable N (with a value range of [0,1]) to represent the overall affinity maturation of the naive B cell population. N depends on both the average antigen affinity of the naive BCR and the total number of naive B cells. At the beginning of the formation of the germinal center, N = 0, because there are no naive B cells in the lymph nodes.

[0113] In the early stages of the humoral immune response, precursor naive B cells undergoing VDJ recombination enter the lymph nodes and seed the germinal centers. ini represents the constant maturation rate of the initial B cell overall affinity level N during this seeding process. The time window of the above process is represented by the rectangular function a(t):

[0114]

[0115] t0 represents the vaccination time point, t iniThe time interval before the onset of the germinal center reaction, during which precursor naïve B cells seed the lymph nodes.

[0116] The germinal center reaction then occurs, in which antigen-specific naive B cells proliferate under somatic high-frequency mutation and T cell selection, increasing the degree of maturity, that is, the overall affinity level N of naive B cells. The maturation rate is driven by the successful binding between the naive B cell receptor and the antigen presented on the surface of antigen-presenting cells (APCs, such as follicular dendritic cells FDCs). Therefore, the maturation rate is proportional to the overall affinity level N of the naive B cell population and the level of antigen presented by APCs. Due to the natural loading capacity of APCs, the antigen level Ag reaches saturation according to the function f(Ag, K).

[0117]

[0118] f(Ag, K)∈[0,1) represents the surface fraction occupied by antigen relative to the maximum loading capacity of APC, and K represents the amount of antigen required to occupy half of the maximum load of antigen presenting cells. Due to the natural limit of affinity between antigen and antibody and the limited size of germinal centers, the maturation degree N approaches the maximum value of 1. To reflect this saturation, the maturation rate is also proportional to (1-N). Parameter s N represents the maximum maturation rate (when f(Ag, K)→1 and N→0). The complete maturation rate is:

[0119]

[0120] Time window a N (t) Describes the duration of the initial immune response:

[0121]

[0122] where t min and t max are the shortest and longest durations of the immune response, respectively. The response lasts at least t min , at t min to max The duration between the two depends on the amount of residual antigen (Ag>0). max , the reaction stops.

[0123] In addition, naive B cells are expressed as N The rate of decay.

[0124] Formula (4) describes the process of memory B cells differentiating from naive B cells and decaying. Generally, only naive B cells with sufficient affinity and maturity can leave the germinal center and differentiate. Therefore, the generation rate of memory B cells is proportional to the overall maturity N of the naive B cell population, expressed as the rate constant k. N2MThe generation rate of memory cells is multiplied by (1-M), which reflects the intrinsic upper limit of the memory B cell population. In addition, memory B cells are expressed as d M The rate of decay.

[0125] Formula (5) describes the secretion and clearance of antibodies. Similar to memory B cell differentiation, only mature naive B cells can differentiate into plasma cells that secrete antibodies. Therefore, the antibody secretion rate is proportional to the maturity N, and the rate constant is p N (Plasma cells are an intermediate and do not need to be explicitly modeled.) Ab The rate of decay.

[0126] 1.3 Mathematical modeling of booster immunization

[0127] 1.3.1 Modeling of homologous boost immunization

[0128] In the secondary immune response during booster immunization, both primary and memory immunity participate in antibody production. We first modeled the simpler homologous booster immunization situation.

[0129] Memory B cells generated during the primary immune response do not immediately participate in the initial immune response. They are activated upon secondary exposure to promote enhanced immune responses. To distinguish between memory B cells in primary and secondary immune responses, we divide them into non-activated (M) and non-activated (N). off ) and activation (M on ) memory cell population.

[0130] M off Represents memory cells newly generated in the initial immune response that have not yet encountered the antigen again. For simplicity, we assume that at the time of the second vaccine exposure, all M off The cells are transformed into activated M on Similar to affinity maturation of naive B cells, these activated memory cells undergo affinity-dependent proliferation, which can be modeled by:

[0131]

[0132] Here M Represents the maximum proliferation rate of memory cells.

[0133] In addition, activated memory cells also differentiate into plasma cells that secrete antibodies, with a rate constant p M Indicates. Time window a M (t) Describe the duration of the memory immune response:

[0134]

[0135] Considering the rapid responsiveness of the memory immune response, we modeled its activation as occurring rapidly at the vaccination time point t0, while keeping other temporal characteristics similar to the initial immune response duration a. N (t)Same.

[0136] By adding M to the primary immune response model on With the help of the following model, we can simulate the same antigen boost immunization:

[0137]

[0138] 1.3.2 Modeling of heterologous boost immunization

[0139] The continuous evolution of viruses often leads to heterologous infection (or vaccination) with related but not identical antigens. Immunological imprints (original antigenic sins) will also be manifested in this situation. Therefore, we extended the model to capture the dynamics of heterologous booster immunity. The model tracks multiple antigenic variants (Agi), as well as the naive B cell population (Ni), memory B cell population (Ni) specific for each variant antigen. and antibodies (Ab i )( Figure 1 a and b, second column).

[0140] The system of ordinary differential equations is expanded to a multivariate form:

[0141]

[0142]

[0143] The key additions to the above formula include:

[0144] 1) Cross-neutralization between antigen and antibody

[0145] Heterologous antigens can be neutralized by multiple antibodies generated against different variants. Each antibody Ab j Antigen Ag i The neutralization effect is summed, and the weight is the cross-neutralization coefficient c i,j .

[0146] Experimental data showed that antibodies induced by different variants have unequal intrinsic affinities for their matching antigens (self-neutralization). i represents the specific affinity constant of variant i (i.e. Ab i Ag i affinity constant).

[0147] Cross neutralization coefficient c i,jIt is defined as the affinity constant (Ka) of variant j-specific antibody to variant i-specific antigen i,j ) and the affinity constant (Ka) of the variant j-specific antibody to its own antigen j ) ratio:

[0148] c i,j =Ka i,j / Ka j (11)

[0149] According to the definition, c i,i =1 (because Ka i,i =Ka i ), and for i≠j,c i,j ∈(0, 1) (assuming that the antibody has a higher affinity for its own antigen than for other variants).

[0150] 2) Competition between memory B cell populations

[0151] When exposed to antigen k again, memory cells specific for different antigens compete for the proliferation signal. Their ability to bind and acquire antigen k determines the strength of the received signal. High affinity for antigen k and large cell population size both favor antigen capture. Therefore, the total binding potential of antigen k is given by the affinity constant Ka. i,k =Ka k c i,k Multiply by the population size of each memory cell population i This interpopulation competition kinetics for antigen K is expressed by dividing the antigen-dependent proliferation rate of each memory cell population by the total potential of all populations To quantify.

[0152] Furthermore, in the absence of competition, the memory cell population The proliferation potential of quantification, where m0 represents the affinity-cell population size product threshold. is much larger than m0, the maximum proliferation rate is reached. Close to m0, proliferation decreases significantly. Far below m0, cross-reactivity can hardly drive further proliferation of memory B cells.

[0153] 3) Inhibition of differentiation of mismatched memory B cells into antibody-secreting cells

[0154] When memory B cells do not completely match the booster antigens of the second vaccination, their differentiation into antibody-secreting plasma cells will be inhibited due to the reduction of cross-reactivity. The degree of this inhibition can be expressed by the formula Here, c0 represents the cross-reactivity threshold, which represents the cross-reactivity boundary where antibody production is significantly reduced.

[0155] This final framework can flexibly simulate the humoral immune responses induced by any vaccination combination of different variants ( Figure 1 c, second column).

[0156] 2 Estimation of prior intervals for parameters

[0157] All the original parameters involved in the above ODE model are shown in Table 2.1. In order to facilitate accurate parameter estimation in the subsequent steps, we first estimated the prior interval of each parameter. Most of the estimates are based on existing experiments or mathematical modeling studies. Some parameters are proposed for the first time in this model, and we also conservatively estimate their approximate reasonable range.

[0158] Table 2.1 Original model parameters and their prior intervals

[0159]

[0160]

[0161] 3 Model simplification

[0162] Next, we simplified some model parameters without affecting model performance.

[0163] 3.1 Simplification of mRNA translation process

[0164] Considering the process of mRNA vaccine translation into antigen alone, it can be expressed by the following simple linear ODE:

[0165]

[0166] Assume that the dose of the injected mRNA vaccine is mμg. The initial conditions of the above ODE can be expressed as:

[0167]

[0168] Solving the ODE yields:

[0169]

[0170] When t→∞, the total amount of antigen translated from mμg mRNA vaccine can be expressed as Assume t 99 It represents the time required for 99% of the antigen to be translated from the mRNA. It satisfies the following equation:

[0171]

[0172] According to γ ​​in Table 2.1 R A priori interval estimate of ,t 99 The estimated value is less than 1 day. This means that due to the rapid degradation of mRNA, 99% of the antigen can be translated within 1 day after injection. This time scale is much shorter than the total duration of the humoral immune response. In addition, according to the interval between vaccination and the onset of antibody production in Table 2.1, t ini ∈[0.5, 2], within one day after vaccination, antibodies have not yet begun to be produced or have just begun to be produced.

[0173] In summary, we can omit the translation and degradation process of mRNA in the model. When injecting mμg mRNA vaccine, the corresponding initial condition of antigen can be equivalently set to Ag(t0)=mP r ,in It represents the equivalent amount of antibody produced by each μg mRNA vaccine. This simplification not only reduces the dimension of the ODE system, but also combines the three parameters into one parameter.

[0174] 3.2 Simplification of Antigen-Antibody Neutralization Parameters

[0175] In the original model, the degradation rate of antibody-antigen complex is γ c and antibody-antigen affinity constant It appears in the equation as a product (see formula (6), (10)). Therefore, we introduce a new parameter To express the specific neutralization rate of variant i (Ab i Ag-mediated i Neutralization rate).

[0176] 3.3 Simplification of parameter units

[0177] According to Table 2.1, parameter P s , P v and the newly defined P r The unit is M / μg, and the order of magnitude is 10 -12 To simplify parameter scale, we rescale the units of these parameters to 10 -12 M / μg. Accordingly, to maintain unit consistency, we rescale the unit of parameter K from M to 10 -12 M, the unit of parameter Ka is from M -1 Rescale to 10 12 M -1 .

[0178] After these simplifications, the original ODE model in formulas (6)-(10) can be simplified to:

[0179]

[0180] The parameters of the simplified model and their corresponding prior intervals are shown in Table 3.1.

[0181] Table 3.1 Simplified model parameters and their values ​​after parameterization by experimental data

[0182]

[0183]

[0184] 4. Simulating vaccination strategies using the DynaVac model

[0185] 4.1 Simulation of monovalent vaccination strategy

[0186] Consider a simulation involving n variants and m consecutive vaccinations. The vaccination strategy can be represented by four m-dimensional vectors T, V, D, P, where:

[0187] T=(t 01 , t 02 , ..., t 0m ), satisfying 0≤t 01 <t 02 <... <t 0m , and t 0k represents the time of the kth vaccination (in days);

[0188] V=(v1,v2,...,v m ), where v k ∈{1, 2, ..., n} represents the variant number used in the kth inoculation;

[0189] D=(d1,d2,...,d m ), where d k >0 indicates the dose of the kth vaccination (in μg);

[0190] P=(p1,p2,...,p m ), where p k ∈{1, 2, 3} represents the type of vaccine used for the kth vaccination (1: recombinant protein vaccine; 2: mRNA vaccine; 3: inactivated vaccine).

[0191] Furthermore, we define:

[0192] n-dimensional vector of the neutralization rate of each variant's own antigen-antibody;

[0193] C=(c i,j ) n×n , n×n antigen-antibody cross-neutralization matrix, where c i,jDenote the cross - neutralization coefficient of the mutant - strain j - specific antibody against antigen i, and c i,i = 1 when i ≠ j, and c i,j ∈(0, 1).

[0194] The method for estimating the vector Γ neu and the matrix C using experimental data will be described in the next section.

[0195] Then, the vaccination strategy can be simulated by numerically solving the initial - value problem for the ODEs shown in equations (12) - (16) for each vaccination. Let t ∈ [t 0k , t 0k+1 be the numerical solution for the k - th vaccination.

[0196] For the first vaccination (k = 1), the initial conditions are:

[0197]

[0198] For subsequent vaccinations (1 < k ≤ m), the initial conditions are:

[0199]

[0200] N k (t 0k ) = N k-1 (t 0k ), Ab k (t 0k ) = Ab k-1 (t 0k )

[0201] Each new vaccination is assumed to activate all memory B - cells. Thus, at the time point t 0k , the level of unactivated memory B - cells M off is emptied and added to the level of activated memory B - cells M on .

[0202] The complete simulation result of the vaccination strategy on [t 01 , t end can be obtained by concatenating the numerical solutions:

[0203] [Ag(t), N(t), M off (t), M on (t), Ab(t)], t ∈ [t 01 , t end

[0204] where t end > t 0m is the end - time of the simulation.

[0205] ​4.2 Simulation of Multivalent Vaccine Immunization Strategies

[0206] For immunization strategies involving multivalent vaccines, we can extend the monovalent vaccine simulation method by replacing the m - dimensional vector V with a q×m matrix:

[0207]

[0208] where q represents the highest vaccine valence in the immunization strategy, and v j,k ∈{0, 1, 2,..., n} represents the variant of the j - th antigenic component in the vaccine used for the k - th vaccination. If the valence of the vaccine used for the k - th vaccination (denoted as q k ) is lower than the highest vaccine valence q in the strategy, then for all j satisfying q k <j≤q, v j,k = 0.

[0209] The initial conditions for multivalent vaccine simulation are adjusted accordingly: For the first vaccination (k = 1) and subsequent vaccinations (k>1), the initial antigen levels are given by:

[0210]

[0211] where q k represents the valence of the vaccine used for the k - th vaccination, and v .k represents the k - th column of the matrix V. To ensure that the total dose of the vaccine used for the k - th vaccination is equal to d k , the dose of each variant antigen in the q k -valent vaccine is set to d k / q k . The initial conditions for other components are the same as those in the monovalent vaccine simulation.

[0212] 5 Parameter Estimation Using Experimental Data

[0213] To validate the DynaVac model and estimate its parameters, we designed a set of 28 comprehensive immunization regimens consisting of 7 mRNA vaccines against SARS - CoV - 2 variants in C57BL / 6J mice. We evaluated the neutralization titers of serum antibodies against 11 SARS - CoV - 2 variants using pseudovirus assays. These immunization regimens included a series of two - dose monovalent or bivalent vaccine vaccinations against various SARS - CoV - 2 variants to determine the neutralizing antibody profiles induced by each variant vaccine alone. Then, we constructed prime - boost immunization scenarios through a series of homologous or heterologous sequential booster immunization regimens following a two - dose primary immunization with prototype inactivated vaccines.

[0214] 5.1 Estimation of Variant - Specific Autoneutralization Rates and Cross - Neutralization Matrices Using Homologous Vaccine Sequential Immunization Regimens

[0215] Of the 28 vaccination regimens designed, 6 involved two consecutive vaccinations with the same monovalent vaccine (inactivated CoronaVac for the prototype strain and mRNA vaccines for Alpha / Beta, Delta, BA.1, BA.2 / 4 / 5, and XBB.1.5) without prior immunostaining. This data can be used directly to estimate the antigen-antibody self-neutralization rate specific to each variant. Table 5.1.1 lists the original pseudovirus neutralization titer data of the antigens involved in the six immunization schemes, where each data point represents the geometric mean titer (GMT) of 8 parallel experiments.

[0216] Table 5.1.1 Original pseudovirus titer data of 6 monovalent vaccines with homologous vaccination strategy

[0217]

[0218] Note: Due to technical limitations, the lowest measurable titer is 15. For data points with a titer close to 15, the titer is set to an unknown parameter greater than 0 and less than 15.

[0219] In order to construct a 6×6 antigen-antibody cross-neutralization matrix C, it is necessary to evaluate the titer of each vaccine against the hypothetical Alpha / Beta and BA.2 / 5 fusion antigens. For simplicity, we use the geometric mean of the GMTs against the two corresponding antigens to express the antibody titer against the fusion antigen, resulting in Table 5.1.2.

[0220] Table 5.1.2 Antigen-antibody titer matrix

[0221]

[0222] Let X = (x i,j ) 6×6 represents the above antigen-antibody titer matrix, where x i,j represents the neutralization titer of variant j-specific antibodies against variant i-specific antigens. The diagonal elements of X form a vector x=(x 11 , x 22 , ..., x 66 ), representing the neutralization titer of each variant-specific antibody against its own antigen.

[0223] Table 5.1.3 Autoantigen-antibody neutralization titers of each variant

[0224]

[0225] If the vaccine type and dose are the same for each variant, x can reflect the antigen-antibody neutralization rate of each variant. The relative ratio between the two. The Delta variant autoantigen-antibody neutralization rate γ neu As a reference, define the relative ratio For mRNA vaccines (Alpha / Beta, Delta, BA.1, BA.2 / 4 / 5, XBB.1.5) administered at a dose of 30 μg, CoronaVac inactivated vaccine The relative value of cannot be directly inferred because it is an inactivated vaccine against the prototype strain administered at a dose of 5 μg. is a vector of relative ratios, which can be calculated as follows:

[0226] Table 5.1.4 Relative ratio of neutralization rate of autoantigens and antibodies of each variant

[0227]

[0228] The autoantigen-antibody neutralization rate of each variant can be calculated by the following formula:

[0229]

[0230] where γ neu , It is an unknown parameter.

[0231] According to formula (12), the antigen-antibody cross-neutralization matrix C can be calculated by the following formula:

[0232] c i,j =x ij / x j,j

[0233] Table 5.1.5 Antigen-antibody cross-neutralization matrix

[0234]

[0235] 5.2 Estimation of other parameters using data from 28 immunization regimens

[0236] For each immunization strategy, we measured neutralizing titers of serum antibodies against 11 pseudoviruses.

[0237] To estimate the cross-neutralization coefficients of the six variant-specific antibodies against the 11 pseudoviruses, we performed the following calculations. The original titer data matrix of 11 pseudoviruses for 6 monovalent vaccine continuous vaccination schemes, including The neutralization titers of variant i-specific antibodies against pseudovirus j are shown (Table 5.2.1).

[0238] Table 5.2.1 Raw titer data of 6 monovalent vaccines against 11 pseudoviruses

[0239]

[0240] Note: Due to technical limitations, the lowest measurable titer is 15. For data points with a titer close to 15, the titer is set to an unknown parameter greater than 0 and less than 15.

[0241] Defining pseudovirus-antibody cross-neutralization coefficient is the ratio of the neutralization titer of variant i-specific antibody against pseudovirus j to its titer against self-antigen:

[0242]

[0243] where x ii is the neutralization titer of variant i-specific antibodies against self-antigens (Table 5.1.3). Pseudovirus-antibody cross-neutralization matrix It can be calculated based on this (Table 5.2.2).

[0244] Table 5.2.2 Pseudovirus-antibody cross-neutralization matrix

[0245]

[0246] Let θ be the vector of 28 unknown parameters in Table 3.1. Each vaccination strategy can be encoded into 4 vectors T, V, D, P according to the method in Section 4. Given an vaccination strategy and parameter set θ, the dynamic changes of the levels of each immune component over time can be obtained through simulation. Under the lth vaccination strategy, the level of variant-specific antibodies at time t can be expressed as:

[0247] Ab l (t,T l , V l , D l , P l ,θ)

[0248] Let t ln is the measurement time of serum antibody under the first immunization strategy (14 days after the last injection), and Ab lm (θ) is the value of the parameter θ at t lm The mutant-specific antibody level vector simulated at each moment:

[0249] Ab lm (θ)=Ab l (t lm , T l , V l , D l , P l ,θ)

[0250] Theorem: Let Ab=(Ab1,Ab2,...,Ab n ) T is an antibody mixture, and N k is the antibody neutralization titer of the mixture against pseudovirus K. Then:

[0251]

[0252] in Representation Matrix The kth row of is the vector of variant-specific neutralization rates.

[0253] Proof: Let Ka i,j Antibody j ) to antigen i(Ag i ) is the affinity constant. In equilibrium:

[0254]

[0255] Among them [Ag i :Ab j ] is the concentration of antibody-antigen complex.

[0256] NT50 titer k Defined as the serum dilution that reduces the biological effect (pseudovirus infection) by 50%. For simplicity, we assume that half of the pseudovirus is neutralized at this dilution:

[0257]

[0258] The antibody concentration is 1 / N of its initial value. k .

[0259] Equilibrium conditions:

[0260]

[0261] Solving for N k :

[0262]

[0263] According to formula (11), Ka is expressed by the cross-neutralization coefficient and variant-specific neutralization rate. k,j :

[0264]

[0265] where γ c is a constant.

[0266] Substituting in:

[0267]

[0268] The proof is complete.

[0269] For the parameter set θ, the DynaVac model predicts the neutralization titers of the l-th immunization strategy against 11 pseudoviruses as:

[0270]

[0271] in

[0272] make Model-predicted neutralization titer matrix for all 28 immunization strategies, Ab m (θ)=(Ab 1m (θ), Ab 2m (θ), ..., Ab 28m (θ)) is the model predicted value of the antibody level matrix at the serum antibody measurement time corresponding to all 28 immunization strategies.

[0273] Then we have:

[0274]

[0275] Among them J 28 =(1, 1, ..., 1) 1×28 is a row vector of all 1s, represents the Hadamard product.

[0276] Let N = (N i,l ) 11×28 Neutralization titer matrix measured experimentally.

[0277] Define the log2 fold change matrix:

[0278]

[0279] The objective function L for parameter estimation is the squared Frobenius norm of the difference between the predicted and measured log2 fold change matrices:

[0280]

[0281] The optimization problem is:

[0282]

[0283] subject to lb≤θ≤ub

[0284] Where lb and ub are the lower and upper limits of the prior interval of each parameter in Table 3.1. Solution θ *is the estimated optimal parameter set.

[0285] According to the objective function defined above, we use genetic algorithm (GA) to iteratively optimize the parameters. Genetic algorithm is a type of heuristic search algorithm inspired by Darwin's theory of natural selection and genetic inheritance. It simulates the natural evolution process, continuously optimizes the solution space, and finally converges to the optimal or approximately optimal solution. We use the 'ga' function in MATLAB's GlobalOptimization Toolbox to implement the genetic algorithm. In our implementation, each parameter set is encoded as a chromosome, and the fitness function is set to the inverse of the objective function L in formula (17) to minimize the objective function value. We use the default options provided by the 'ga' function. By fitting the titer data of the 28 immunization schemes in this study, the estimated parameter values ​​are shown in Table 3.1.

[0286] 6 Results display

[0287] 6.1 Comprehensive experimental validation in mice

[0288] The fitted DynaVac model accurately reproduces the experimentally measured neutralization titers for almost all vaccination regimens and variants ( Figure 2 Further evaluation showed strong agreement between the DynaVac model predictions and experimental observations, with a coefficient of determination (R 2 ) is 0.834( Figure 3 ).

[0289] 6.2 Typical simulation results display

[0290] The success in modeling the evolution of immunogenicity of SARS-CoV-2 variants provides an opportunity to continuously explore the space of immune strategy-variant combinations using computer simulation experiments. We simulated a vaccination regimen that included two doses of prototype vaccination (day 0, day 30) followed by two doses of variant vaccination (day 390, day 420) and systematically investigated the quantitative relationship between antigenic distance and immunoprinting-mediated suppression of variant-specific antibody responses in prime-boost immunization ( Figure 4 a).

[0291] Using simulated data, we compared the dynamic changes in the levels of antigen, antibody, naive B cells, and memory B cells against prototype and variant antigens after each vaccination at three representative antigenic distances (defined as the 1-cross neutralization ratio between prototype and variant antigens): 0.5, 0.95, and 0.99 ( Figure 4(b) After the first vaccination with the prototype vaccine, the initial B cell response was relatively high, but the level of antibody production was limited. This initial response induced the accumulation of prototype-specific memory B cells, which responded rapidly and expanded upon the second dose of the prototype vaccine. Pre-existing memory B cells were rapidly activated, rapidly proliferated and differentiated into antibody-secreting cells, resulting in rapid and robust production of prototype-specific antibodies. High levels of antibodies led to rapid clearance of antigens, thus limiting the initial B cell response. At the same time, the memory B cell population underwent a massive expansion driven by antigenic stimulation ( Figure 4 b, first column).

[0292] When the variant vaccine is administered as a booster (heterologous boost), the dynamic changes in the antibody specificity spectrum are regulated by the antigenic distance between the primary immunization antigen (prototype) and the boosting antigen (variant). Figure 4 In middle b, second column), heterologous boosting elicited strong prototype-specific memory B cell recall and correspondingly high levels of prototype-specific antibodies. Due to the cross-neutralizing inhibitory effect of prototype-specific antibodies on antigens, the induction of variant-specific naive B cells and antibodies was minimal. Therefore, the antibody spectrum elicited by antigenically similar variants is similar to that of homologous boosting.

[0293] For booster antigens with an antigenic distance of 0.95 from the primary immunization antigen, the first booster dose triggered a modest initial B cell response against the variant, resulting in a small number of variant-specific memory B cells and antibodies. However, even after two doses of heterologous booster vaccination, prototype-specific memory B cell responses still dominated due to the high efficiency of memory B cell recall, resulting in a much larger proportion of prototype-specific antibodies than variant-specific antibodies ( Figure 4 b, third column).

[0294] For booster antigens with an antigenic distance of 0.99 from the primary immunization antigen, the variant-specific initial B cell response induced by the first booster dose was almost equivalent to the initial B cell response of the primary immunization. The initial B cell response to the variant antigen was followed by a large population of variant-specific memory B cells after the first booster dose. Nevertheless, similar to the primary immunization, the strong initial B cell response did not induce a strong corresponding antibody response. Therefore, a second dose of variant booster is required to recall and expand variant-specific memory B cells, thereby significantly increasing variant-specific antibodies ( Figure 4 (b, fourth column). Interestingly, although the antibodies produced by the first booster were still mainly prototype-specific, the second booster shifted the antibody composition to mainly variant-specific antibodies. This phenomenon of delayed shift in antibody specificity after repeated booster immunization is consistent with experimental observations and further verifies the accuracy of our model.

Claims

1. A method for constructing an immune response kinetic model, comprising the following steps: 1) Establish a set of ordinary differential equations describing the dynamics of the immune response to the viral vaccine as follows: Among them, i, j, and k represent the numbers of the virus variants, and n represents the total number of virus variants involved; Ag stands for antigen, then Ag i 、Ag j 、Ag k Represent the specific antigen levels of virus variants i, j, and k, respectively; Ab stands for antibody, then Ab i ,Ab j Represent the levels of specific antibodies against virus variants i and j, respectively; N represents the initial B cells, then N i represents the level of the initial B cell population specific for the viral variant i antigen; M stands for memory B cells, M off Indicates the non-activated memory B cell population, M on indicates the activated memory B cell population, then Respectively represent the size of the non-activated memory B cell population specific to the antigen of virus variants i and j, They represent the size of activated memory B cell populations specific for the antigens of virus variants i and j, respectively; γ Aeu represents the antigen-antibody neutralization rate, then They represent the neutralization rates of specific antigens and antibodies against virus variants i and k, respectively; γ Ag represents the rate of antigen degradation, γ Ab Represents the antibody degradation rate; s N represents the maximum rate of initial B cell affinity maturation, s M represents the maximum expansion rate of memory B cells; d N represents the initial B cell decay rate, d M represents the decay rate of memory B cells; k N2M represents the maximum rate of differentiation of initial B cells into memory B cells; p N represents the maximum initial antibody production rate, p M represents the maximum memory antibody production rate; m0 represents the cross-reaction affinity threshold for memory B cell expansion; c0 represents the cross-reaction affinity threshold for memory B cells to differentiate into antibody-producing plasma cells; c i,j 、c j,i 、c i,k 、c j,k represents the cross-neutralization coefficient, expressed as c i,j For example, it is defined as the affinity constant Ka of the specific antibody of virus variant j to the specific antigen of virus variant i. i,j The affinity constant Ka of the specific antibody of the virus variant j to its own antigen j Ratio: C i,j =The i,j / The j By definition, c i,i =1, and for i≠j, C i,j ∈(0,1); Time window a N (t) Describes the duration of the initial immune response: Among them, t0 represents the vaccination time point, t ini is the time interval before the onset of the germinal center reaction, t min and t max are the shortest and longest durations of the immune response, respectively. The response lasts at least t min , at t min to max The duration between the two depends on the amount of residual antigen, and the max , the reaction stops; The specific antigen of virus variant i is expressed as function f(Ag i , K) reaches saturation: f(Ag i , K)∈[0,1) represents the surface fraction occupied by antigen relative to the maximum loading capacity of antigen-presenting cells, where K represents the amount of antigen required to occupy half of the maximum loading capacity of antigen-presenting cells; Time window a M (t) Describes the duration of the memory immune response: Among them, t0 represents the vaccination time point, t ini is the time interval before the onset of the germinal center reaction, t min and t max are the shortest and longest durations of immune response, respectively; memory immune response occurs rapidly at the vaccination time point t0; 2) Establish simulation algorithms for any vaccination strategy, including: For any vaccination strategy that only contains monovalent vaccines, it is first encoded into four m-dimensional vectors T, V, D, P, T = (t 01 , t 02 , ..., t 0m ), V=(v1,v2,...,v m ), D=(d1, d2, ..., d m ), P=(p1, p2, ..., p m ); This strategy involves n virus variants and m consecutive vaccinations, let k be an integer from 1 to m, t 0k is the time of the kth vaccination, v k is the variant number of the vaccine used for the kth vaccination, d k is the vaccination dose of the kth vaccination; p k ∈{1, 2, 3} represents the type of vaccine used in the kth vaccination, where 1 represents protein vaccine, 2 represents mRNA vaccine, and 3 represents inactivated vaccine; The initial conditions of the immune response kinetic model after the first vaccination are: The initial condition of the immune response kinetic model after the kth vaccination is: N k (t 0k )=N k-1 (t 0k ),Ab k (t 0k )=Ab k-1 (t 0k ) Among them, i represents the virus variant number, P s , P r , P v They are the antigen content per unit dose of recombinant protein vaccine, mRNA vaccine and inactivated vaccine; Ag k-1,i (t 0k ) represents the antigen level of variant i before the kth vaccination; Ag k,i (t 0k ) represents the antigen level of variant i after the kth vaccination; N k-1 (t 0k ), Ab k-1 (t 0k ), N k (t 0k ), Ab k (t 0k ) are all n-dimensional vectors, where N k-1 (t 0k ), Ab k-1 (t 0k ) represent the levels of n variant-specific non-activated memory B cells, activated memory B cells, initial memory B cells, and antibodies before the kth vaccination, respectively; N k (t 0k ), Ab k (t 0k ) represent the levels of n variant-specific non-activated memory B cells, activated memory B cells, initial memory B cells and antibodies after the kth vaccination, respectively; For vaccination strategies involving multivalent vaccines, simply replace the above m-dimensional vector V with a q×m matrix: where q represents the highest vaccine potency in the multivalent vaccination strategy, and v j,k ∈ {0, 1, 2, ..., n} represents the variant number of the j-th antigen component in the vaccine used for the k-th vaccination; if the potency q k of the vaccine used for the k-th vaccination is lower than the highest vaccine potency q in this strategy, then the missing antigen components are filled with 0, that is, for the k-th column of the matrix V, if q k < j ≤ q, then v j,k = 0; the initial conditions of the antigen levels for the first vaccination and subsequent vaccinations are adjusted respectively as follows: Among them, q k represents the potency of the vaccine used for the kth vaccination, v .k represents the kth column of the matrix V, and the initial conditions of the remaining components are consistent with the monovalent vaccination strategy; According to the above initial conditions and the ordinary differential equations of step 1), the dynamics of the humoral immune response after each vaccination can be calculated in the time period [t 0k , t 0k+1 ] to perform numerical simulation: ] is the simulation result of the kth inoculation, where are all n-dimensional vector functions of time t, representing the time period [t 0k , t 0k+1 ] The dynamic changes of specific antigens, initial memory B cells, non-activated memory B cells, activated memory B cells and antibody levels of n variants over time t; the complete simulation results of the vaccination strategy can be obtained by connecting the simulation results of each vaccination: [Ag(t),N(t),M off (t),M on (t),Ab(t)],t∈[t 01 ,t end ] Among them, t end is the simulation end time, t end >t 0m ; 3) Based on the simulation algorithm of step 2), by fitting the antibody-pseudovirus neutralization titer data measured in the experiment, an intelligent optimization algorithm is used to estimate the parameters in the ordinary differential equation group to obtain a parameterized immune response kinetic model.

2. The construction method according to claim 1, characterized in that: Step 3) Estimating the parameters of the immune response kinetic model by fitting the serum antibody-pseudovirus neutralization titration experimental data of various immunization schemes, the fitting process includes: I. Data Collection: Design a variety of vaccination schemes for experimental animals, covering both homologous and heterologous immunization strategies, and collect serum antibody-pseudovirus neutralization titer data under each scheme; II. Construction of Neutralization Titer Matrix: Using the serum antibody-pseudovirus neutralization titer data of homologous immunization, the antigen-antibody neutralization titer matrix X = (x i,j ) n×n , where x i,j represents the neutralization titer of variant j-specific antibodies against variant i-specific antigens, n represents the total number of virus variants involved in the experiment; the vector x composed of the diagonal elements of X = (x 11 , x 22 , ..., x nn ), indicating the neutralization titer of each antibody against its own antigen; III. Calculate the relative ratio of autoantigen-antibody neutralization rate: A variant strain is selected as the benchmark, which is numbered as δ, and its autoantigen-antibody neutralization rate is defined as γ neu For other variants, the relative value of the autoantigen-antibody self-neutralization rate of each variant and the variant δ is calculated based on the experimentally measured autoantigen-antibody neutralization titer data x Then construct the relative ratio vector of self-antigen-antibody neutralization rate: We further obtain the autoantigen-antibody neutralization rate vector Γ neu : IV. Construction of antigen-antibody cross-neutralization matrix: The neutralization efficiency of each variant-specific antibody against non-self antigens was calculated through the antigen-antibody neutralization titer matrix X, thereby constructing the antigen-antibody cross-neutralization matrix C. The matrix elements were calculated using the following formula: c i,j =x i,j / x j,j Among them, x ij represents the neutralization titer of antibody j against antigen i, x j,j is the neutralization titer of antibody j against its own antigen; V. Construction of pseudovirus-antibody cross-neutralization matrix: Construction of pseudovirus-antibody neutralization titer matrix based on experimental data in represents the neutralization titer of variant j-specific antibody against pseudovirus i, n′ represents the total number of pseudoviruses involved in the experiment, and n represents the total number of virus variants involved in the experiment; the pseudovirus-antibody cross-neutralization matrix is ​​constructed by the following formula The antibody-pseudovirus cross-neutralization coefficient is defined as the ratio of the neutralization titer of the antibody to any pseudovirus to the neutralization titer to the self-antigen; VI. Parameter optimization using all neutralization titer data: Assume that the experiment involves L vaccination strategies, T l , V l , D l , P l Respectively represent the vaccination time, variant number, vaccination dose and vaccine type of the first vaccination strategy encoded in step 2); take the model parameter set to be determined as θ, and simulate the algorithm in step 2) to obtain the specific antibody level vector Ab of each variant corresponding to the first vaccination strategy l , written as time t, vaccination strategy T l , V l , D l , P l And the parameter set θ as a function of the independent variable: Ab l (t,T l ,V l ,D l ,P l ,θ) Ab lm (θ) represents the serum antibody measurement time t corresponding to the lth vaccination strategy when the parameter set is θ lm The following vectors show the specific antibody levels for each variant: Ab lm (θ)=Ab l (t lm ,T l ,V l ,D l ,P l ,θ) make Represents the model prediction value of the neutralization titer matrix of serum antibodies against n' pseudoviruses corresponding to all L vaccination strategies, Ab m (θ)=(Ab 1m (θ), Ab 2m (θ), ..., Ab Lm (θ)) is the model prediction value of the antibody level matrix at the serum antibody measurement time corresponding to all L vaccination strategies, then: in, represents the pseudovirus-antibody cross-neutralization matrix, Γ neu represents the antigen-antibody self-neutralization rate vector, J L =(1, 1, ..., 1) 1×L is a row vector of all 1s, represents the matrix Hadamard product; N = (N i,l ) n′×L It represents the experimental observation value of the neutralization titer matrix of serum antibodies against n' pseudoviruses corresponding to all L vaccination strategies; Among them, rN and are the log2 fold change matrices of the neutralization titer matrices of the actual experiment and the model prediction, respectively; The objective function Loss, which measures the goodness of fit of the model, is defined as the model-predicted value of the neutralization titer The deviation measure from the experimental observation value rN is: Where θ is the model parameter vector and the subscript F represents the Frobenius norm of the matrix.

3. The construction method according to claim 2, characterized in that: In step 3), according to the defined objective function Loss, a genetic algorithm is used to iteratively optimize the parameters.

4. The construction method according to claim 3, characterized in that: The genetic algorithm is implemented using the ‘ga’ function in the Global Optimization Toolbox of MATLAB. Each parameter set is encoded as a chromosome, and the fitness function is set to the inverse of the objective function L to minimize the objective function value.

5. A method for optimizing viral vaccine components and immunization strategies based on an immune response kinetic model, wherein the immune response kinetic model is obtained by the construction method described in any one of claims 1 to 4, and the simulation algorithm in step 2) is used to simulate the multi-dose vaccination process of the viral variant vaccine to predict the corresponding immune response dynamics, and the neutralization level of the target variant induced by the immunogenicity induced by the candidate immunization strategy is evaluated by comparison, to obtain the optimal vaccine components, dosages, and schedule.

6. A readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method of optimizing viral vaccine components and immunization strategies based on an immune response kinetic model as described in claim 5.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for optimizing viral vaccine components and immunization strategies based on an immune response kinetic model as described in claim 5 is implemented.