Methods and systems for optimal vaccine design

The method optimizes epitope-based vaccine design by simulating diverse populations and minimizing the likelihood of no immune response, addressing the genetic heterogeneity challenge in existing vaccine design methods to enhance population coverage.

JP7582310B2Active Publication Date: 2024-11-13NEC CORP
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
JP2022525858
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Priority Date
2020-04-20
Filing Date
2020-06-26
Publication Date
2024-11-13
Estimated Expiration
2040-06-26

AI Technical Summary

Technical Problem

Existing epitope-based vaccine design methods fail to account for the genetic heterogeneity of human populations, leading to inadequate coverage and response variability among individuals, with traditional approaches either favoring common HLA alleles or ignoring rare alleles, resulting in vaccines that do not effectively protect all citizens.

Method used

A computer-implemented method that selects amino acid sequences for a vaccine by considering the immune profiles of a population, optimizing for maximum immune response across all HLA alleles and rare genotypes, using a 'digital twin' framework to simulate a diverse population and minimize the likelihood of no immune response through mathematical optimization.

Benefits of technology

Ensures that a higher likelihood of immune response is achieved across a diverse population, including rare genotypes, by explicitly considering the immune profiles of simulated citizens, thereby enhancing the effectiveness of epitope-based vaccines.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007582310000020
    Figure 0007582310000020
  • Figure 0007582310000021
    Figure 0007582310000021
  • Figure 0007582310000022
    Figure 0007582310000022
Patent Text Reader

Abstract

According to one aspect of the present invention, a computer-implemented method for selecting one or more amino acid sequences for inclusion in a vaccine from a set of predicted immunogenic candidate amino acid sequences includes the steps of: identifying an immune profile reactivity value for each candidate amino acid sequence for each sample component of a plurality of sample components of an immune profile, the immune profile reactivity value indicating whether the candidate amino acid sequence will generate an immune response for the sample component of the immune profile; capturing a plurality of immune profiles for a population; generating a plurality of representative immune profiles for the population, the representative immune profiles overlapping with the sample components of the immune profile; and selecting, based on the immune profile reactivity values, one or more amino acid sequences for inclusion in the vaccine that minimize the likelihood of no immune response for each representative immune profile. A computer-readable medium is also provided along with a method for making a vaccine.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to methods and systems for optimal vaccine design. [Background technology]

[0002] Epitope-based vaccines (EVs) utilize short antigen-derived peptides corresponding to immune epitopes that are administered to trigger protective humoral and / or cellular immune responses. EVs allow for tight control over immune response activation by targeting potentially most important, immunogenic and conserved antigenic regions. Experimental screening of large sets of peptides is time-consuming and costly. Therefore, in silico methods that facilitate T-cell epitope mapping of protein antigens are paramount for the development of EVs. Prediction of T-cell epitopes targets the process of peptide antigen presentation by proteins encoded by the major histocompatibility complex (MHC). Since various MHCs have different specificities and T-cell epitope repertoires, in a genetically heterogeneous human population, each individual is likely to respond to a different set of peptides from a given pathogen. Furthermore, protective immune responses are only expected if T-cell epitopes are restricted by MHC proteins that are expressed at high frequency in the target population. Therefore, without careful consideration of the specificity and prevalence of MHC proteins, EVs will not be able to adequately address the target population.

[0003] Vaccine design in the context of genetically heterogeneous human populations faces two major problems: first, each individual expressing a different set of alleles, possibly with different binding specificities, is likely to respond to a different set of peptides from a given pathogen, and second, alleles are expressed at vastly different frequencies in different ethnicities.

[0004] Computational tools can be important in addressing these issues in vaccine design. Available computational methods for T cell epitope vaccine design are primarily directed at the epitope prediction stage of MHC-binding peptides. A smaller number of tools and algorithms have been developed to guide the selection of putative epitopes and optimize the design of polypeptide vaccine structures by maximizing coverage in the target population and / or with respect to pathogen diversity.

[0005] The current state of the art is moving towards epitope-based vaccine design, specifically the problem of selecting putative epitopes, broadly classified as HLA supertype-based and allele-based (Oyarzun, P. and Kobe, B. Computer-aided design of T-cell epitope-based vaccines: addressing population coverage. International Journal of Immunogenetics, 2015, 42, 313-321).

[0006] Supertype-based methods have proven inadequate for populations with diverse HLA backgrounds by prioritizing only the most common HLA alleles (Schubert, B.; Lund, O. and Nielsen, M. Evaluation of peptide selection approaches for epitope-based vaccine design. Tissue Antigens, 2013, 82, 243-251).

[0007] Traditional allele-based approaches do not consider individual citizens when selecting elements for inclusion in the vaccine; rather, they aim to maximize the average likelihood of response for all individuals. This is problematic because the proposed approach focuses on eliciting the strongest (or most promising) possible response rather than on ensuring that each citizen is protected by the vaccine. (Vider-Shalit, T.; Raffaeli, S. and Louzoun, Y. Virus-epitope vaccine design: Informatic matching the HLA-I polymorphism to the virus genome. Molecular Immunology, 2007, 44, 1253-1261; Toussaint, NC; Donnes, P. and Kohlbacher, O. A Mathematical Framework for the Selection of an Optimal Set of Peptides for Epitope-Based Vaccines. PLOS Computational Biology, 2008, 4, e1000246; Lundegaard, C.; Buggert, M.; Karlsson, AC; Lund, O.; Perez, C. and Nielsen, M. PopCover: A Method for Selecting of Peptides with Optimal Population and Pathogen Coverage. Proceedings of the 1^st ACM International Conference on Bioinformatics and Computational Biology, 2010).

[0008] Other known methods have used graph-based approaches to design epitope vaccines, but none of these have been shown to generate optimal vaccine designs (Theiler, J. and Korber, B. Graph-based optimization of epitope coverage for vaccine antigen design. Statistics in Medicine, 2018, 37, 181-194).

[0009] Thus, there is a need to improve upon existing methods for selecting candidate elements for inclusion in vaccines. [Prior art documents] [Patent documents]

[0010] [Patent Document 1] WO2020 / 070307 [Patent Document 2] WO2017 / 186959 [Non-patent literature]

[0011] [Non-Patent Document 1] Oyarzun, P. and Kobe, B. Computer-aided design of T-cell epitope-based vaccines: addressing population coverage. International Journal of Immunogenetics, 2015, 42, 313~321 [Non-Patent Document 2] Schubert, B.; Lund, O. and Nielsen, M. Evaluation of peptide selection approaches for epitope-based vaccine design. Tissue Antigens, 2013, 82, 243~251 [Non-Patent Document 3] Vider-Shalit, T.; Raffaeli, S. and Louzoun, Y. Virus-epitope vaccine design: Informatic matching the HLA-I polymorphism to the virus genome. Molecular Immunology, 2007, 44, 1253~1261 [Non-Patent Document 4] Toussaint, NC; Donnes, P. and Kohlbacher, O. A Mathematical Framework for the Selection of an Optimal Set of Peptides for Epitope-Based Vaccines. PLOS Computational Biology, 2008, 4, e1000246 [Non-Patent Document 5] Lundegaard, C.; Buggert, M.; Karlsson, AC; Lund, O.; Perez, C. and Nielsen, M. PopCover: A Method for Selecting of Peptides with Optimal Population and Pathogen Coverage. Proceedings of the 1^st ACM International Conference on Bioinformatics and Computational Biology, 2010 [Non-Patent Document 6] Theiler, J. and Korber, B. Graph-based optimization of epitope coverage for vaccine antigen design. Statistics in Medicine, 2018, 37, 181~194 [Non-Patent Document 7] Patronov, A. and Doytchinova, I. T-cell epitope vaccine design by immunoinformatics. Open Biology, 2013, 3, 120139 [Non-Patent Document 8] Caoili, SEC Benchmarking B-Cell Epitope Prediction for the Design of Peptide-Based Vaccines: Problems and Prospects. Journal of Biomedicine and Biotechnology, 2010 [Non-Patent Document 9] Shiina, T.; Hosomichi, K.; Inoko, H. and Kulski, JK The HLA genomic loci map: expression, interaction, diversity and disease. Journal of Human Genetics, 2009, 54, 15~39 [Non-Patent Document 10] Cao, K.; JillHollenbach; Shi, X.; Shi, W.; Chopek, M. and Fernandez-Vina, MA Analysis of the frequencies of HLA-A, B, and C alleles and haplotypes in the five major ethnic groups of the United States reveals high levels of diversity in these loci and contrasting distribution patterns in these populations. Human Immunology, 2001, 62, 1009~1030 [Non-Patent Document 11] Jensen, KK; Andreatta, M.; Marcatili, P.; Buus, S.; Greenbaum, JA; Yan, Z.; Sette, A.; Peters, B. and Nielsen, M. Improved methods for predicting peptide binding affinity to MHC class II molecules. Immunology, 2018, 154, 394~406 [Non-Patent Document 12] Dawood, RM; Moustafa, RI; Abdelhafez,TH; El-Shenawy, R.; El-Abd, Y.; Bader El Din, NG; Dubuisson, J.; El Awady, MK A multiepitope peptide vaccine against HCV stimulates neutralizing humoral and persistent cellular responses in mice. BMC Infectious Diseases, 2019, 19 [Non-Patent Document 13] Woo, W.-P.; Doan, T.; Herd, KA; Netter, H.-J. and Tindle, RW Hepatitis B Surface Antigen Vector Delivers Protective Cytotoxic T-Lymphocyte Responses to Disease-Relevant Foreign Epitopes. Journal of Virology, 2006, 80, 3975~3984 [Non-Patent Document 14] Trovato, M. and De Berardinis, P. Novel antigen delivery systems. World Journal of Virology, 2015, 4, 156~168 Summary of the Invention [Means for solving the problem]

[0012] Aspects of the present invention provide methods and systems for selecting a set of candidate elements for inclusion in a vaccine such that the likelihood that every member of a population will have a positive response to the vaccine is maximized.

[0013] According to an aspect of the present invention, a computer-implemented method is provided for selecting one or more amino acid sequences for inclusion in a vaccine from a set of predicted immunogenic candidate amino acid sequences, the computer-implemented method comprising the steps of: identifying an immune profile response value for each candidate amino acid sequence for each sample component of a plurality of sample components of an immune profile, the immune profile response value indicating whether the candidate amino acid sequence produces an immune response for the sample component of the immune profile; capturing a plurality of immune profiles for a population; generating a plurality of representative immune profiles for the population, the representative immune profiles overlapping with the sample components of the immune profile; and selecting one or more amino acid sequences for inclusion in the vaccine that minimize the likelihood of no immune response for each representative immune profile based on the immune profile response value.

[0014] Advantageously, the proposed approach, in contrast to conventional approaches, explicitly considers the various components that make up immune profiles and is optimized with respect to these components, maximizing the chances that a vaccine will be successful across a given population.If the population represents the world population, this approach can be considered as aiming for an optimal universal vaccine, i.e., aiming to maximize the chances that the combination of vaccine elements included in the vaccine will generate an immune response.For example, if the sample components are multiple sample HLA alleles, the proposed approach explicitly considers all alleles and is optimized with respect to all alleles.

[0015] In essence, the method of the above aspect of the invention formulates vaccine design for a particular population as an optimization problem in which the goal is to maximize the likelihood of response for each citizen.

[0016] This technique may be considered an allele-based approach, but unlike traditional methods, this approach considers the most frequently occurring alleles in a population and considers each individual citizen, rather than seeking to obtain the average of that set. It is noted in the art that a population range corresponds to the portion of a population in which an epitope-based vaccine is theoretically effective.

[0017] The predicted immunogenic candidate amino acid sequence may be a short peptide sequence or a long peptide sequence, and the long peptide sequence may include multiple short peptide sequences. The set of predicted immunogenic candidate amino acid sequences is generally taken from a prediction engine that calculates a certain score that the peptide will cause some immune response (e.g., binding, antigen presentation, cytokine release, etc.). Examples of publicly available databases and tools that may be used for such predictions include the Immune Epitope Database (IEDB) (https: / / www.iedb.org / ), NetMHC prediction tool (http: / / www.cbs.dtu.dk / services / NetMHC / ), and NetChop prediction tool (http: / / www.cbs.dtu.dk / services / NetChop / ). Other techniques are disclosed in WO2020 / 070307 and WO2017 / 186959.

[0018] The score from the prediction engine associated with each sequence may be used to determine an immune response value. Alternatively, the immune response value may be populated from a database implemented using the data in the aforementioned publications, for example by extracting univariate response statistics.

[0019] The predicted candidate amino acid sequence or sequences may have a fixed length or may have a variable length. For example, when considering MHC Class I HLA alleles, epitope lengths of 8, 9, 10, 11, and 12 amino acids may be candidates, and when considering MHC Class II HLA alleles, each epitope is generally 15 amino acids in length. Alternatively, the candidate amino acid sequence may be a group of sequences. For example, the candidate amino acid sequences include (1) short peptide sequences, such as 9-mer amino acid sequences, (2) longer peptide sequences, such as 27-mer amino acid sequences, based on short peptide sequences and may include adjacent regions, (3) longer amino acid sequences that may include multiple short peptide sequences as well as intervening naturally occurring sequences, and (4) entire protein sequences.

[0020] Selecting one or more amino acid sequences for inclusion in the vaccine may be based on the correspondence between sample components of the immune profile and components of the immune profile present in each representative immune profile.

[0021] In some embodiments, the immune profile may include one or more immune profiles selected from the group including a set of HLA alleles, the presence (or absence) of tumor-infiltrating lymphocytes, the presence (or absence) of immune checkpoint markers such as PD1, PD-L1, or CTLA4, the presence (or absence) of hypoxia markers such as HIF-1a or BNIP3, the presence (or absence) of chemokine receptors such as CXCR4, CXCR3, and CX3CR1, and previous infection with human papillomavirus. Each of these features has been found to contribute positively or negatively to the immune response of a particular epitope or candidate vaccine element. Thus, the immune response value associated with each candidate amino acid sequence may represent the contribution of the candidate sequence to the extent to which it generates an immune response to the particular variable of interest.

[0022] In certain embodiments, the sample components of the immune profile include sample HLA alleles, whereby the immune profile reactivity value includes an HLA allele immune reactivity value for each candidate amino acid sequence for each sample HLA allele of the plurality of sample HLA alleles. The immune profile for a population may include a plurality of HLA genotypes for the population. Generating a plurality of representative immune profiles may include generating a plurality of representative sets of HLA alleles for the population. The representative set of HLA alleles may overlap with the sample HLA alleles.

[0023] The sample HLA alleles of the immune profile may be the set of alleles that occur most frequently in the population, or may be all alleles of the population. The degree of overlap between the sample HLA alleles and the representative immune profile may include (1) all sample HLA alleles occur in at least one representative immune profile, and / or (2) all HLA alleles of the representative immune profile occur in the sample HLA alleles. Preferably, at least one allele for each representative immune profile should be present in the set of sample HLA alleles. Preferably, each sample HLA allele should be present in at least one of the representative sets. Similar variation in the degree of overlap is also possible between the components of the immune profile and the representative immune profile.

[0024] In an implementation, the candidate amino acid sequences are vaccine elements, and each representative set is a simulated citizen of a given population.

[0025] The method may further comprise the step of importing the set of predicted immunogenic candidate amino acid sequences, which may be from a local memory, from a database, or from a remote data repository.

[0026] In a preferred embodiment, the generating step includes (i) creating a first distribution for a plurality of immune profiles, and (ii) sampling the first distribution to create a plurality of representative immune profiles. In an example, the immune profiles may include HLA genotypes.

[0027] More preferably, the first distribution is a distribution for a plurality of immune profiles for each region of the population.

[0028] Each region may be a population group having an ethnic population group (eg, Caucasian, African, Asian) or a geographic population group (eg, Lombardy, Wuhan).

[0029] More preferably, the first distribution is a posterior distribution over genotypes in each region based on a prior distribution and the observed genotypes in a plurality of immune profiles in each region of the population.

[0030] In some particular implementations, the first distribution is a symmetric Dirichlet distribution, in which case the method further comprises collecting all genotypes observed at least once across all regions, and the sampling comprises sampling a desired number of genotypes from each region based on the count of each genotype in the sample. An alternative to the Dirichlet may be a multivariate Gaussian distribution with a logistic function transformation.

[0031] Advantageously, the approach can take into account insufficient input data and properly account for limitations of the data sample used to populate the input database. To do so, the method preferably includes simulating a digital population based on a plurality of captured immune profiles for the population, and the step of creating a first distribution is based on the simulated population such that the step of sampling is performed on the simulated population.

[0032] Such a simulation may be viewed as creating a "digital twin" of citizens in the population present in the database, where the "digital twin" is an immune profile, which may include, for example, a set of HLA alleles and other indicators of immune response, such as previous infection with human papillomavirus. Thus, the method employs a "digital twin" framework in which a synthetic population is simulated, with respect to which optimal selection of vaccine elements is made.

[0033] For example, if the input database contains 400 people in a particular area, it may be desirable to increase the available data: the proposed statistical model can create or simulate people that match the real people in this area to create a larger number of citizens, such as 10,000 people.

[0034] The proposed model includes a measure of variance: by creating a posterior distribution over the genotypes, the variance may be proportional to the amount of genotypes in the database.

[0035] Specifically, simulating the digital population includes defining a population size and creating a second distribution for each region.

[0036] In a particular implementation, the second distribution is a Dirichlet distribution. A possible alternative to the Dirichlet distribution is a multivariate Gaussian distribution, which is then subjected to a logistic function transformation.

[0037] The proposed model emphasizes rare genotypes so that population coverage is maximized. This contrasts with existing approaches, which consider the most frequently occurring alleles to try to maximize vaccine coverage. Such approaches essentially ignore rare genotypes and are therefore inappropriate for universal vaccines, since they are useful for the majority of the population, but the vaccine does not benefit minorities. Furthermore, such approaches are biased to inherent shortcomings of the input database by considering frequently occurring alleles. For example, if there is insufficient data for a region, the alleles that frequently occur in that region will not be emphasized, and the vaccine elements selected will be inherently biased towards regions with sufficient data coverage in the input database.

[0038] Generally, representative immune profiles are generated such that the representative immune profiles maximize the range of immune profile combinations in a population.

[0039] The selecting step is generally performed to select the amino acid sequences that will result in the most likely vaccine. In a preferred implementation, the selecting step comprises applying a mathematical optimization algorithm to minimize the maximum likelihood of no immune response for each representative immune profile.

[0040] In essence, this approach aims to calculate the likelihood of no reaction for a given representative immune profile and a given set of amino acid sequences, which may be viewed as the sum of immune reactivity values ​​for sample components of the immune profile that correspond to components in the representative immune profile.

[0041] The mathematical optimization algorithm may be constrained by one or more predetermined thresholds. In embodiments, the amino acid sequences may be selected based on the particular vaccine delivery platform.

[0042] A general algorithm may have difficulty in dealing with such computational complexity, and the method may be configured to provide one or more surrogate variables for the mathematical optimization algorithm to achieve efficiency and improvement. The surrogate variables may include the log likelihood of no reaction for the representative set. In a particular preferred implementation, the variables of the mathematical optimization algorithm include (a) a binary indicator variable for each candidate amino acid sequence indicating whether or not the vaccine includes this amino acid sequence, (b) a continuous variable for each representative immune profile that provides the log likelihood of no immune reaction, (c) a continuous variable for each sample component that provides the log likelihood of no reaction, and (d) a continuous variable that provides the maximum log likelihood that none of the representative immune profiles react to the selected one or more amino acid sequences, and the mathematical optimization algorithm minimizes the continuous variable that provides the maximum log likelihood that none of the representative immune profiles react to the selected one or more amino acid sequences.

[0043] Thus, in some implementations, the immune profile may include a set of HLA alleles, and the sample components of the immune profile may include sample HLA alleles. In these embodiments, the mathematical optimization algorithm may include (a) a binary indicator variable for each candidate amino acid sequence indicating whether the candidate amino acid sequence should be included in the vaccine, (b) a continuous variable for each representative immune profile that provides the log likelihood of no immune response, (c) a continuous variable for each sample component of the immune profile that provides the log likelihood of no response, and (d) a continuous variable that provides the maximum log likelihood that none of the representative immune profiles respond to the selected one or more amino acid sequences, and the mathematical optimization algorithm minimizes the continuous variable that provides the maximum log likelihood that none of the representative immune profiles respond to the selected one or more amino acid sequences.

[0044] The objective of the mathematical optimization algorithm is to minimize the variable (d). In an embodiment, the setting of the binary variables corresponds to the optimal selection of amino acid sequences for a given population. Advantageously, the mathematical optimization algorithm is a mixed integer linear program.

[0045] Thus, because the decision is binary, ie, to include an amino acid sequence in the vaccine or not, the optimization can take advantage of programming as described above.

[0046] Preferably, the selection of amino acid sequences to include in the vaccine is not an open-ended exercise, but rather the selection is constrained in some way. Preferably, the method further comprises a step of assigning a cost to each candidate amino acid sequence, and the selecting step is constrained based on the cost assigned to each candidate amino acid sequence, whereby the one or more amino acid sequences selected have a total cost below a predetermined threshold budget.

[0047] Thus, the amount of amino acid sequence to be included in the vaccine can be selected based on the current state of the vaccine platform and vaccine delivery method selected. Additionally or alternatively, the selecting step is constrained based on the maximum amount of amino acid sequence allowed in the vaccine delivery platform.

[0048] In some cases, this may be done by assigning a cost 1 to each amino acid sequence and budgeting depending on the number of amino acid sequences that can be included in the vaccine.

[0049] In addition to being considered an allele-based approach, the proposed embodiment may also be considered a graph-based approach, in which case the method further comprises the step of creating a tripartite graph, where a first set of nodes corresponds to candidate amino acid sequences, a second set of nodes corresponds to sample components of an immune profile, and a third set of nodes corresponds to a representative immune profile for a population, the edge weights between the first set of nodes and the second set of nodes are immune response values, and the edge weights between the second set of nodes and the third set of nodes represent the correspondence between the sample components and each representative immune profile.

[0050] This implementation may therefore be viewed as a network flow problem in a graph where a minimax problem is addressed with the goal of selecting the set of vaccine elements that minimizes the log-likelihood of no reaction for each hypothetical citizen. Traditional graph-based approaches do not take into account the population HLA background.

[0051] In a preferred embodiment, the immune reactivity value is a log-likelihood based on amino acid subsequences of the candidate amino acid sequence.

[0052] The vaccine design method is applicable to any method that assigns log-likelihood values. Most short peptide prediction engines calculate some kind of score that the peptide will generate some immune response (e.g., binding, antigen presentation, cytokine release, etc.), and this score generally takes into account the specific HLA allele. In some cases, this is already a probability, and in other cases, it can be converted to a probability using a conversion function such as a logistic function. Additionally, the identifying step includes a step of selecting the best likelihood value from the likelihood values ​​for each amino acid subsequence as the immune response value.

[0053] Thus, when a candidate amino acid sequence includes multiple peptide sequences, the likelihood value can be determined based on the score for the long peptide sequence or for each short peptide sequence that falls within a longer peptide sequence.

[0054] In particularly preferred embodiments, the one or more candidate amino acid sequences are contained in one or more proteins of a coronavirus, preferably the SARS-CoV-2 virus.

[0055] Thus, this approach is suitable for providing an optimized universal vaccine design for a population of subjects for the SARS-CoV-2 virus. In an embodiment, the one or more candidate amino acid sequences may be one or more of the viral spike (S), nucleoprotein (N), membrane (M), and envelope (E) proteins, and open reading frames such as orflab. Thus, the method of the present invention may be applied to the entire viral proteome. This is particularly advantageous for identifying candidate elements for vaccine design.

[0056] The method may further comprise the step of synthesizing one or more selected amino acid sequences.

[0057] The method may further include encoding the one or more selected amino acid sequences as a corresponding DNA or RNA sequence. Additionally, the method may include incorporating the DNA or RNA sequence into the genome of a bacterial or viral delivery system to generate a vaccine.

[0058] Thus, according to one aspect of the present invention, there is provided a method of making a vaccine comprising the steps of selecting one or more amino acid sequences for inclusion in the vaccine from a set of candidate immunogenic amino acid sequences predicted by a method according to any of the above aspects, and synthesizing the one or more amino acid sequences or encoding the one or more amino acid sequences as corresponding DNA or RNA sequences and / or incorporating the DNA or RNA sequences into the genome of a bacterial or viral delivery system to generate the vaccine.

[0059] According to a further aspect of the invention, there is provided a computer-implemented method for selecting one or more amino acid sequences for inclusion in a vaccine from a set of predicted immunogenic candidate amino acid sequences, the method comprising: importing a set of predicted immunogenic candidate amino acid sequences; identifying an HLA allele immune reactivity value for each candidate amino acid sequence for each sample HLA allele of a plurality of sample HLA alleles, the HLA allele immune reactivity value indicating whether the candidate amino acid sequence will elicit an immune response for the sample HLA allele; importing a plurality of HLA genotypes for a population; generating a plurality of representative sets of HLA alleles for the population, the HLA alleles in the representative sets overlapping with the sample HLA alleles; and selecting one or more amino acid sequences for inclusion in the vaccine that minimize the likelihood of no immune response for each representative set of HLA alleles based on the HLA allele immune reactivity values ​​and the correspondence between the sample HLA alleles and the HLA alleles present in each representative set of HLA alleles.

[0060] According to a further aspect of the present invention, there is provided a system for selecting one or more amino acid sequences for inclusion in a vaccine from a set of predicted immunogenic candidate amino acid sequences, the system comprising at least one processor in communication with at least one memory device, the at least one memory device storing instructions for causing the at least one processor to perform a method according to any of the above aspects.

[0061] According to a further aspect of the present invention, there is provided a computer readable medium having stored thereon computer executable instructions for carrying out the method of any of the above aspects.

[0062] Embodiments will now be described in detail, by way of example only, with reference to the accompanying drawings, in which: [Brief description of the drawings]

[0063] [Figure 1]FIG. 2 is a schematic diagram of a tripartite graph according to an embodiment of the present invention; [Diagram 2] 1 is a high-level flowchart of the proposed method. [Diagram 3] FIG. 2 is an alternative schematic diagram of a tripartite graph according to an embodiment of the present invention. [Figure 4] FIG. 1 illustrates an exemplary output. [Diagram 5] FIG. 2 illustrates a method according to one embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0064] According to some embodiments described herein, methods and systems are proposed for selecting a small set of candidate elements for inclusion in a vaccine such that the likelihood that every member of the population will have a positive response to the vaccine is maximized. Specifically, epitope-based vaccines are targeted. A "digital twin" framework is employed where a synthetic population is simulated and an optimal selection of vaccine elements is made for the simulation.

[0065] In this document, a method and system for designing effective vaccines against SARS-CoV-2 and other infections is proposed. It is aimed at epitope-based vaccines consisting of a set of epitopes or short amino acid sequences (Patronov, A. and Doytchinova, I. T-cell epitope vaccine design by immunoinformatics. Open Biology, 2013, 3, 120139 and Caoili, SEC Benchmarking B-Cell Epitope Prediction for the Design of Peptide-Based Vaccines: Problems and Prospects. Journal of Biomedicine and Biotechnology, 2010). In detail, the system preferably selects from a set of candidate elements for inclusion in the vaccine by simulating a population of "digital twin" citizens, which in this context may include the human leukocyte antigen (HLA) profile of the citizens. The HLA profile is an important determinant of the immune response that a particular citizen can mount in response to an infection (Shiina, T.; Hosomichi, K.; Inoko, H.; and Kulski, JK The HLA genomic loci map: expression, interaction, diversity and disease. Journal of Human Genetics, 2009, 54, 15-39) and is also an important factor in determining whether a vaccine will be effective in establishing immunity in a particular individual.

[0066] This method is applicable to further embodiments in which the digital twin may contribute to the consideration of the immune profile of a population, including HLA profile and / or immune response to a particular vaccine. For example, components of such immune profile may include the presence (or absence) of tumor infiltrating lymphocytes, the presence (or absence) of immune checkpoint markers such as PD1, PD-L1, or CTLA4, the presence (or absence) of hypoxia markers such as HIF-1a or BNIP3, the presence (or absence) of chemokine receptors such as CXCR4, CXCR3, and CX3CR1, and prior infection with human papillomavirus.

[0067] In the following, a specific example of the selection of candidate elements for a vaccine is described. It should be noted that in the proposed implementation described below, all reference signs given in this specification are incorporated by reference. We propose to select a set of vaccine elements to be included in the vaccine (while considering the budget of what can be included in the vaccine) based on the HLA profile of citizens in the population.

[0068] The population may be viewed as a set C of "digital twin" citizens c, and the vaccine as a set V of vaccine elements v. The likelihood that all citizens will have a positive response to the vaccine is denoted herein as P(R = +|C, V). The goal is to design the vaccine, i.e., to select the set of vaccine elements that maximizes this probability.

number

[0069] In this setting, maximizing the probability of a positive response is equivalent to minimizing the probability of no response, so we can approach vaccine design by minimizing the probability of no response for the citizen with the highest probability of no response, P(R = -|V, c).

number

[0070] A vaccine may be considered to produce a response if at least one of its components produces a positive response. That is, the probability of no response is the joint likelihood that all components fail. j For, the probability is given as follows:

number

[0071] Note that the conditioning set of likelihoods includes V.

[0072] The initial optimization problem can be expressed as:

number

[0073] Since logarithmic functions are monotonic, the value of V that minimizes the logarithm of the function also minimizes the original function.

number

[0074] Furthermore, each citizen may be considered as an immune profile. The immune profile may include a set of HLA alleles and / or further components, as described below. Each vaccine element v i It can be hypothesized that each allele or each component of the immune profile may independently produce a response. j The allele or component for A(c j ) Therefore, the final goal is

number

[0075] In this implementation, we approach this minimax problem as a type of network flow problem, where one set of nodes corresponds to vaccine elements, one set corresponds to immune profile components (e.g., HLA alleles), and one set corresponds to citizens. The goal is to select the set of vaccine elements such that the likelihood of no response is minimized for each citizen. Figure 1 shows an overview of the problem setup.

[0076] The vaccine design process Specifically, we approach the vaccine design process in four steps, as shown in Figure 2. 1. Select a set of candidate vaccine elements for inclusion in the vaccine (S201). 2. Create a set of “digital twin” citizens for the population of interest, where the digital twin is a representative immune profile (e.g., a set of HLA alleles, S202). 3. Create a tripartite graph (S203) in which nodes correspond to vaccine elements, immune profile components (e.g., HLA alleles), and citizens, and edges correspond to associated biological terms as described below. 4. Select a set of vaccine components (given a given budget) such that the likelihood of each citizen having a positive response is maximized (or equivalently, the log-likelihood of no response for each citizen is minimized, S204).

[0077] These steps are now described in detail.

[0078] Step 1. Select a set of candidate vaccine elements.

[0079] Some of these candidate vaccine elements are selected for inclusion in the vaccine. Four examples of vaccine elements are: (1) short peptide sequences, such as 9-mer amino acid sequences, (2) longer peptide sequences, such as 27-mer amino acid sequences that are based on short peptide sequences and may include flanking regions, (3) longer amino acid sequences that may include multiple short peptide sequences as well as intervening naturally occurring sequences, and (4) entire protein sequences.

[0080] Each vaccine component v i is cost C i v while a total budget b is available for including the components in the vaccine. The budget and cost explanations depend on the vaccine platform.

[0081] Some vaccine platforms are primarily limited to a fixed number of vaccine components, in which case each cost C i v will be 1, and the budget will indicate the total number of elements that can be included.

[0082] Some other vaccine platforms are limited in the maximum length of the included elements. In this case, each cost C i v is the length of the vaccine element, and the budget is the maximum length of the element that can be included.

[0083] Step 2. Create a set of “digital twin” citizens.

[0084] The inventive approach is based on the simulation of a set of "digital twin" citizens. In this exemplary implementation, it targets vaccine elements whose effectiveness is determined in part by the HLA of each citizen. Each digital twin may therefore correspond to a set of HLA alleles (or immune profile, as further described below).

[0085] It has been found that citizens in different regions of the world tend to have different sets of HLA alleles, and furthermore, some combinations of HLA alleles are more common than others (Cao, K.; JillHollenbach; Shi, X.; Shi, W.; Chopek, M. and Fernandez-Vina, MA Analysis of the frequencies of HLA-A, B, and C alleles and haplotypes in the five major ethnic groups of the United States reveals high levels of diversity in these loci and contrasting distribution patterns in these populations. Human Immunology, 2001, 62, 1009-1030). In some implementations, these relationships can be accurately modeled using complete HLA genotypes from actual citizens, genotypes available from high quality samples in the Allele Frequency Net Database (AFND, http: / / www.allelefrequencies.net / ).

[0086] A distribution for the genotypes is generated for each region.

[0087] In detail, AFND assigns each sample to a region based on where the sample was obtained (e.g., "Europe" or "Sub-Saharan Africa"). In a first step, a posterior distribution over genotypes in each region may be created based on the observations and an uninformative (Jeffreys) prior distribution.

[0088] Specifically, all genotypes observed at least once across all regions can be collected, and each genotype can be assigned an index g. The total number of unique genotypes can be referred to as G. Second, a prior distribution on the genotypes can be specified. In some implementations, a symmetric Dirichlet distribution can be used with a concentration parameter of 0.5 because this distribution is non-informative in the information-theoretic sense and does not reflect the strong conventional idea that any particular genotype is more likely to appear in any particular region. Then, for each region, the posterior distribution on the genotypes is calculated as a Dirichlet distribution as follows: θ1,…,θ G |x1,…,x G ~Dirichlet(α1+ x1,…α G + x G ) Here, α g is the (conventional) concentration parameter for the gth genotype (always 0.5 in this case), and x g is the number of times the gth genotype is observed in this region.

[0089] This distribution can then be used to sample genotypes from the region using a two-step process. θ1,…,θ G ~Dirichlet(α1+ x1,…,α G + x G ) y1,…,y G ~Multinomial(θ1,…,θ G ;n) where n is the desired number of genotypes to sample from the region, and y1,...y G are the counts of each genotype in the sample.

[0090] Creating a set of "digital twins" of citizens

[0091] An exemplary implementation continues by creating a set of digital twin citizens using a two-step approach, where a population size p and a distribution over each domain are preferably given. Specifically, the inputs are a Dirichlet distribution over each domain as well as p (this Dirichlet is completely independent from the Dirichlet for genotypes described in the previous section).

[0092] The Dirichlet distribution over each region has one "concentration" parameter for each region, each parameter reflecting the proportion of digital twins for the population that comes from that region. As an example, the parameters can be based on the actual population of each region (e.g., https: / / www.worldometers.info / world-population / population-by-region / ). The Dirichlet parameters must be positive, but do not have to sum to one. The samples obtained from the Dirichlet distribution are categorical distributions; that is, the samples obtained from this Dirichlet distribution (and the population sizes) give a multinomial distribution. That distribution may then be sampled to determine the number of citizens in each region. Mathematically, we have the following two-step sampling process: θ1,…,θ R ~Dirichlet(α1,…,α R ) d1,…,d R ~Multinomial(θ1,…,θ R ;p) where R is the number of regions, p is the desired population size, and d1,…d R is the count of digital twins from each domain, and α1,…α R is the Dirichlet concentration parameter (given by the user).

[0093] Second, genotypes for each region are sampled using the posterior distribution over genotypes described above. The number of genotypes sampled for region r is d r is given by:

[0094] In essence, there are two Dirichlet distributions: one over the immune profile or HLA genotype (and based on the observed genotypes), while a second Dirichlet distribution is over each region (and in some implementations may be provided by the user when running the simulation).

[0095] In that case, the simulation of the ensemble is a two-step process. 1. Choose (using a second, user-defined Dirichlet) how many times to capture the digital twin from each domain. 2. Select a genotype for each digital twin based on its domain (using Dirichlet 1 based on observed data).

[0096] Step 3. Create a tripartite graph.

[0097] In this presented example, a tripartite graph may be created. The graph may represent how a particular problem may be solved, although it will of course be understood that a graph need not be created, but may simply be a representative graph. Thus, in a next step of an exemplary implementation, the vaccine elements and the digital twin may be used to build a tripartite graph that forms the basis of an optimization problem for vaccine design. The graph has three sets of nodes. 1. All candidate vaccine components identified in step 1 2. All components of the immune profile, e.g., all HLA alleles in all digital twin genotypes 3. Digital Twin of Everything

[0098] A graph may have two sets of weighted edges. 1. Each vaccine element vi to each component, e.g., HLA allele a k The weight of this edge is logP(R = -|v i ,a k), i.e. the likelihood of no response for the components derived from that particular vaccine component. Note that below, a method for calculating this value for short peptides is described. Also below, a specific method is described when the components of the immune profile are not HLA alleles. 2. Edges from each component or HLA allele to each citizen (or component in their immune profile) who has that allele in their genotype. The weight of these edges is typically 1.

[0099] Intuitively, we call the edges from the vaccine element to the allele when the vaccine element is selected (and from the allele to each patient who has that allele in that case) "active". The log-likelihood of a response for a citizen is then the sum of all active incoming edges; that is, the flow from the selected vaccine element to the citizen gives the likelihood of no response for that citizen.

number

[0100] Calculating the likelihood of no response for a given digital twin and vaccine component

[0101] Below, we present the logP(R = -|v i ,a k An exemplary method for calculating logP(R = -|v i ,a k ) is applicable to any technique for assigning values ​​of

[0102] 1. Short peptide sequences. Most short peptide prediction engines calculate some kind of score that the peptide will produce some immune response (e.g., binding, antigen presentation, cytokine release, etc.), and this score generally takes into account the specific HLA alleles (Jensen, KK; Andreatta, M.; Marcatili, P.; Buus, S.; Greenbaum, JA; Yan, Z.; Sette, A.; Peters, B. and Nielsen, M. Improved methods for predicting peptide binding affinity to MHC class II molecules. Immunology, 2018, 154, 394-406). In some cases, this is already a probability, and in other cases, it can be converted to a probability using a conversion function such as a logistic function. Below, we describe an example where the response is a score that is a response for components other than the HLA allele.

[0103] It should be noted that generally in the art, terms such as likelihood and probability are used interchangeably, and are used interchangeably herein.

[0104] Therefore, the prediction engine calculates P(R = +|v i ,a k ), where v i is a peptide, k is an allele. Then logP(R = -|v i ,a k ) = log[1 - P(R = +|v i , a k )] can be obtained.

[0105] 2. Longer peptide sequences. Longer peptide sequences may contain multiple shorter peptide sequences with different scores from the prediction engine. logP(R = -|v i ,a kIn an exemplary approach for calculating the logP(R = -|p,a k ) where p is v i It is any short peptide included in the above.

[0106] 3. Longer amino acid sequences: Longer amino acid sequences may contain many more short peptide sequences, where the same techniques can be used as for the long peptide sequences.

[0107] Step 4. Designing vaccine components

[0108] Finally, the vaccine design problem can be posed as a kind of network flow problem through the graph defined in step 3. In particular, the minimization problem can be posed as an integer linear programming (ILP) problem and can therefore be solved explicitly and optimally using known ILP solvers.

[0109] Dealing with minimax problems

[0110] As stated above, the goal is to select the set of vaccine elements that minimizes the log-likelihood of no response for each patient or individual.

[0111] The minimax problem is simplified as follows:

number

[0112] Therefore, each term on the inside of the sum is exactly the term that was calculated in step 3 as the weight for the edges in the graph.

[0113] Standard ILP solvers cannot directly solve this minimax problem, but the approach in the proposed exemplary implementation uses a set of surrogate variables to address this problem. j c Citizen Cj is defined as the log-likelihood of no response for

number

number

[0114] ILP systemization

[0115] The exemplary ILP formulation consists of three types of variables.

[0116] x i v : one binary indicator variable for each vaccine element indicating whether it should be included in the vaccine for a given population. In general, vaccine elements may be annotated with an index i.

[0117] x j c : a continuous variable for each citizen in the population giving the log-likelihood of non-response for that citizen. In general, citizens may be tagged with an indicator j.

[0118] x k α : a continuous variable for each HLA allele giving the log-likelihood of no response for that allele. In general, alleles may be tagged with an index k.

[0119] z: some continuous variable giving the maximum log-likelihood that any citizen will not respond to the vaccine (the goal is to minimize this value).

[0120] Additionally, the ILP uses the following constants:

[0121] pi , k : The log-likelihood that vaccine element vi will not produce a response for allele k.

[0122] c i v : Vaccine elements v i "Cost".

[0123] b: Maximum cost of vaccine components that can be selected.

[0124] Finally, the ILP uses the following constraints:

[0125]

number

[0126]

number

[0127]

number

[0128] z≧x j c : As above, we use z as a method for solving the minimax problem. These constraints mean that z is the minimum log-likelihood that any individual patient will respond to the vaccine.

[0129] The goal of ILP is to minimize z.

[0130] 2-value xi v The setting of variables corresponds to the optimal selection of vaccine components for a given population.

[0131] The maximum flow problem and its connections to other problems with apparently efficient solutions.

[0132] We propose a connection between the maximum flow problem and other problems that have apparently efficient solutions. It is highly relevant to several network flow problems that can be solved efficiently. The proposed optimization problem is essentially a minimum flow problem with multiple sinks (each citizen is one sink), but the goal is to minimize the flow to each individual sink rather than the flow to all sinks. In particular, a (nonlinear) "min" operator is required rather than the "sum" operator that is typically used to transform a multi-sink flow problem into a single-sink problem. Hence, efficient minimum flow formulations are not applicable in this setting.

[0133] The goal of ILP is again to minimize z.

[0134] 2-value x i v The setting of variables again corresponds to the optimal selection of vaccine components for a given population.

[0135] Immune Profile

[0136] The concepts described above and representing a set of HLA alleles for a population may be used to represent an immune profile for a population, where the immune profile may include a set of HLA alleles as well as other components, possibly, or may simply include a set of other components that represent how the vaccine components will react in that representative population.

[0137] Below are described examples of how the above-described implementations are typically adjusted for and explained in the context of a set of HLA alleles.

[0138] In these examples, various other immune profile components may be represented as central nodes in the graph. In one implementation, only discretized versions of each variable may be considered. For example, components may represent "Tumor Infiltrating Lymphocytes (TILs) Present=High" or "CTLA4 Present=Low" rather than "TILs=73.8". Similarly, human papillomavirus (HPV) may be considered to be represented as a discrete binary variable ("HPV=False"). Thus, they may still be sampled using the Dirichlet distribution already used to sample HLA for each immune profile.

[0139] Although it was noted above that the central nodes represent other components for HLA alleles, the score or degree of immune response (used as edges of the graph) may be determined differently. In a particular implementation, an immune response value can be calculated for each of the above markers by extracting the univariate response statistics of the aforementioned literature. This value may again be considered as the log-likelihood of no response. For example, assume that published statistics show that 52 patients have a "high" TIL presence, while 110 have a "low" TIL presence. This allows a distribution for TIL presence to be created. Thus, each digital twin or representative immune profile for a population (i.e., the nodes on the right side of the graph) will have values ​​for each of these profile elements in addition to HLA.

[0140] For example, if the "high" group has a probability of response of 80% and the "low" group has a probability of response of (approximately) 45%, then these numbers can be used to give an immune response value for TIL presence. Similar approaches can be used for all of the other elements of the immune profile.

[0141] When creating the graph, each immune profile element and value (e.g., "TILs present = high" or "CTLA4 present = low") may be represented as a central node, and each of these nodes is connected to the appropriate digital twin node (same as for HLA).

[0142] In some example implementations, a new node may be added to the first set of nodes in the graph (i.e., the candidate amino acid sequence), and all of these immune profile element nodes are connected to this node, with the weights becoming the calculated immune response values ​​as described above. Such a graph is shown in FIG.

[0143] In practice, this graph configuration means that the selected amino acid sequence does not "influence" immune profile elements. Nevertheless, this configuration facilitates this vaccine design to aid digital twins with poor prognosis (e.g., "TILs present = low").

[0144] Vaccine production for specific vaccine platforms

[0145] The choice of vaccine delivery platform may be important in determining the budget for how many vaccine components can be selected, the cost of each vaccine component, and ultimately, how to create the actual vaccine based on the vaccine components. Below, two specific examples of vaccine platforms and the resulting budgets, costs, and use of the selected components are provided.

[0146] In the first example, HCVp6-MAP vaccine is used. This "multiple antigen peptide" (MAP) vaccine is designed as a prophylactic vaccine for Hepatitis C virus (HCV). In the first study, the authors selected short peptides as vaccine components based on several criteria. After selection, the short peptides were synthesized using the 9-fluorenylmethoxycarbonyl method. The peptides were then dissolved in DMSO to a concentration of 10 μg / μL and stored at -20°C. Just before immunization, the peptides were diluted to the desired concentration (e.g., 800 ng peptide per μL DMSO) and kept at 4°C. The vaccine was then administered subcutaneously (Dawood, RM; Moustafa, RI; Abdelhafez,TH; El-Shenawy, R.; El-Abd, Y.; Bader El Din, NG; Dubuisson, J.; El Awady, MK A multiepitope peptide vaccine against HCV stimulates neutralizing humoral and persistent cellular responses in mice. BMC Infectious Diseases, 2019, 19).

[0147] When the HCVp6-MAP vaccine is mapped onto this vaccine design problem, where each vaccine element is a short peptide, the total budget is 6 and the cost of each vaccine element is 1. To produce the vaccine, the selected vaccine elements can be processed as described above.

[0148] As a second example, consider a chimeric Hepatitis B surface antigen (HBsAg) DNA vaccine (Woo, W.-P.; Doan, T.; Herd, KA; Netter, H.-J. and Tindle, RW Hepatitis B Surface Antigen Vector Delivers Protective Cytotoxic T-Lymphocyte Responses to Disease-Relevant Foreign Epitopes. Journal of Virology, 2006, 80, 3975-3984). Roughly speaking, this vaccine platform replaces two peptide sequences in the HBsAg small envelope protein with vaccine elements. To ensure the immunogenicity of the molecule, the total length of the replaced vaccine element must be approximately 36 amino acids (Trovato, M. and De Berardinis, P. Novel antigen delivery systems. World Journal of Virology, 2015, 4, 156-168). For the present vaccine design scheme, the total budget is 36, and the cost of each vaccine element is the length of that element in amino acids. Further details are known in the art regarding the technical details of the synthesis of DNA-based vaccines after the vaccine elements have been selected (Woo, W.-P.; Doan, T.; Herd, KA; Netter, H.-J. and Tindle, RW Hepatitis B Surface Antigen Vector Delivers Protective Cytotoxic T-Lymphocyte Responses to Disease-Relevant Foreign Epitopes. Journal of Virology, 2006, 80, 3975-3984).

[0149] In overview, the proposed approach includes the following steps: 1. Select a set of candidate vaccine elements for inclusion in the vaccine. 2. Create a set of “digital twin” citizens for the population of interest. The digital twin is a set of HLA alleles or immune profile. 3. Create a tripartite graph in which nodes correspond to vaccine elements, HLA alleles (or parts of immune profiles), and citizens, and edges correspond to the relevant biological terms described below. 4. Choose the set of vaccine components (given a given budget) such that the likelihood of each citizen having a positive response is maximized (or equivalently, the log-likelihood of no response for each citizen is minimized).

[0150] Example implementations of the present invention have particular application for selecting peptide sequences for use in prophylactic vaccines against SARS-CoV-2.

[0151] A specific exemplary implementation will now be described with reference to FIG. 5. In step S501, the method identifies an immune profile response value for each candidate amino acid sequence for each sample component of a plurality of sample components of an immune profile. The immune profile response value represents whether the candidate amino acid sequence produces an immune response for the sample component of the immune profile. In step S502, the method captures a plurality of immune profiles for the population. In step S503, the method generates a plurality of representative immune profiles for the population. The representative immune profiles overlap with the sample components of the immune profile. Finally, in step S504, the method selects one or more amino acid sequences for inclusion in the vaccine that minimize the likelihood of no immune response for each representative immune profile based on the immune profile response value.

[0152] Working Example Below are presented examples of implementations of the above processes and concepts.

[0153] Graph-based "digital twin" optimization prioritizes epitope hotspots to select a universal blueprint for vaccine design.

[0154] Developing a blueprint for a viable universal vaccine against SARS-CoV-2 requires prioritizing the selection of 1) a broad portion of the human population with a given fidelity and 2) a much smaller number of regions (the exact number may depend on the bin size and the vaccine platform under consideration). It is therefore necessary to identify optimal constellations of hotspots or relevant feasible segments that can deliver a limited targeted vaccine "payload" to a broad range of subjects in the human population. To this end, we developed and applied a "digital twin" method, which models the specific HLA background of each distinct geographic population. A graph-based mathematical optimization technique is then used to select the optimal combination of immunogenic epitope hotspots that induce immunity in a broad range of human populations. An example output from the analysis is shown in Figure 3. The output shows a subset of identified hotspots that may be combined to simulate a robust immune response in the world population.

[0155] Graph-Based Optimization in Digital Twin Simulation of Epitope Hotspots

[0156] Consider the population as a set C of "digital twin" citizens c, and the vaccine as a set V of vaccine elements v. We denote the likelihood that every citizen has a positive response to the vaccine as P(R = +|C, V). Our goal is to design the vaccine, i.e., select the set of vaccine elements, to maximize this probability.

number

[0157] In this setting, maximizing the probability of a positive response is equivalent to minimizing the probability of no response. Therefore, we choose the highest probability of no response P(R = -|V, c jWe approach vaccine design by minimizing the probability of non-response for citizens with the virus.

number

[0158] In the present invention, a vaccine is considered to produce a response if at least one of the vaccine's components produces a positive response. That is, the probability of no response is the joint likelihood that all components fail. j For , this probability is given by:

number

[0159] In that case, the initial optimization problem can be expressed as:

number

[0160] Since logarithmic functions are monotonic, the value of V that minimizes the logarithm of the function also minimizes the original function.

number

[0161] Furthermore, in the present invention, each citizen is considered as a set of HLA alleles, and each vaccine element v i Suppose that each allele can independently produce a response, and citizen c j The allele for A(c j Therefore, the final object of the present invention is as follows:

number

[0162] We approach this minimax problem as a kind of network flow problem, where one set of nodes corresponds to vaccine elements, one set to HLA alleles, and one set to citizens. The goal is to select the set of vaccine elements such that the likelihood of non-response is minimized for each citizen.

[0163] The vaccine design process

[0164] Specifically, the present invention approaches the vaccine design process in four steps. 1. Select a set of candidate vaccine elements for inclusion in the vaccine. 2. Create a set of "digital twin" citizens for the population of interest. The digital twin is a set of HLA alleles. 3. Construct a tripartite graph whose nodes correspond to vaccine elements, HLA alleles, and citizens, and whose edges correspond to the associated biological terms described below. 4. Choose the set of vaccine elements such that the likelihood of each citizen having a positive response is maximized (or equivalently, the log-likelihood of no response for each citizen is minimized). [Explanation of symbols]

[0165] a k Allergy Budget c j citizen p Short peptide v i Vaccine elements z Maximum log-likelihood

Claims

1. 1. A computer-implemented method for selecting one or more amino acid sequences for inclusion in a vaccine from a set of predicted immunogenic candidate amino acid sequences, comprising: determining an immune profile reactivity value for each candidate amino acid sequence for each sample component of a plurality of sample components of an immune profile, the immune profile reactivity value indicating whether the candidate amino acid sequence generates an immune response for the sample component of the immune profile; capturing a plurality of immune profiles for a population; generating a plurality of representative immune profiles for the population, wherein the representative immune profiles overlap with the sample components of an immune profile; selecting one or more amino acid sequences for inclusion in the vaccine based on the immune profile response values, which minimizes the likelihood of no immune response for each representative immune profile; A computer-implemented method comprising:

2. The generating step includes: (i) generating a first distribution for said plurality of immune profiles; (ii) sampling the first distribution to generate the plurality of representative immune profiles; 2. The computer-implemented method of claim 1, comprising:

3. 3. The computer-implemented method of claim 2, wherein the first distribution is a distribution for the plurality of immune profiles for each region of the population.

4. 4. The computer-implemented method of claim 3, wherein the first distribution is a posterior distribution for genotypes in each region based on a prior distribution and observed genotypes in the plurality of immune profiles in each region of the population.

5. 5. The computer-implemented method of claim 4, wherein the first distribution is a symmetric Dirichlet distribution, the method further comprising collecting all genotypes that are observed at least once across all regions, and the sampling comprises sampling a desired number of genotypes from each region based on a count of each genotype in the sample component.

6. 6. The computer-implemented method of any one of claims 3 to 5, further comprising simulating a digital population based on the captured plurality of immune profiles for the population, wherein the creating a first distribution is based on the simulated population such that the sampling step is performed on the distribution of the simulated population.

7. simulating a digital population, Defining a population size; generating a second distribution for the region; 7. The computer-implemented method of claim 6, comprising:

8. 8. The computer-implemented method of claim 7, wherein the second distribution is a Dirichlet distribution.

9. 9. The computer-implemented method of any one of claims 1 to 8, wherein the representative immune profile is generated such that the representative immune profile maximizes the range of immune profile combinations in the population.

10. 10. The computer-implemented method of any one of claims 1 to 9, wherein the selecting step comprises applying a mathematical optimization algorithm to minimize the maximum likelihood of no immune response for each of the representative immune profiles.

11. the immune profile comprises a set of HLA alleles, the sample component of the immune profile comprises sample HLA alleles, and the variables of the mathematical optimization algorithm are: (a) a binary indicator variable for each candidate amino acid sequence indicating whether the candidate amino acid sequence should be included in the vaccine; (b) a continuous variable for each representative immune profile giving the log-likelihood of no immune response; (c) a continuous variable for each sample component of the immune profile giving the log-likelihood of no response; and (d) a continuous variable that gives the maximum log-likelihood that no representative immune profile reacts to the selected one or more amino acid sequences; Including, The computer-implemented method of claim 10, wherein the mathematical optimization algorithm minimizes the continuous variables that give the maximum log-likelihood that no representative immune profile responds to the selected one or more amino acid sequences.

12. 12. The computer-implemented method of claim 10 or 11, wherein the mathematical optimization algorithm is a mixed integer linear program.

13. further comprising the step of assigning a cost to each candidate amino acid sequence; 13. The computer-implemented method of any one of claims 1 to 12, wherein the selecting step is constrained based on the cost assigned to each candidate amino acid sequence such that the total cost is below a predetermined threshold budget.

14. 14. The computer-implemented method of claim 1, wherein the selecting step is constrained based on a maximum amount of amino acid sequences allowed in the vaccine delivery platform.

15. Creating a tripartite graph, a first set of nodes corresponds to the candidate amino acid sequence; a second set of nodes corresponding to said sample components of an immune profile; a third set of nodes represents the representative immune profile for the population; a weight of an edge between the first set of nodes and the second set of nodes is an immune response value; 15. The computer-implemented method of claim 1, further comprising the step of: weights of edges between the second set of nodes and the third set of nodes representing correspondences between the sample components and their respective representative immune profiles.

16. 16. The computer-implemented method of claim 15, wherein the immune reactivity value is a log-likelihood based on amino acid subsequences of the candidate amino acid sequence.

17. 17. The computer-implemented method of claim 15 or 16, wherein the identifying step comprises selecting a best likelihood value from the likelihood values ​​for each amino acid subsequence as the immune response value.

18. 18. The computer-implemented method of any one of claims 1 to 17, wherein the one or more candidate amino acid sequences are contained in one or more proteins of a coronavirus, preferably a SARS-CoV-2 virus.

19. 19. The computer-implemented method of any one of claims 1 to 18, wherein the representative immune profile may comprise one or more immune profiles selected from the group comprising a set of HLA alleles, the presence of tumor infiltrating lymphocytes, the presence of immune checkpoint markers, the presence of hypoxia markers, the presence of chemokine receptors, and prior infection with human papillomavirus.

20. 20. The computer-implemented method of any one of claims 1 to 19, wherein the step of selecting the one or more amino acid sequences for inclusion in the vaccine is further based on a correspondence between the sample components of an immune profile and the respective representative immune profiles.

21. 21. A system for selecting one or more amino acid sequences for inclusion in a vaccine from a set of predicted immunogenic candidate amino acid sequences, comprising at least one processor in communication with at least one memory device, the at least one memory device storing instructions for causing the at least one processor to perform a method according to any one of claims 1 to 20.

22. A computer readable medium having stored thereon computer executable instructions for carrying out the method of any one of claims 1 to 20.

Citation Information

Patent Citations

  • Machine learning algorithm for identifying peptides that contain features positively associated with natural endogenous or exogenous cellular processing, transportation and major histocompatibility complex (MHC) presentation

    WO2017186959A1

  • Method and system for binding affinity prediction and method of generating a candidate protein-binding peptide

    WO2020070307A1