A system and method for selecting multivalent tumor neoantigen combinations

Abstract

The present invention provides a system for efficiently selecting a combination of multivalent tumor neoantigens, which is intended to input multiple tumor neoantigen peptide sequences and screen out a combination with the most optimized protein sequence from a random combination peptide pool composed of these sequences. The system includes three key modules: a library construction module, an optimal path selection module, and a linker addition module. The library construction module is responsible for generating a library containing all possible combinations of input peptide sequences and calculating the potential utility of each combination. The optimal path selection module is further responsible for screening out the best protein sequence combination from these combinations to ensure that the association between peptides is both accurate and diverse. This module selects the best path for a multivalent tumor neoantigen vaccine sequence that can be generated with the least linkers, avoids the generation of unexpected neoantigens, and has a loose structure through precise screening. Finally, the linker addition module accurately adds linkers at the desired position, aiming to form a complete protein sequence that is stable and flexible in structure, while minimizing the use of linkers and optimizing the design of the overall sequence. Through this innovative method, the present invention generates a multivalent tumor neoantigen combination sequence that not only uses fewer linkers and has a simple structure, but also minimizes the generation of unexpected neoantigens, effectively addressing the challenges of multivalent antigen design.

Description

Technical Field

[0001] The present invention relates to the technical field of tumor immunotherapy, and in particular to a system and method for selecting a combination of multivalent tumor neoantigens. Background Art

[0002] Neoantigens are substances that are specifically expressed in tumor cells and can trigger strong and specific anti-tumor immune responses, making them an emerging method of efficient and safe immunotherapy. Compared with previous tumor-associated antigens, tumor-specific neoantigens have unique advantages in designing cancer vaccines. They are effective and can avoid triggering autoimmune responses. In the latest developments in this field, an emerging form of vaccine, namely multivalent neoantigen vaccines, has emerged. The therapeutic effects of multivalent neoantigen vaccines have been confirmed and have become a powerful strategy to enhance the immunogenicity of subunit vaccine neoantigens. Compared with traditional monovalent neoantigen vaccines, multivalent neoantigen vaccines have stronger immunogenicity and the ability to induce broad-spectrum immune protection, and have significant advantages in the field of tumor immunotherapy. In order to give full play to the therapeutic effect of multivalent neoantigen vaccines, researchers have started from optimizing adjuvants, vaccination plans, delivery methods, and the prediction and selection of neoantigens to achieve better therapeutic effects.

[0003] In the development of multivalent neoantigen vaccines, the random combination of neoantigen peptides may generate a large number of protein sequences. At present, there is a lack of effective methods to screen the optimal protein sequence from a large number of options. Studies have shown that traditional linkers may reduce the immunogenicity of vaccines. In addition, the presence of linkers may lead to an increase in repetitive sequences, affecting the efficiency of sequence synthesis. Complex protein structures may also need to rely on larger-scale proteasomes for effective processing. The appearance of unexpected neoantigens or functional domains in vaccine sequences may trigger unpredictable immune responses or interact with other proteins, leading to adverse reactions. The presence of functional fragments may also interfere with the processing of sequences by proteasomes, resulting in the inability to effectively produce the required neoantigen short peptides. Therefore, the ideal protein sequence should minimize the number of linkers, maintain a simple protein structure, and reduce the generation of unexpected neoantigens, which will inspire us to screen out the optimal protein sequence from a large number of protein sequences generated by random combinations of many neoantigens. Summary of the invention

[0004] The present invention provides a system and method for selecting a combination of multivalent tumor neoantigens, which screens the optimal protein sequence from a sequence pool generated by random combinations of neoantigen peptides. It has the advantages of comprehensiveness, reliability, flexibility, high efficiency and wide applicability, and can effectively optimize vaccine sequence design and provide beneficial effects for immunotherapy.

[0005] In order to achieve the above object, the present invention provides a system for selecting a multivalent tumor neoantigen combination, characterized in that it comprises the following steps:

[0006] A library construction module is used to calculate the optional libraries corresponding to the front and back ends of each neoantigen peptide sequence in the input list;

[0007] The optimal path selection module is used to determine the optimal ordering of peptide sequences in the input list. The target peptide sequences with the minimized objective function value are added to the path one by one in an iterative manner. The final path is the selected optimal path, which represents the optimal ordering of peptide sequences in the input list.

[0008] The linker adding module is used to check the part of the best-ranked sequence that needs to add a linker, and screen out suitable linkers from the linker library to ensure that the best-ranked sequence does not produce new antigens.

[0009] Furthermore, the library construction module uses the NetMHCpan algorithm to identify new antigens, sets thresholds for EL rank and BA rank parameters, and obtains a union of new antigens that simultaneously meet the parameter setting thresholds.

[0010] Furthermore, the library construction module uses the MHCflurry algorithm to identify new antigens, sets affinity parameter thresholds, and obtains a union of new antigens that simultaneously meet the parameter setting thresholds.

[0011] Furthermore, the library construction module uses the NetChop algorithm to identify new antigens, sets a proteasome cleavage probability threshold, and obtains a union of new antigens that simultaneously meet the parameter setting threshold.

[0012] Furthermore, the optimal path selection module performs the following operations:

[0013] If there is only one alternative peptide in the alternative library associated with the peptide, the peptide will be combined with the only alternative peptide and form the starting point of the pathway by direct connection;

[0014] If the optional library associated with each peptide segment includes multiple optional peptide segments, then all optional libraries will be statistically analyzed and sorted to determine the peptide segment associated with the optional library containing the least optional peptide segments as the starting point. Further, the optimal path selection module performs the following operations:

[0015] Based on the custom objective function, the target peptides with the minimized custom function value are gradually added to the path in an iterative manner. Further, the custom objective function is: l*num+domain+looseness, where num is the number of optional peptides contained in the optional library associated with the peptide, which is an important parameter associated with minimizing the linker, domain is the predicted number of domains, and looseness is a custom function based on the secondary structure software GOR4.

[0016] Furthermore, the looseness custom function is:

[0017] Looseness = (spiral + fold - random curl).

[0018] Furthermore, the number of domains is obtained based on HMMER software.

[0019] Furthermore, the linker adding module checks whether there is a part in the peptide sequence that needs to add a linker, and selects a suitable linker from the linker library to ensure that no unexpected antigens will be generated after being added to the sequence.

[0020] In another aspect, the present invention also provides a method for selecting a multivalent tumor neoantigen combination, comprising:

[0021] Library construction, calculating the optional libraries corresponding to the front and back ends of each new antigen peptide sequence in the input list;

[0022] The optimal path selection is to add the target peptide sequences with minimized objective function values ​​to the path one by one in an iterative manner to determine the optimal order of the peptide sequences in the input list; add linkers, check the parts of the optimally ordered sequences that need to add linkers, and screen out suitable linkers from the linker library to ensure that the optimally ordered sequences do not produce new antigens.

[0023] The present invention has the following technical effects and improvements

[0024] 1. Accuracy and reliability: The present invention consists of three modules, including library construction, optimal path selection and linker addition. Each module is closely linked and logically tight, ensuring the accuracy of the vaccine sequence design process. It combines relevant databases and advanced algorithms, uses prediction tools such as NetMHCpan, NetChop and MHCflurry to reliably predict new antigens, uses tools such as HMMER and GOR4 to evaluate protein quality, ensures that the generated sequences have good domain composition and secondary structure, and ensures the accuracy of the design results.

[0025] 2. Optimization design: The optimal path selection module uses a greedy algorithm as the core strategy to accurately screen the best path at each step by balancing factors such as linkers and structural parameters. This can generate a tumor multivalent neoantigen vaccine sequence with minimal linkers, no unexpected neoantigens, and a loose structure. Through this optimized design, the immunogenicity and stability of the vaccine sequence can be improved.

[0026] 3. Flexibility and customization: The present invention allows users to customize trade-offs as needed. Users can define thresholds for new antigen predictions to balance the number of unexpected new antigens and the number of linkers. This allows the rigor of vaccine sequence design to be adjusted according to specific needs.

[0027] 4. High efficiency: The present invention runs on computers of various configurations, showing universality, fast processing speed and low resource consumption. It uses multi-threading technology to improve the efficiency of the tool, thereby providing an efficient vaccine sequence design solution.

[0028] 5. Wide applicability: neoDesign is applicable to different types of tumor vaccines, including mRNA, DNA, dendritic cells (DCs), and peptide vaccines, etc. It provides a timely and promising solution for personalized cancer immunotherapy, and provides a comprehensive strategy for optimizing tumor multivalent neoantigen vaccine sequences, minimizing linkers, and improving immunogenicity and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0030] Figure 1 This is an architectural diagram of a system for selecting a multivalent tumor neoantigen combination according to an embodiment of the present invention.

[0031] Figure 2 The figure is a flowchart of a library construction module according to an embodiment of the present invention.

[0032] Figure 3 The figure is a flow chart of an optimal path selection module according to an embodiment of the present invention.

[0033] Figure 4 A is a distribution diagram of the number of linkers and peptide segments in one embodiment of the present invention.

[0034] Figure 4 B is an area diagram of the structural domain parameters of an embodiment of the present invention.

[0035] Figure 4 C is an area graph of the looseness parameter according to an embodiment of the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0038] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0039] The present invention proposes a system for selecting a combination of multivalent tumor neoantigens for the design of multivalent neoantigen vaccine sequences. The present invention selects the best combination from a random combination of multiple neoantigen peptides by avoiding unexpected neoantigens, reducing the number of linkers, and simplifying the overall spatial structure, so that the vaccine is easy to produce and achieves the best immune effect and safety. The process of optimizing the design of multivalent tumor neoantigen vaccines is simplified, making it more convenient to select ideal sequences from complex combinations. The present invention also has great potential in the development of diversified multivalent tumor neoantigen vaccines, making an important contribution to the development of immunotherapy strategies.

[0040] The present invention uses a series of advanced computational tools and algorithms. HMMER is used to predict common domains in proteins to help determine the functional regions of proteins, and the GOR4 algorithm is used for accurate protein secondary structure prediction to evaluate the protein secondary structure of multiple sequences. In order to predict neoantigens, advanced algorithms such as NetMHCpan, NetChop, and MHCflurry are used, which can accurately predict the presence of neoantigens. In the process of determining the protein sequence, the greedy algorithm is used as the core strategy to determine the optimal path and protein sequence selection. Ultimately, the present invention is able to screen out ideal combinations from random combinations of neoantigen peptides and generate optimal protein sequences. This technical route provides an innovative and efficient solution for the selection and design of vaccine sequences.

[0041] Example 1

[0042] Figure 1 This is a schematic diagram of a system for selecting a multivalent tumor neoantigen combination according to Example 1 of the present invention. Figure 1As shown, a system for selecting multivalent tumor neoantigen combinations of the present invention includes a library construction module, an optimal pathway selection module and a linker addition module.

[0043] The library construction module is responsible for calculating the optional libraries for each neoantigen peptide sequence in the input list. These constructed libraries are then used in the optimal path selection module. The optimal path selection module selects the peptide associated with the optional library containing the least optional peptides as the starting point. This module combines the HMMER and GOR4 methods and introduces a parameter called "l" to balance the structural parameters and minimize the linker-related parameters. In order to make the best path decision at each step, we customized an objective function (l*num+domain+looseness). Starting from the starting point, peptide sequences are gradually added based on this customized objective function as the decision criterion. The corresponding peptide sequences with the minimized customized objective function value are selected and added. This process is repeated until all peptide sequences in the input list are included, and the optimal path and the best sequence are finally determined. Finally, in the linker addition module, the position where the linker needs to be inserted is determined based on the obtained optimal sequence, and the linker that will not produce unexpected neoantigens after being added to the sequence is selected from the existing linker library. The goal is to avoid the generation of unexpected neoantigens.

[0044] In a specific embodiment, the input data of the present invention is a series of high-quality new epitopes for designing new antigen vaccines. Typically, each new antigen vaccine contains about 20 to 30 different peptides. These peptides can come from mutant sequences selected by whole exome sequencing (WES) technology and prediction algorithms, or they can be based on mass spectrometry technology. Once high-quality new antigen peptides are collected, they will be stored as a complete library for subsequent analysis.

[0045] In order to add a linker when needed, a linker library is also established in this embodiment. Based on existing research, linkers usually meet the following criteria: 1) The length is usually between 10 and 15 amino acids. 2) Linkers are usually composed of non-polar hydrophobic amino acids, such as glycine (Gly), serine (Ser) and proline (Pro). 3) Linkers are usually composed of repeating units, such as GGGGS, GS or GPG and S [24, 25]. Therefore, the linker library of this embodiment contains linker sequences that meet the above characteristics, such as GGGGGGSGGS, GSGSGSGSGS, etc.

[0046] In a specific embodiment, the library construction module inputs a series of antigen peptide sequences, such as Figure 2As shown in Figure 1, the library construction module calculates the front-end and back-end optional libraries for each peptide in the input list. Among them, the front-end optional library of each peptide sequence contains other peptide sequences from the new antigen peptide list, which can be connected to the specified peptide in pairs at the front end and meet specific conditions. Similarly, the back-end optional library consists of other peptide sequences that can be connected to the specified peptide in pairs at the back end and meet certain conditions. The specific condition is that the specified peptide sequence (i.e., the sequence corresponding to each library) and any other peptide sequence do not produce unexpected new antigens after pairing without adding a linker. The unexpected new antigens mentioned here refer to the new antigens predicted by NetMHCpan-EL (version 4.1), NetMHCpan-BA (version 4.1), NetChop (version 3.1) and MHCflurry (version 2.0) at the junction of the connecting peptides, rather than inside the respective peptides. The peptide sequences that are paired with the specified peptide sequence and meet the above specific conditions will be added to the optional library of the specified peptide. The final output is the front-end and back-end optional libraries corresponding to each specific peptide sequence.

[0047] This approach enabled us to construct a comprehensive library of options for each peptide sequence, facilitating the optimal pathway selection and linker addition steps in the present process.

[0048] In a specific embodiment, Figure 3 As shown, the input of the optimal path selection module includes the front-end optional library and the back-end optional library corresponding to all peptides. Each optional library is statistically analyzed and evaluated and ranked according to the number of optional peptides they contain. During the selection process, peptide sequences are selected based on the statistical analysis of all optional libraries. If the optional library corresponding to a peptide sequence contains only one optional peptide, it means that the specified sequence can only be combined with a unique peptide sequence. In this case, the two sequences are preferentially selected and directly connected to form the starting point of the path.

[0049] Alternatively, the peptide sequence with the least number of optional peptides in the corresponding optional library is selected as the starting point. Importantly, the parity of the number of neoantigen peptide sequences in the input is taken into account when determining the starting point. If the number is odd, the starting point is determined to connect three peptides; if it is even, the starting point is determined to connect two peptides.

[0050] After the starting point is determined, a greedy algorithm is used to select peptides at each step of the path. The selection decision is based on a custom objective function: l*num+domain+looseness. Among them, the parameter "num" represents the number of optional peptides contained in the peptide selection library and is an important parameter related to minimizing the linker. The parameter "l" is used to balance the structural parameters and minimize the linker-related parameters, and its value depends on the preference for minimizing the linker or a looser structure. The default and recommended "l" input is 1, but larger values ​​can be used to prioritize the case of minimizing the linker, while smaller values ​​can emphasize structural factors. The looseness parameter is a custom function based on the predicted secondary structure software GOR4, which calculates the content of α-helix, β-sheet and coil. The parameter "domain" is the number of domains predicted by HMMER.

[0051] Peptides are iteratively added to both ends of the sequence, and each iteration selects peptides that meet the minimum value of the objective function. In order to ensure the optimal selection, there is a specific order in the peptide sequences selected in each iteration, and the peptides corresponding to the optional library containing a smaller number of optional peptides are given priority. This method ensures flexibility in late iterative selection and helps minimize the generation of linkers and unexpected new antigens in the entire sequence. Repeat this step until there are no peptides remaining in the input list of new antigen peptides, thereby obtaining the optimal path and generating the final optimal sequence.

[0052] Although we have minimized the generation of unexpected neoantigens that can be avoided without adding linkers during the optimal pathway selection stage, in some regions of the sequence, unexpected neoantigens may still be generated if linkers are not added.

[0053] To solve this problem, the linker addition module checks whether there are parts in the sequence that need to add linkers, and selects suitable linkers from the linker library. The selection of these linkers ensures that after they are added, the sequence will not produce unexpected new antigens. This process is integrated into the entire process, and the final output is an optimal sequence with as few linkers and unexpected new antigens as possible and a relatively loose structure.

[0054] Example 2

[0055] In one embodiment, 20 new antigen peptides were used as the content of the input file for testing. Through the screening of the present invention's process for selecting a combination of multivalent tumor new antigens, an ideal final sequence was finally obtained, which contained the least number of linkers while ensuring that there were no unexpected new antigens. Structurally, the sequence structure was loose, which was consistent with the design intent of this system.

[0056] In order to evaluate the universality of this system, this embodiment is experimented on two servers with different configurations. Server 1 is equipped with a 4-core Intel Xeon Gold 6133 CPU with a frequency of 2.50GHz, which is suitable for processing medium workloads. Server 2 is equipped with a 40-core Intel Xeon CPU E5-2630 v4 with a frequency of 2.20GHz, which is suitable for processing heavy workloads that require a lot of parallel processing.

[0057] By evaluating the system on servers with different performances, its universality can be judged. Server 2 represents the performance under optimal conditions, while Server 1 represents the performance under average conditions. On Server 1, the system program has relatively low resource consumption, occupying only 0.1% of CPU resources and 0.1% of memory resources, and the running time is about 40 minutes. On Server 2, the program consumes 0.2% of CPU resources and 0.0% of memory resources, and the running time is about 20 minutes, which is faster than the running time on Server 1.

[0058] These results demonstrate the universality of this system and its applicability to scenarios with different performance levels. In addition, regardless of the server performance, this system exhibits fast processing speed and minimal resource consumption. It is worth noting that all performance evaluations were performed with the program's 'cpu' parameter set to 2. When the 'cpu' parameter is set higher, the corresponding CPU resource consumption will increase and the execution speed will be faster.

[0059] Example 3

[0060] In one embodiment, the system is developed using Python and Shell languages, and relies on some Python packages and native functions, such as GOR4, MHCflurry2.0, NetMHCpan4.1, NetChop3.1, and HMMER (>3.4). Before starting to use this method, it is necessary to complete the installation of basic packages in advance.

[0061] The main commands of this system are: python main.py[OPTIONS]

[0062] The available options are:

[0063] -p<peptides file> : Neoantigen peptide file.

[0064] -l <parameter>: Used to balance structural parameters and minimize linker-related parameters (recommended to be 1).

[0065] -s <species>: Species (only human and mouse are supported).

[0066] -e <cutoff>: Threshold for domain prediction (recommended as 1e-5).

[0067] -e1 <cutoff1>: The threshold of netMHCpanEL is recommended to be 0.05.

[0068] -e2 <cutoff2>: The threshold of netMHCpanBA is recommended to be 0.05.

[0069] -e3 <cutoff3>: The threshold of MHCflurry is recommended to be 0.05.

[0070] -e4 <cutoff4>: NetChop threshold, recommended value is 0.5.

[0071] -c <cpu>: Number of CPU cores.

[0072] -d <path>: Pfam catalogue.

[0073] -d2 <path>: NetMHCpan4.1 directory. -d3 <path>: NetChop3.1 directory.

[0074] Example usage: python main.py -p MC38.txt -l 1 -s mouse -e 1e-5 -c 2 -d / home / neoDesign / Pfam-A.hmm-d2 / home / neoDesign / netMHCpan-4.1 / netMHCpan-d3 / home / neoDesign / netChop3.1 / netChop

[0075] Finally, the sequence results of the best neoantigen peptide combination generated by given multiple neoantigen peptides will be output.

[0076] In order to test the effect of the present system, this example prepared target proteins and reference proteins for evaluating the effect of the present invention.

[0077] This system extracted 100 patient samples from the Cancer Genome Atlas (TCGA) public database, and the corresponding neoantigen peptides of these samples were derived from single nucleotide variations (SNVs) in pancreatic cancer and lung cancer. To ensure its practicality, we limited the number of peptides corresponding to each sample to between 10 and 30. The equilibrium structure parameter was set to 1, the screening threshold parameters of NetMHCpan4.1-BA, NetMHCpan4.1-EL and MHCflurry were set to 0.05, the screening threshold parameter of NetChop was set to 0.5, and the recommended screening threshold parameter of HMMER was set to 1e-5, and a total of 100 target proteins were generated. These proteins represent the best neoantigen vaccine sequences designed by the system for these patients.

[0078] In this example, 100 proteins with known functions, such as ubiquitin protein ligase E3A and DNA damage binding protein, were randomly selected from the protein data bank (PDB) database as reference proteins. These reference proteins were used to compare with the best vaccine sequences predicted by us to evaluate their domain composition and looseness.

[0079] like Figure 4 A to Figure 4 As shown in Figure C, this example evaluates the parameters of the 100 target proteins obtained, including the number of linkers, the number of peptides, the distribution of the number of domains and the distribution of looseness parameters, and also compares the differences between the target proteins and the reference proteins in the number of domains and the looseness characteristics.

[0080] in, Figure 4 A is a distribution diagram of the number of linkers and peptides. The upper part represents the distribution of the number of peptides, and the right part represents the distribution of the number of linkers. Figure 4 B is the area plot of the domain parameter, and the distribution rate is the percentage of the number of sequences in the domain to the total number of sequences. Figure 4 C is the area plot of the looseness parameter, and the number of sequences is the number of sequences corresponding to the looseness number.

[0081] The evaluation results show that the sequences designed by this system do contain as few linkers as possible. Among the 100 sequences, 62 have no linkers, 31 have one linker, 6 have two linkers, and only one has three linkers. In theory, more peptide sequences may lead to more linkers, but in practice, the correlation between the two is weak ( Figure 4 A). The purpose of this system design is to minimize the generation of domains. Compared with the reference protein, the protein sequence domain value designed by this system is 0, and the looseness function value is low, which meets the ideal definition ( Figure 4 B, Figure 4 C) Therefore, the design sequence of this system successfully achieves the expected looseness.

[0082] Although the embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and cannot be understood as limiting the present invention. A person of ordinary skill in the art can change, modify, replace and modify the above embodiments within the scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.< / path> < / path> < / path> < / cpu> < / cutoff> < / species> < / parameter>

Claims

1. A system for selecting a combination of multivalent tumor neoantigens, characterized in that: include: A library construction module is used to calculate the optional libraries corresponding to the front and back ends of each neoantigen peptide sequence in the input list; The optimal path selection module is used to determine the optimal ordering of peptide sequences in the input list, and iteratively adds target peptide sequences with minimized custom objective function values ​​to the path one by one. The final path is the selected optimal path, which represents the optimal ordering of peptide sequences in the input list; the linker addition module is used to check the part of the sequence with the best order to which linkers need to be added, and screen suitable linkers from the linker library to ensure that the sequence with the best order does not produce unexpected new antigens; The optimal path selection module performs the following operations: Based on the custom objective function, the target peptides with the minimized custom objective function value are gradually added to the pathway in an iterative manner. The custom objective function is: l*num+domain+looseness, where num is the number of optional peptides contained in the optional library associated with the peptide, domain is the number of domains obtained based on the HMMER software, looseness is a custom function built based on the secondary structure software GOR4, and the parameter "l" is used to balance the structural parameters and minimize the linker-related parameters, and its value depends on the preference for minimizing the linker or a looser structure; The looseness custom function is: Looseness = (spiral + fold - random curl).

2. A system for selecting a multivalent tumor neoantigen combination as claimed in claim 1, characterized in that: The library construction module uses the NetMHCpan algorithm to identify new antigens, sets thresholds for EL rank and BA rank parameters, and obtains a union of new antigens that simultaneously meet the parameter setting thresholds.

3. A system for selecting a multivalent tumor neoantigen combination as claimed in claim 1, characterized in that: The library construction module uses the MHCflurry algorithm to identify new antigens, sets affinity parameter thresholds, and obtains a union of new antigens that simultaneously meet the parameter setting thresholds.

4. A system for selecting a multivalent tumor neoantigen combination as claimed in claim 1, characterized in that: The library construction module uses the NetChop algorithm to identify new antigens, sets a proteasome cleavage probability parameter threshold, and obtains a union of new antigens that simultaneously meet the parameter setting threshold.

5. A system for selecting a multivalent tumor neoantigen combination as claimed in claim 1, characterized in that: The optimal path selection module performs the following operations: If there is only one alternative peptide in the alternative library associated with the peptide, the peptide will be combined with its only alternative peptide and form the starting point of the pathway by direct connection; If the alternative library associated with each peptide contains multiple alternative peptides, all alternative libraries will be statistically analyzed and ranked to determine the peptide associated with the alternative library containing the least alternative peptides as the starting point.

6. A system for selecting a multivalent tumor neoantigen combination as claimed in claim 1, characterized in that: The linker adding module checks whether there is a part in the best sorted sequence to which a linker needs to be added, and selects a suitable linker from the linker library to ensure that no unexpected antigens will be generated after being added to the sequence.

7. A method for selecting a multivalent tumor neoantigen combination, characterized in that: include: Library construction, calculating the optional libraries corresponding to the front and back ends of each new antigen peptide sequence in the input list; Optimal pathway selection, in which the target peptide sequences with the minimized custom objective function value are added to the pathway one by one in an iterative manner to determine the best ordering of the peptide sequences in the input list; The optimal path selection includes: Based on the custom objective function, the target peptides with the minimized custom objective function value are gradually added to the pathway in an iterative manner. The custom objective function is: l*num+domain+looseness, where num is the number of optional peptides contained in the optional library associated with the peptide, domain is the number of domains obtained based on the HMMER software, looseness is a custom function built based on the secondary structure software GOR4, and the parameter "l" is used to balance the structural parameters and minimize the linker-related parameters, and its value depends on the preference for minimizing the linker or a looser structure; The looseness custom function is: Looseness = (spiral + fold - random curl) Add linkers, check whether there are parts in the optimally sorted sequence to which linkers need to be added, and screen suitable linkers from the linker library to ensure that no unexpected new antigens are generated after being added to the sequence.

Images