Method for screening new antigens immunogenic for aml and multi-parameter prediction model

Abstract

The present application relates to the field of biotechnology, and particularly relates to a method for screening AML immunogenic neoantigens and a multi-parameter model. The present application takes acute myeloid leukemia (AML) as the research object, integrates multi-omics data of large-scale AML patient samples, and combines machine learning algorithms to construct a multi-parameter prediction model of AML immunogenic neoantigens. Compared with the existing method of screening neoantigens using a single parameter, the model can significantly improve the accuracy and efficiency of screening immunogenic neoantigens in AML. The screened immunogenic neoantigen targets can help researchers more accurately select and design mRNA vaccines based on tumor neoantigen epitopes, and provide personalized precision treatment strategies for AML patients.

Description

Technical Field

[0001] This invention relates to the field of biotechnology, and in particular to a method for screening novel immunogenic antigens of AML and a multi-parameter prediction model. Background Technology

[0002] Acute myeloid leukemia (AML) is a malignant blood disease that poses a serious threat to human life and is characterized by typical heterogeneity. It is mainly manifested by the abnormal accumulation of immature myeloid precursors in the bone marrow and other tissues, which not only leads to impaired hematopoietic function but also causes bone marrow failure. In China, the incidence of AML is approximately 2.57 per 100,000 people, while the annual mortality rate from this disease is 1.57 per 100,000 people, posing a serious threat to human health.

[0003] Currently, approved drugs and investigational treatments for AML include a variety of advanced regimens: First, for elderly AML patients or those unsuitable for intensive chemotherapy, a combination of epigenetic-associated hypomethylating agents (such as azacitidine and decitabine) and venetoclax can be used; for younger AML patients, a combination of intensive chemotherapy and venetoclax can be considered. Second, for AML patients carrying FLT3 gene mutations, a combination therapy of FLT3 inhibitors (including gilteritinib, midostorin, sorafenib, quizartinib, and crenolanib) with intensive chemotherapy or low-intensity chemotherapy / HMA can be chosen. Furthermore, AML patients with IDH1 and IDH2 mutations can be treated with a combination of IDH inhibitors (ivosidenib targeting IDH1, enasidenib targeting IDH2) and / or venetoclax. For AML patients with TP53 gene mutations, treatment options include the TP53 modulator APR246 and magrolimab (an anti-CD47 monoclonal antibody that enhances macrophage-mediated phagocytosis). For mixed lineage (MLL) leukemia patients, Menin inhibitors offer a novel treatment approach. Finally, oral anti-AML therapies (such as oral decitabine and azacitidine) are being developed to replace or improve the efficacy of traditional parenteral treatments. Despite the availability of various treatment options for AML patients, the five-year survival rate remains below 30%, and most patients still face relapse and ultimately die from the disease. This situation urgently necessitates the development of more effective treatment strategies to improve the prognosis of AML patients.

[0004] Cancer immunotherapy has shown great promise in combating chemotherapy resistance or malignant clones, offering long-term benefits. However, its application in AML is hampered by several factors, including the low level of endogenous immune response and inherent immune escape mechanisms in AML, and especially the lack of AML-specific neoantigen targets. Extensive genomic and transcriptomic sequencing has revealed highly reproducible patterns of genetic abnormalities in AML, distinguishing it from many other cancer types. Therefore, neoantigens generated by somatic mutations in these high-frequency genes are ideal targets for immunotherapeutic intervention and have the potential to trigger strong anti-tumor immune responses.

[0005] Unlike passenger mutations, driver mutations are generally more stable and less susceptible to immune editing and loss of expression, as they play a crucial role in maintaining tumor cell phenotypes. In fact, multiple studies have confirmed neoantigen-specific T-cell responses targeting driver mutations, such as NPM1 and the leukemia initiator gene fusion CBFB-MYH11, confirming the potential immunogenicity of AML and its applicability to neoantigen-based cancer vaccines or selective T-cell therapies. Driver mutations are considered an ideal source of neoantigens, but neoantigen abundance is limited by HLA allele diversity among patients, highlighting the need for further identification of the neoantigen landscape in AML.

[0006] Tumor neoantigens are protein fragments generated by somatic mutations specific to tumor cells. They can be recognized by the immune system, triggering an immune response. Screening for effective tumor neoantigens is a crucial step in cancer immunotherapy research, particularly in developing personalized cancer vaccines and employing targeted T-cell therapies. Current methods for screening tumor neoantigens mainly consider the following aspects: the binding affinity of the neoantigen to HLA molecules, the binding stability of the neoantigen to HLA molecules, the hydrophobicity of the neoantigen peptide, the location of peptide mutations, the diversity of T-cell receptors, and the expression level of the neoantigen peptide-HLA complex on the cell surface. While predictive models built based on these characteristics can achieve high accuracy in predicting neoantigens, only about 20%–40% of neoantigens have been shown to activate cellular immunity, indicating a significant lack of accuracy in predicting immunogenic neoantigens. Therefore, providing a method and predictive model that can accurately identify immunogenic tumor neoantigens in AML, facilitating the development of personalized cancer vaccines and providing patients with more precise and effective treatment, has profound clinical application significance. Summary of the Invention

[0007] In view of this, the present invention provides a method and a multi-parameter prediction model for screening immunogenic neoantigens in AML. This method can more comprehensively evaluate the immunogenicity of neoantigens and significantly improve the accuracy and efficiency of screening immunogenic neoantigens in AML.

[0008] To achieve the above-mentioned objectives, the present invention provides the following technical solution:

[0009] The method for screening new immunogenic antigens for acute myeloid leukemia includes the following steps:

[0010] S1. Identify driver mutation genes in AML based on targeted sequencing data, transcriptome data, and clinical data from patients with acute myeloid leukemia.

[0011] S2. Predict neoantigen epitopes generated by driver mutation genes in HLA-I type and AML, take the neoantigen epitopes generated by gene mutation and fusion gene mutation as neoantigen epitope set, and calculate the antigen presentation characteristic value of neoantigen epitope.

[0012] S3. Identify the optimal threshold set for enriching immunogenic neoantigens, and based on the optimal threshold set, construct a multi-parameter model for screening immunogenic neoantigens.

[0013] The optimal threshold set includes threshold set A or threshold set B;

[0014] The threshold set A includes: short peptides with a length of 9 amino acids, binding affinity <94nM, and tumor abundance >30TPM;

[0015] The threshold set B includes: short peptides with a length of 9 amino acids, binding affinity <157 nM, and binding stability >1.2 hours;

[0016] S4. Based on the multi-parameter model and HLA typing, peptides that meet the optimal threshold set are screened from the neoantigen epitope set as immunogenic neoantigens.

[0017] In this invention, the multi-parameter model is constructed based on AML antigen presentation characteristics and T cell recognition characteristics; the T cell recognition characteristics include affinity difference and / or exogenousness, and correspondingly, the optimal threshold set, threshold set A and threshold set B, further include: affinity difference values ​​less than 10. -4 Or the exogenous value is greater than 10 -5 .

[0018] In some specific embodiments, step S2 specifically includes:

[0019] Step S2 specifically includes:

[0020] S21. Analyze transcriptome RNA-seq, whole genome and / or whole exome sequencing data of AML patient samples to predict HLA type; use PRIME and DeepHLAPan software to calculate the immunogenicity score of a set of tumor neoantigen epitopes with known immunogenicity;

[0021] S22. Subsequently, ROC curves were used to identify the optimal immunogenicity score thresholds calculated by PRIME and DeepHLAPan as 0.172 and 0.449, respectively.

[0022] S23. Take the neoantigens generated by gene mutations and fusion gene mutations as a set of neoantigen epitopes, and use PRIME and DeepHLAPan to calculate the corresponding immunogenicity scores. Neoantigen epitopes with an immunogenicity score greater than 0.172 calculated by PRIME and an immunogenicity score greater than 0.449 calculated by DeepHLAPan are marked as immunogenic neoantigen epitopes; otherwise, they are marked as non-immunogenic neoantigen epitopes.

[0023] In this invention, in step S3, the optimal threshold set for enriching neoimmunogenic antigens is identified based on the Monte Carlo cross-validation algorithm.

[0024] In some specific embodiments, step S3 specifically includes: using the numerical parameters of all neoantigen peptides generated by AML-driven mutations, the categorical parameters of the neoantigen peptides, and the label information of each peptide as input to the Monte Carlo cross-validation algorithm; for each unique parameter combination, selecting 10 random subsets of 70% of the peptides in the entire dataset, and using the feature values ​​of each unique parameter combination to stratify the neoantigen peptides; subsequently, using Fisher's exact test to calculate the enrichment degree of immunogenic peptides under a given parameter combination; finally, taking the parameter set with the smallest average p-value among the 10 random subsets as the parameter set with the best overall stratification ability, and obtaining the optimal threshold set related to antigen presentation characteristics.

[0025] In some embodiments, step S3 of the present invention further includes:

[0026] The Monte Carlo cross-validation algorithm uses the affinity differences and exogenous values ​​of all neoantigen peptides generated by AML-driven mutations, along with the tag information of each peptide, as inputs. For each unique affinity difference or exogenous value, 10 random subsets of 70% of the peptides in the entire dataset are selected, and the neoantigen peptides are stratified using the feature values ​​of each unique parameter combination. Subsequently, Fisher's exact test is used to calculate the enrichment degree of immunogenic peptides under a given parameter combination. Finally, the parameter set with the smallest average p-value among the 10 random subsets is taken as the parameter set with the best overall stratification ability, thus obtaining the optimal threshold set related to T cell recognition features.

[0027] This invention also provides a multi-parameter prediction model for neoimmunogenic antigens in acute myeloid leukemia, wherein the threshold set includes threshold set A or threshold set B;

[0028] The threshold set A includes: short peptides with a length of 9 amino acids, binding affinity <94nM, and tumor abundance >30TPM;

[0029] The threshold set B includes: short peptides with a length of 9 amino acids, binding affinity <157 nM, and binding stability >1.2 hours.

[0030] In the multi-parameter prediction model of this invention, the threshold set A further includes: affinity difference values ​​less than 10. -4 Or exogenous values ​​greater than 10 -5 .

[0031] The threshold set B also includes: affinity difference values ​​less than 10. -4 Or exogenous values ​​greater than 10 -5 .

[0032] The multi-parameter prediction model (or optimal parameter set) provided by this invention integrates the antigen presentation characteristics and T cell recognition characteristics of neoantigen epitopes. Based on the multi-level characteristic values ​​of neoantigen peptides, including peptide length, binding affinity, tumor abundance, affinity difference or exogenousness, the immunogenicity of neoantigen peptides is comprehensively evaluated. The stronger the immunogenicity of the neoantigen peptide, the stronger the immune response it induces, and the better the effect of the mRNA vaccine designed using it in vivo.

[0033] The present invention also provides a method for predicting neoantigens in acute myeloid leukemia, which includes: using the multi-parameter prediction model described in the present invention to predict the neoantigens of the tumor to be tested.

[0034] Using the multi-parameter prediction model of this invention, when predicting a peptide, the sequence, HLA molecules, and tumor abundance of the peptide to be predicted are first input. Then, the corresponding binding affinity, peptide length, binding stability characteristic values, affinity difference value, and exogenous value are calculated. The multi-parameter prediction model of this invention (or the optimal threshold set of this invention) can then classify the peptides to be predicted into two categories: those selected through threshold screening (short peptides with a length of 9 amino acids, binding affinity <94 nM, tumor abundance >30 TPM, and affinity difference value less than 10). -4 Or exogenous values ​​greater than 10 -5 Peptides that pass the threshold screening are classified as immunogenic peptides; those that fail the threshold screening are classified as non-immunogenic peptides.

[0035] This invention focuses on acute myeloid leukemia (AML), integrating multi-omics data from a large-scale AML patient samples. It utilizes machine learning algorithms to integrate different characteristic data of neoantigens generated by AML mutations, including binding affinity, binding stability, hydrophobicity, and T-cell recognition, to construct a multi-parameter prediction model for AML immunogenic neoantigens. Compared to existing methods using single parameters to screen neoantigens, this model can more comprehensively assess the immunogenicity of neoantigens, significantly improving the accuracy and efficiency of immunogenic neoantigen screening in AML. Identifying immunogenic tumor neoantigens in AML has profound clinical significance, not only helping researchers more accurately select and design mRNA vaccines based on tumor neoantigen epitopes, but also providing personalized precision treatment strategies for AML patients. Furthermore, it is crucial for improving the personalization, effectiveness, and safety of cancer treatment, demonstrating significant practical application value. Attached Figure Description

[0036] Figure 1 Demonstrating the driver mutation map in AML;

[0037] Figure 2 The spectrum of tumor neoantigens derived from driver mutations in AML is shown below; A, the most frequent HLA type in AML patients; B, the average number of neoantigens generated by Indel (insertion or deletion), Fusion (fusion), and SNV (single nucleotide variant); C, the number of neoantigens generated by each mutated gene; D, the number of neoantigens generated by each fusion gene; E, the correlation between the number of mutation sites occurring on each driver mutation and the number of neoantigens generated; FH, the immunogenicity scores of neoantigens generated by Indel, Fusion, and SNV calculated by PRIME (F), DeepHLAPan (G), and IEDB (H) software.

[0038] Figure 3This paper presents a multi-parameter model for screening immunogenic neoantigens based on antigen presentation characteristics. The model includes: A) Correlation between binding affinity score and binding stability; B) Correlation between presentation characteristics of neoantigens generated by mutations; C) Correlation between presentation characteristics of neoantigens generated by fusion genes; D) A diagram illustrating the optimal set of antigen presentation parameter thresholds identified using a Monte Carlo cross-validation algorithm (repeated-random-subsample-based approach); E) Significant enrichment of immunogenic neoantigens from mutation sources by the presentation feature threshold set; F) Significant enrichment of immunogenic neoantigens from fusion genes by the presentation feature threshold set; G) Results of in vitro peptide exchange experiments based on ELISA; H) Binding strength of neoantigens generated by Indel, SNV, and Fusion with HLA-A:0201; I) Binding strength of 9-amino acid and 10-amino acid neoantigen peptides with HLA-A:0201; J) Significant enrichment of neoantigen peptides with strong binding affinity to HLA-A:0201 by the optimal set of antigen presentation parameter thresholds.

[0039] Figure 4 The following are examples of multi-parameter models for screening immunogenic neoantigens based on antigen presentation and T cell recognition features: A) Correlation between antigen presentation and T cell recognition features; B) Monte Carlo cross-validation algorithm for selecting optimal T cell feature parameter values; C) Screening for overlapping epitopes of neoantigens from mutation sources using different feature thresholds; D) Screening for overlapping epitopes of neoantigens from fusion genes using different feature thresholds; E) Precision-recall curves of the ranking model based on binding affinity; F) Precision-recall curves of the ranking model based on antigen presentation features; G) Precision-recall curves of the ranking model based on both antigen presentation and T cell recognition features.

[0040] Figure 5 The accuracy of the multi-parameter model was validated using immunological experiments. A) Immunogenicity of neoantigen peptides was detected by enzyme-linked immunospot (ELISPOT) assay; B) Immunogenicity of neoantigen peptides was detected by peptide-MHC I-tetramer assay; C) Statistical results of ELISPOT and peptide-MHC I-tetramer assays were presented; D) The proportion and number of peptide-MHC I-tetramer-positive CD8 T cells induced by neoantigen peptides generated by Indel and SNV were compared; F) The optimal parameter threshold set significantly enriched strong neoantigen peptides. Detailed Implementation

[0041] This invention provides a method and a multi-parameter prediction model for screening novel immunogenic antigens for AML. Those skilled in the art can refer to the content of this document and appropriately modify the process parameters to achieve the desired result. It should be particularly noted that all similar substitutions and modifications are obvious to those skilled in the art and are considered to be included in this invention. The methods and applications of this invention have been described through preferred embodiments, and those skilled in the art can clearly modify or appropriately change and combine the methods and applications described herein without departing from the content, spirit, and scope of this invention to implement and apply the technology of this invention.

[0042] Unless otherwise specified, the test materials used in this invention are all commercially available products that can be purchased on the market.

[0043] The present invention will be further illustrated below with reference to the embodiments:

[0044] Example 1

[0045] (I) Data Collection and Processing

[0046] Targeted sequencing data, transcriptome data, and corresponding clinical data from 304 adult AML patients were collected from Ruijin Hospital. Furthermore, whole-genome / whole-exome and transcriptome data of AML patients were downloaded from the public databases Gene Expression Omnibus (GEO, https: / / www.ncbi.nlm.nih.gov / geo / ) and The Cancer Genome Atlas Program (TCGA, https: / / www.cancer.gov / ccg / research / genome-sequencing / tcga). Driver mutation genes in AML were identified using MutSigCV, 2020Plus, OncodriveCLUST, and OncodriveFML software, respectively. Driver mutation genes occurring in AML were systematically defined based on detection by at least two software programs. Ultimately, 49 AML driver genes were defined. Figure 1 ).

[0047] (II) Predicting neoantigenic epitopes generated by driver mutations in AML

[0048] Patient samples with transcriptome RNA-seq sequencing data and whole genome / whole exome / targeted sequencing data were analyzed, and the HLA type of all AML patients was identified using OptiType software. Figure 2 A shows the most frequent HLA type in AML patients. Figure 2B shows the average number of neoantigens generated by three mutation categories: Indel (insertion or deletion), Fusion (fusion), and SNV (single nucleotide variant). Figure 2 C represents the number of neoantigens produced by each mutated gene. Figure 2 D represents the number of neoantigens generated by each fusion gene. Pearson correlation analysis revealed a significant positive correlation between the number of mutation sites occurring on each driver mutation and the number of neoantigens generated. Figure 2 E). Immunogenicity scores of neoantigens generated by Indel, Fusion, and SNV were assessed using PRIME, DeepHLAPan, and IEDB software.

[0049] (III) Using machine learning algorithms to identify the optimal set of thresholds associated with AML antigen presentation characteristics

[0050] In our data focus on neoantigens generated by AML-driven mutations (gene mutations and fusion genes), multiple antigen presentation-related features are not independent of each other; there is a strong interdependence among presentation features. For example, the binding affinity score and binding stability of neoantigen epitopes derived from AML-driven mutations or fusions are both significantly negatively correlated. Figure 3 A). The hydrophobicity of peptides also showed a slight negative correlation with tumor abundance and binding affinity, although this correlation was not significant in peptides of mutant origin. Figure 3 To this end, we used a Monte Carlo cross-validation algorithm (repeated-random-subsample-based approach) to identify the optimal set of antigen presentation parameter thresholds to identify immunogenic peptides generated by AML-driven variants. Figure 3 (D) Specifically, we use the numerical parameters (MHC binding affinity BA, tumor abundance TA, MHC binding stability BS) and categorical parameters (mutation location MP, peptide length PepLen) of all neoantigen peptides generated by AML-driven mutations (including gene mutations and fusion genes), as well as the tag information (immunogenicity) of each peptide, as input to the algorithm. For each unique parameter combination, we select 10 random subsets of 70% of the peptides in the entire dataset and use the feature values ​​of each unique parameter combination to stratify the neoantigen peptides (including immunogenic and non-immunogenic peptides). Subsequently, Fisher's exact test is used to calculate the enrichment of immunogenic peptides under a given parameter combination. Finally, we select the parameter set with the smallest average p-value among the 10 random subsets as the parameter set with the best overall stratification ability.

[0051] Using this method, we identified a set of antigen presentation-related variables and corresponding thresholds. This threshold set included short peptides with a length of 9 amino acids, binding affinity <94 nM, and tumor abundance >30 TPM. Using this set of three threshold parameters, 93.6% of mutation-derived non-immunogenic peptides could be filtered out. Figure 3 E). Similarly, the threshold set of neoantigen peptides derived from fusion genes consists of short peptides with a length of 9 amino acids, binding affinity <157 nM, and binding stability >1.2 hours, which can filter out 95.2% of non-immunogenic peptides. Figure 3 F).

[0052] To validate this multi-parameter model, we randomly selected 30 short peptides (27 from mutant sources and 3 from fusion genes) that bind to HLA-A:0201 for synthesis. We first calculated the multi-parameter characteristics of these 30 peptides, including binding affinity, tumor abundance, peptide length, and binding stability. Subsequently, we performed an ELISA-based in vitro peptide exchange assay to directly verify whether the neoantigen peptides bind to the HLA-A:0201 molecule. We found that approximately 73% (22 out of 30) of the neoantigen peptides could bind to HLA-A:0201. Figure 3 G). We labeled the bound neoantigen peptides as positive (Positive, Pos) and the unbound neoantigen peptides as positive (Negative, Negative). Using these experimental data, we verified that the neoantigen peptides generated by the Indel mutation have a stronger binding to the HLA-A:0201 molecule. Figure 3 H), a neoantigen peptide with a length of 9 amino acids has a significantly stronger binding affinity to the HLA-A:0201 molecule. Figure 3 I). Finally, based on the calculated multi-parameter characteristic values ​​of the 27 mutation-derived neoantigen peptides, we divided these 27 neoantigen peptides into predicted presented (True) and non-predicted presented (False). We then used Fisher's exact test to assess whether binding-positive (Pos) neoantigen peptides were enriched under our multi-parameter characteristic model screening (predicted presented group). We confirmed that our defined multi-parameter threshold set can significantly enrich strongly binding neoantigen epitopes. Figure 3 J, P-value=0.033, OR=10.163).

[0053] (iv) Constructing a multi-parameter model for screening immunogenic neoantigens by combining antigen presentation and T cell recognition characteristics.

[0054] A durable and effective immune response induced by a neoepitaxy requires not only presentation but also recognition of the neoantigen epitope as a foreign substance by T cells. Two short peptide properties, affinity difference (defined as the ratio of HLA-binding affinity of mutant to wild-type neoepitaxy) and exogeneity (defined as the probability of T cell receptors recognizing mutant neoepitaxy), have been shown to significantly predict the immunogenicity of tumor neoantigens. Importantly, these two properties are mutually exclusive and are only associated with the ability to induce an immune response in neoepitaxy most likely to be presented. Therefore, we computed these two properties for neoantigen epitopes screened by presentation characteristics in our dataset. We observed that affinity difference and exogeneity were not associated with antigen presentation-related properties ( Figure 4 A) indicates that these are independent parameters related to the immunogenicity of neoantigen epitopes. Similarly, we use the affinity difference and exogenous value of neoantigen epitopes selected by presentation features, along with the label information (immunogenicity) of each peptide, as inputs to the Monte Carlo cross-validation algorithm. For each unique parameter combination, we select 10 random subsets of peptides representing 70% of the entire dataset and stratify them using the feature values ​​of each unique parameter combination. Subsequently, Fisher's exact test is used to calculate the enrichment of immunogenic peptides under a given parameter combination. Finally, we select the parameter set with the smallest average p-value among the 10 random subsets as the parameter set with the best overall stratification ability. Based on this method, we identify the parameter set with affinity differences less than 10... -4 Or exogenous greater than 10 -5 The optimal threshold for distinguishing between immunogenic and non-immunogenic neoepitopes ( Figure 4 B). We will consider affinity differences to be less than 10. -4 Or exogenous greater than 10 -5 The new epitope is called the new epitope recognized by T cells.

[0055] Our results highlight the importance of two distinct sets of feature thresholds for neoantigen immunogenicity, each consisting of four unique and independent peptide features, for screening immunogenic neoantigen epitopes generated by AML-driven mutations and fusions. Figure 4 CD demonstrates the overlap between these features of all neoantigenic epitopes derived from driver mutations. Regardless of whether the neoantigenic epitopes are generated by driver mutations or fusions, both sets of thresholds screened out over 98% of non-immunogenic peptides. Figure 4 Finally, we used precision-recall curves to compare the impact of antigen presentation and T cell recognition-related feature thresholds on neoantigen immunogenicity models. We found that ranking neoantigen epitopes presented and recognized by T cells, followed by ranking based on their binding affinity, significantly improved the model's precision and recall compared to ranking solely based on binding affinity or prior to ranking based on presentation status followed by binding affinity. Figure 4 (EG). In summary, our findings highlight the contribution of features associated with antigen presentation and T cell recognition to improving neoantigen immunogenicity models.

[0056] (V) Verification of the accuracy of the multi-parameter model based on immunological experiments

[0057] To verify the accuracy of our established multi-parameter prediction model in screening immunogenic neoantigen epitopes, we conducted ELISPOT and MHC tetramer assays. The detailed steps are as follows:

[0058] (1) Inducing the production of monocyte-derived dendritic cells and neoantigen-specific T cells

[0059] CD14 cells were isolated from peripheral blood mononuclear cells (PBMCs) of healthy individuals using magnetic bead sorting. + Monocytes were cultured for 8 days in X-vivo 15 medium containing 5% human serum, GM-CSF, and IL-4. On day 7 of culture, immature dendritic cells were stimulated to mature with LPS and IFN-γ. On day 8, the stimulated and matured dendritic cells were irradiated with 30 Gy using an X-ray irradiator and pulsed with a short peptide at 10 ng / µl overnight. At this point, the dendritic cells were ready for subsequent DC-T co-culture experiments.

[0060] CD8 was isolated from PBMCs of the same healthy individual using magnetic beads. + T cells were co-cultured with corresponding short-peptide-pulsed dendritic cells at a DC:T ratio of 1:10 in X-vivo 15 medium containing 8% human serum. During the first week of co-culture, IL-12 and IL-21 were added to the medium, followed by IL-2 two days later. From the second week onwards, IL-2, IL-7, IL-15, and IL-21 were added to expand neoantigen-specific T cells until the fourth week. During this period, fresh complete medium containing cytokines was added every two days. Neoantigen-specific T cells were analyzed every three days using flow cytometry and ELISApot during weeks 3 and 4 of co-culture.

[0061] (2) Synthesis of antigen-specific short peptides and peptide-MHC I-tetramer complexes

[0062] Lyophilized short peptides with the corresponding sequences were ordered from GenScript and dissolved in DMSO to a storage concentration of 10 mM. Short peptide-specific tetramers were prepared using the QuickSwitch™ Quant Tetramer Kit-PE from MBL. Specifically, the storage concentration short peptide solution was diluted to 1 mM with sterile triple-distilled water, and then 50 μL of Tetramer was mixed with 1 μL of the 1 mM short peptide and 1 μL of Peptide Exchange Factor. The mixture was incubated at room temperature in the dark for 4 hours before use.

[0063] (3) Flow cytometry

[0064] To detect the neoantigen-specific CD8 + T cells, 1 x 10 6 Each T cell was incubated with the prepared corresponding short peptide tetramer at 4°C in the dark for 40 minutes, followed by incubation with APC / Cyanine7 anti-human CD8 antibody at room temperature in the dark for 20 minutes. The cells were then processed using a FACS Fortessa flow cytometer, and subsequent data analysis was performed using FlowJo V10 software.

[0065] (4) IFN-γ enzyme-linked immunospot assay (ELISpot)

[0066] The IFN-γ produced by neoantigen-specific T cells was detected using the IFN-γ ELISpot kit. 1 x 10⁻⁶ cells were added to each well of a 96-well plate. 5 CD8+ T cells were collected in each well and stimulated with the corresponding short peptide. After co-culturing for 18-20 hours, the strips were washed with ice-cold deionized water, followed by incubation with the detection antibody at 37°C for 1 hour. After washing 6 times with washing buffer, streptavidin-HRP working solution was added and incubated at 37°C for 1 hour. After washing 6 times again with washing buffer, AEC solution was added to the strips for color development, and incubation was carried out at room temperature in the dark for 15-30 minutes. Finally, the strips were washed with deionized water to stop the reaction. After the strips dried, they were scanned and analyzed using an ELISpot CTL reader.

[0067] We selected 16 novel, highly immunogenic antigens binding to HLA-A:0201 and 4 binding to HLA-A:1101 for synthesis, followed by enzyme-linked immunospot (ELISPOT) and peptide-MHC I-tetramer assays. Figure 5 AB). Enzyme-linked immunospot assay revealed that 8 neoantigens exhibited significant immunogenicity compared to the negative control group; peptide-MHC I-tetramer assay revealed that 13 neoantigens exhibited significant immunogenicity compared to the negative control group. Figure 5C). We define neoantigen peptides that test positive by any method as immunogenic peptides. Through statistical analysis, we found a total of 15 neoantigen peptides to be immunogenic. Figure 5 C). By comparing the proportion of tetramer-positive cells induced by neoantigens generated by Indel mutations and SNV mutations, and the number of spots generated by IFN-γ stimulation, we confirmed that the neoantigens generated by Indel mutations have stronger immunogenicity. Figure 5 D).

[0068] Finally, based on the calculated multi-parameter characteristic values ​​of the 20 neoantigen peptides, including presentation and T cell recognition characteristics, we divided these 20 neoantigen peptides into those with both presentation and recognition (Presented & Recognized: Yes) and those without (Presented & Recognized: No). We then used Fisher's exact test to assess whether binding-positive (Positive, Pos) neoantigen peptides were enriched under our multi-parameter characteristic model screening (presented and recognized group). Using ELISPOT and peptide-MHC I-tetramer experimental data, we confirmed that our defined set of multi-parameter thresholds related to antigen presentation and T cell recognition can enrich immunogenic neoantigen peptides. Figure 5 E, P-value = 0.073, OR = 8.308. Of the neoantigen peptides that passed the multi-parameter threshold screening (Presented and Recognized group), 86.7% (13 / 15) were immunogenic; while of the neoantigen peptides that did not pass the multi-parameter threshold screening, 40% (2 / 5) were immunogenic. Figure 5 E).

[0069] The above results confirm that the multi-parameter model of this invention can significantly enrich immunogenic neoantigen peptides.

[0070] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for screening new immunogenic antigens for acute myeloid leukemia, characterized in that, Includes the following steps: S1. Identify driver mutation genes in AML based on targeted sequencing data, transcriptome data, and clinical data from patients with acute myeloid leukemia. S2. Predict neoantigen epitopes generated by driver mutation genes in HLA-I type and AML, take the neoantigen epitopes generated by gene mutation and fusion gene mutation as neoantigen epitope set, and calculate the antigen presentation characteristic value of neoantigen epitope. S3. Identify the optimal threshold set for enriching immunogenic neoantigens, and construct a multi-parameter model for screening immunogenic neoantigens based on the optimal threshold set. The optimal threshold set includes threshold set A or threshold set B; The threshold set A includes: short peptides with a length of 9 amino acids, binding affinity <94 nM, and tumor abundance >30 TPM; The threshold set B includes: short peptides with a length of 9 amino acids, binding affinity <157 nM, and binding stability >1.2 hours; S4. Based on the multi-parameter model and HLA typing, peptides that meet the optimal threshold set are screened from the set of neoantigen epitopes as immunogenic neoantigens; The multi-parameter model is constructed based on AML antigen presentation characteristics and T cell recognition characteristics; The threshold set A and threshold set B also include: affinity difference values ​​less than 10. -4 And / or exogenous values ​​greater than 10 -5 .

2. The method according to claim 1, characterized in that, Step S2 specifically includes: S21. Analyze transcriptome RNA-seq, whole genome and / or whole exome sequencing data of AML patient samples to predict HLA type; use PRIME and DeepHLAPan software to calculate the immunogenicity score of a set of tumor neoantigen epitopes with known immunogenicity; S22. Subsequently, ROC curves were used to identify the optimal immunogenicity score thresholds calculated by PRIME and DeepHLAPan as 0.172 and 0.449, respectively. S23. Take the neoantigens generated by gene mutations and fusion gene mutations as a set of neoantigen epitopes, and use PRIME and DeepHLAPan to calculate the corresponding immunogenicity scores. Neoantigen epitopes with an immunogenicity score greater than 0.172 calculated by PRIME and an immunogenicity score greater than 0.449 calculated by DeepHLAPan are marked as immunogenic neoantigen epitopes; otherwise, they are marked as non-immunogenic neoantigen epitopes.

3. The method according to claim 1, characterized in that, In step S3, the optimal threshold set for enriching neoimmunogenic antigens is identified based on the Monte Carlo cross-validation algorithm.

4. The method according to claim 1, characterized in that, Step S3 specifically includes: using the numerical parameters of all neoantigen peptides generated by AML-driven mutations, the categorical parameters of the neoantigen peptides, and the label information of each peptide as inputs to the Monte Carlo cross-validation algorithm; for each unique parameter combination, selecting 10 random subsets of 70% of the peptides in the entire dataset, and using the feature values ​​of each unique parameter combination to stratify the neoantigen peptides; subsequently, using Fisher's exact test to calculate the enrichment degree of immunogenic peptides under a given parameter combination; finally, taking the parameter set with the smallest average p-value among the 10 random subsets as the parameter set with the best overall stratification ability, and obtaining the optimal threshold set related to antigen presentation characteristics.

5. The method according to claim 4, characterized in that, Step S3 further includes: The Monte Carlo cross-validation algorithm uses the affinity differences and exogenous values ​​of all neoantigen peptides generated by AML-driven mutations, along with the tag information of each peptide, as inputs. For each unique affinity difference or exogenous value, 10 random subsets of 70% of the peptides in the entire dataset are selected, and the neoantigen peptides are stratified using the feature values ​​of each unique parameter combination. Subsequently, Fisher's exact test is used to calculate the enrichment degree of immunogenic peptides under a given parameter combination. Finally, the parameter set with the smallest average p-value among the 10 random subsets is taken as the parameter set with the best overall stratification ability, thus obtaining the optimal threshold set related to T cell recognition features.

6. A multi-parameter prediction model for neoimmunogenic antigens in acute myeloid leukemia, characterized in that, Its threshold set includes threshold set A or threshold set B; The threshold set A includes: short peptides with a length of 9 amino acids, binding affinity <94 nM, and tumor abundance >30 TPM; The threshold set B includes: short peptides with a length of 9 amino acids, binding affinity <157 nM, and binding stability >1.2 hours; The multi-parameter prediction model takes the sequence of the peptide to be predicted, HLA molecules, and tumor abundance as input, and calculates the corresponding binding affinity, peptide length and binding stability characteristic values, affinity difference values, and exogenous values. Then, according to the threshold set, the peptides to be predicted are divided into two categories: those that pass the threshold screening are classified as immunogenic peptides; those that do not pass the threshold screening are classified as non-immunogenic peptides. The threshold set A and threshold set B also include: affinity difference values ​​less than 10. -4 Or the exogenous value is greater than 10 -5 .

7. A method for predicting neoantigens in acute myeloid leukemia, characterized in that, The multi-parameter prediction model described in claim 6 is used to predict the neoantigens of the tumor to be tested.

Images