Method and computer program for predicting neoantigens by processing lengths of peptides of different lengths via folding

Abstract

Disclosed is a method for treating a peptide sequence exceeding a unit length by a folding process when predicting a neoantigen using the peptide sequence and an HLA class I and / or HLA class II allele sequence. According to this method, the peptide sequence contained in the cancer tissue can identify the neoantigen in the cancer tissue regardless of the length of the peptide sequence being diversified. Accordingly, it is possible to overcome information imbalance and loss in training data on length, and to more accurately predict binding force to determine neoantigens in cancer tissue.

Description

Technical Field

[0001] Embodiments of the present invention relate to a method and computer program for predicting the neoantigen probability of peptides extracted from cancer tissue and cancer-derived DNA relative to HLA class I and / or class II alleles. Background Art

[0002] Cancer is one of the most common causes of death worldwide. Approximately 10 million new cases occur each year, accounting for approximately 12% of all deaths, making it the third leading cause of death. Recently, third-generation immunotherapy drugs have gained attention through first-generation chemotherapy drugs and second-generation targeted anticancer drugs due to progress in the development of new anticancer drugs. In particular, third-generation immunotherapy drugs have significantly reduced side effects due to treatment strategies that utilize the patient's own immune system, which is different from previous anticancer drugs. However, despite these advantages, there is a limitation that treatment strategies using immunotherapy drugs can only be established for patients expressing marker genes such as PD-L1 and microsatellite instability-high (MSI-H). Due to these limitations, it is necessary to establish strategies for treating patients who are difficult to accept traditional anticancer drugs, and one alternative has proposed cancer vaccines that utilize new antigens.

[0003] The immune system has learned not to attack normal cells by eliminating self-recognizing T cells in the thymus (thymic selection), which is called central tolerance. Cancer is a disease that occurs when mutations accumulate in the genome. For example, solid cancer cells such as gastric cancer, colon cancer and breast cancer have an average of 60 mutations, while lung cancer and melanoma caused by carcinogens have more mutations, about 150.

[0004] When a mutation occurs, the immune system produces abnormal proteins that it has never seen before. In this case, since the immune system has not encountered the mutant protein during the thymic selection process, it does not produce immune central tolerance to the mutant protein. Therefore, all mutant proteins may be considered immunogenic. However, most mutations include intracellular proteins that cannot contact the surface of tumor cells and therefore cannot directly contact immune cells. Mutations can only contact immune cells in one way, that is, by presenting 8 to 10 short peptide fragments called epitopes on the surface of cancer cells through the class I major histocompatibility complex (class I MHC, also known as human leukocyte antigen (HLA) in humans). At this time, the mutant peptide that binds to the class I MHC to be presented on the surface of the tumor cell is called a new epitope or new antigen, and the new epitope and MHC complex is called pMHC (called pHLA in humans). The new epitope may be a marker that helps CD8 cytotoxic T cells recognize cancer cells as foreign attackers.

[0005] However, in order for the immune response to mutations in cancer cells to be activated, cancer cell-derived neoantigens pMHC need to be cross-presented to T immune cells through antigen presenting cells (APCs) (especially dendritic cells other than cancer cells) together with accessory proteins expressed only on dendritic cells. Unlike other somatic cells, APCs present longer peptides as neoantigens through class II MHC, thereby activating CD4 helper T cells as helper cells. Ultimately, class I pMHC, class II pMHC, and co-stimulatory molecules all act to generate CD8 cytotoxic T cells with strong immunogenicity against neoantigens.

[0006] Through this process, activated T immune cells try to attack cancer cells, but cancer cells express a barrier called immune checkpoints that paralyze activated T cells and suppress immune responses (T cell exhaustion). Most of the immunotherapeutics currently used in clinical practice are immune checkpoint inhibitors. Immune checkpoint inhibitors are only effective in certain cancers where there are already a large number of activated and then paralyzed T cells. However, in many cases, a state of immune ignorance occurs in which T cells are not initially activated for new antigens and no immune response occurs. Recently, a phenomenon has been discovered in which an immune response is activated even for new antigens that are immune ignorant during treatment with new antigen-targeted cancer vaccines. In other words, it is expected that new antigen-targeted cancer vaccines can be applied to cancer types that do not use traditional immune checkpoint inhibitors, and even in the target population of immune checkpoint inhibitors, combination therapy may be expected to have greater therapeutic effects.

[0007] Traditional cancer vaccines target "tumor-associated antigens," which are "normal" proteins that are abnormally expressed in cancer cells at higher levels than in normal cells. Existing vaccines can be made using these common targets, but have the disadvantage of low antigenicity due to central tolerance, and no vaccine has yet been shown to have significant clinical efficacy. Unlike these "tumor-associated antigens," neoantigens do not respond to central tolerance and therefore may be ideal targets for cancer vaccines.

[0008] On the other hand, the HLA genes encoding MHC proteins have more than 13,000 highest polymorphisms. These polymorphisms lead to differences in the peptide binding groove sequences of the MHC proteins, and thus the MHC proteins have different affinities for peptides. Therefore, peptides with the same mutation may be presented as neoantigens only in patients with specific HLA types that have high affinity for certain MHC proteins. Due to this MHC restriction, less than 10% of mutant peptides are presented by MHC, and most neoantigens are expressed differently in each patient, which is why cancer vaccines need to be customized for each patient.

[0009] One of the many challenges that arise during cancer vaccine development comes from the different lengths of new antigens that bind to HLA. These different lengths limit the establishment of AI-based prediction models (including deep learning). To address these problems, traditional methods have adopted some strategies, such as independently developing models for each length or developing models after calibrating the model to a specific length. However, this preprocessing process has disadvantages such as insufficient learning data and information loss. Therefore, the present invention aims to address the diversity of new antigen sequences while minimizing data shortages and information loss. Summary of the invention

[0010] Technical issues

[0011] The present invention is based on the above-mentioned necessity, and the purpose of the present invention is to use a model learned from data (in such data, peptide sequences of different lengths are folded and unified to unit length) to predict the immunogenicity of peptides, predict binding ability (i.e., binding ability to HLA alleles), and determine new antigens in cancer tissues based on the predicted binding ability and immunogenicity.

[0012] Technical Solution

[0013] According to an embodiment of the present invention, a method may include: inputting a peptide sequence extracted from a target cancer tissue by a neoantigen prediction device, and determining whether the length of the peptide sequence exceeds a predetermined unit length; when the peptide sequence exceeds the unit length, determining the remaining region in the peptide sequence except for a predetermined fixed region, and processing the remaining region so that the value of the region exceeding the unit length is folded to shorten the peptide sequence to the unit length; inputting a sequence of an HLA class I or class II allele complex by the neoantigen prediction device; extracting a feature value from input data of a sequence value contained in the processed peptide sequence by the neoantigen prediction device, and outputting an immunogenicity prediction value of the feature value; extracting a peptide feature value from data processed to shorten the peptide sequence to a unit length by the neoantigen prediction device; and The α chain sequence or β chain sequence of class I or class II alleles is one-hot processed, and the HLA feature value corresponding to the processed chain sequence is extracted; the neo-antigen prediction device outputs a combinability prediction value by combining the peptide feature value and the HLA feature value; and the neo-antigen prediction device determines the neo-antigen of the peptide sequence based on the immunogenicity prediction value or the combinability prediction value.

[0014] Shortening the peptide to the unit length can be processed by a folding process in which a first value at a first position and a second value at a second position in the remaining region of the peptide are folded to be expressed as the sum of the first value and the second value, and the length of the peptide is reduced by 1.

[0015] The fixed region of the peptide may include a first value, a second value, and a last value, and the values ​​contained in the fixed region may not be unfolded.

[0016] The folding process can be repeated until the peptide reaches the unit length.

[0017] The outputting of the combinatorial prediction value may be performed by outputting the combinatorial prediction value using a combinatorial prediction model, and the combinatorial prediction model may be a model trained in the following manner: inputting a peptide feature value extracted from a peptide sequence, an HLA-α combinatorial feature value extracted from an α chain sequence of an HLA class I or class II allele, and an HLA-β feature value extracted from a β chain sequence of an HLA class II allele, and outputting a combinatorial prediction value corresponding to the binding force.

[0018] The combinatorial prediction model may also include a peptide processing model for extracting peptide feature values, an HLAα processing model for extracting HLA-α feature values, and an HLAβ processing model for extracting HLA-β feature values, and the combinatorial processing models of peptides, HLAα, and HLAβ are encoders in different encoder forms, respectively.

[0019] The outputting of the immunogenicity prediction value may be performed by outputting the immunogenicity prediction value using an immunogenicity prediction model, and the immunogenicity prediction model is a model trained in the following manner: inputting the feature value extracted from the peptide sequence and outputting the immunogenicity prediction value corresponding to the immunogenicity of the peptide sequence.

[0020] The target cancer tissue may include cells engineered to express a single HLA class I and / or class II allele.

[0021] The target cancer tissue may be obtained from a variety of organisms or may include mutant cells derived therefrom.

[0022] The target cancer tissue may include fresh or frozen tumor cells obtained from a variety of organisms.

[0023] The target cancer tissue may include fresh or frozen tissue cells obtained from a variety of organisms.

[0024] Effects of the Invention

[0025] According to the above-mentioned embodiments of the present invention, a model can be trained by folding peptide sequences of different lengths separately and unifying them into data of unit length to predict the immunogenicity of peptides, predict the binding ability to HLA alleles, and determine new antigens in cancer tissues based on the predicted binding ability and immunogenicity. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a block diagram of a neoantigen prediction apparatus 100 according to an embodiment of the present invention.

[0027] Figure 2 is an exemplary diagram of a learning model 200 according to an embodiment of the present invention.

[0028] Figure 3a is an exemplary diagram of a peptide sequence processed by a method according to an embodiment of the present invention. Figure 3b is an exemplary diagram of input data processed by conventional methods.

[0029] Figure 4a is an exemplary diagram of a unique-hot encoded peptide sequence according to an embodiment of the present invention.

[0030] Figure 4b is an exemplary diagram of a folded peptide sequence according to an embodiment of the present invention.

[0031] Figure 4c is an exemplary diagram of a peptide sequence to which weight values ​​are applied according to an embodiment of the present invention.

[0032] Figure 5 is a flow chart of a method for predicting immunogenicity according to an embodiment of the present invention.

[0033] Figure 6 is a diagram comparing a neoantigen prediction model according to an embodiment of the present invention with a traditional neoantigen prediction model. DETAILED DESCRIPTION

[0034] The configuration and operation of the present invention will be described in detail below with reference to the embodiments of the present invention shown in the accompanying drawings.

[0035] The present invention may have various modifications and various embodiments, and specific embodiments will be shown in the drawings and described in detail in the specific embodiments. The advantages and features of the present invention, and the methods for achieving these, will be more clearly understood from the embodiments described in detail below with reference to the drawings. However, the present invention is not limited to the embodiments disclosed below, but can be implemented in various different forms.

[0036] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings, wherein the same reference numerals designate the same or corresponding elements regardless of the reference numerals, and repeated descriptions thereof will be omitted.

[0037] In the following embodiments, the terms first, second, etc. are not used in a limiting sense but are used for the purpose of distinguishing one component from another.

[0038] In the following embodiments, a singular expression includes a plural expression unless the context clearly indicates otherwise.

[0039] In the following embodiments, the terms “comprising”, “having” or similar expressions mean that the features or components described in the present specification are present, but do not preclude the possibility that one or more other features or components may be added.

[0040] In the drawings, the size of the components may be enlarged or reduced for convenience of explanation. For example, the size and thickness of each configuration shown in the drawings are arbitrarily shown for convenience of explanation, and therefore the present invention is not necessarily limited to those shown above.

[0041] When any embodiment is feasible in other aspects, a specific processing sequence may also be performed in a sequence different from that to be described. For example, two processes to be described sequentially may be performed substantially simultaneously, or may be performed in a sequence opposite to that to be described.

[0042] Here, the target cancer tissue refers to the tissue to be tested. For example, the target cancer tissue can be a cancer tissue for detecting an antigen capable of causing an immune response. Preferably, the target cancer tissue can be an aggregate of tumor cells or cancer cells.

[0043] Here, mutation refers to all phenomena in which the arrangement of the base sequence A (adenine), T (thymine), G (guanine) and C (cytosine) of the gene containing genetic information in each organism changes and is different from the original genetic information of the corresponding species. These mutations lead to small or large structural mutations. Small mutations can include point mutations (occurring when a single base sequence changes) and mutations in which the base sequence is additionally inserted or deleted. Mutations that occur on a large scale and affect the structure include gene duplication, gene deletion, chromosome inversion, interstitial deletion, chromosome translocation, loss of heterozygosity, etc.

[0044] Depending on the type of cell to be generated, mutations are roughly divided into germline mutations and somatic mutations. Somatic mutations are gene mutations that occur in somatic cells, also known as somatic mutations or somatic variations, which may be caused by gene mutations or chromosomal abnormalities.

[0045] The occurrence of these mutations may lead to changes in the function of the proteins produced by the corresponding genes, and a specific function may be lost or activated to another function. Since these changes in protein function may cause or accelerate the development of cancer, these mutations may have a profound relationship with the development and progression of cancer, either directly or indirectly.

[0046] The base sequence in DNA, which contains the genetic information of an organism, consists of A, T, G, and C. Three of these base sequences are combined in a row to form a code for forming specific amino acids. Several such codes are combined to form proteins. Amino acids include alanine (Ala), cysteine ​​(Cys), aspartic acid (Asp), glutamic acid (Glu), phenylalanine (Phe), glycine (Gly), histidine (His), isoleucine (Ile), lysine (Lys), leucine (Leu), methionine (Met), asparagine (Asn), pyrrolysine (Ply), proline (Pro), glutamine (Gln), arginine (Arg), serine (Ser), threonine (Thr), selenocysteine ​​(Sec), valine (Val), tryptophan (Trp), and tyrosine (Tyr).

[0047] A peptide may refer to a peptide or polypeptide consisting of an amino acid sequence. In the body, there is an immune system to eliminate foreign substances that are not derived from the genetic information of each species, and in particular, among foreign peptides, there are immunogenic peptides that can induce an immune response. Mutations that occur during the development of cancer that are different from the original genetic information can also produce immunogenic peptides, and these immunogenic peptides produced in this way can bind to HLA class I or HLA class II proteins through a series of processes within the immune system. In addition, the immunogenic peptide may have a mutated amino acid sequence, and its amino acid length may be 8 to 12 amino acids for a new antigen that binds to HLA class I, or 13 to 20 amino acids for a new antigen that binds to HLA class II, but is not limited thereto and may be various lengths.

[0048] Here, machine learning refers to using a model containing multiple parameters to optimize the parameters using given data. Depending on the type of learning problem, machine learning can be divided into supervised learning, unsupervised learning, and reinforcement learning. Supervised learning is the process of learning the mapping between input and output, and can be applied when input and output pairs are given as data.

[0049] In this specification, a new antigen can be a peptide that induces an immune response, i.e., an immunogenic peptide. A new antigen can be induced by a specific mutation in a tumor cell and can be expressed as an epitope on a tumor cell. Hereinafter, for the sake of brevity, the immunogenic peptide will be referred to as a new antigen.

[0050] The new antigen prediction device according to an embodiment of the present invention can analyze the peptide sequence of the target cancer tissue and the HLA class I allele sequence of the organism to determine whether a specific peptide of the target cancer tissue used to treat the target cancer tissue can be determined as a new antigen. The length of the peptide sequence of the target cancer tissue may not be constant.

[0051] Here, human leukocyte antigen (HLA) is a protein present on the surface of human cells, and may have various types of gene sequences for each organism. In addition, the protein sequence of HLA is different for each organism. Based on this, each organism has different antigenicity.

[0052] Figure 1 is a block diagram of a neoantigen prediction apparatus 100 according to an embodiment of the present invention.

[0053] The neo-antigen prediction apparatus 100 may include a first data input unit 111 , a second data input unit 112 , an immunogenicity prediction unit 120 , a combinability prediction unit 130 , and a neo-antigen prediction unit 140 .

[0054] The new antigen prediction device 100 can determine the new antigen, which is an immunogenic peptide sequence for treating cancer tissues and cancer cells based on the peptide sequence of cancer tissues, cancer cells, etc. and the HLA class I and / or class II sequence of the organism. At this time, the new antigen prediction device 100 can determine the new antigen in the peptide sequence based on the immunogenicity of the peptide sequence and / or the combinability between the peptide sequence and HLA class I and / or class II.

[0055] The neoantigen prediction device 100 can use peptides derived from mutations in cancer tissues of an organism as neoantigens, and predict the binding force between the mutation-derived peptides and specific HLA class I and / or class II alleles of the organism, so that the organism's immune system can recognize and attack the neoantigens of the peptides.

[0056] The neoantigen prediction apparatus 100 can determine whether mutation-derived peptides having various lengths are neoantigens by pre-processing peptide sequences.

[0057] The new antigen prediction device 100 can fold the sequence values ​​of a partial region of the peptide sequence to unify the length of the peptide into a predetermined unit length. In the case of a peptide, natural folding occurs, and even in the peptide region, a folded region and an unfolded region can be determined. More specifically, it can be seen that the peptide spontaneously folds in the middle part. The new antigen prediction device 100 can fix the region in the peptide sequence where the probability of folding is relatively low so that it is not folded. The new antigen prediction device 100 can determine the fixed region in the peptide sequence that is not unfolded based on data about folding each peptide sequence.

[0058] The sequence value of the region to be folded in the peptide sequence can be calculated by applying a predetermined type of weight value. The weight value can be determined according to the number of folding times, the number of values ​​combined during the folding process, etc. For example, if folding is performed 4 times, a weight value of 4 can be applied to the value of the folded region. When five values ​​are combined and converted into one value, a weight value of 5 can be applied to the value of the folded region.

[0059] The neoantigen prediction apparatus 100 can predict the combinability corresponding to the binding force of the peptide sequence by pre-processing a complex of specific HLA class I and class II alleles of the organism. According to one embodiment, the neoantigen prediction apparatus 100 can infer the combinability corresponding to the binding force of the peptide sequence after pre-processing the HLA class I and class II alleles.

[0060] The new antigen prediction device 100 can calculate the immunogenicity prediction value using the immunogenicity prediction model trained by inputting the data obtained by folding the peptide sequence. The new antigen prediction device 100 can output the combinability prediction value of the peptide sequence using the combinability prediction model trained by inputting the data obtained by folding the peptide sequence.

[0061] In addition, the new antigen prediction device 100 can use a method such as heat treatment to input data encoded with HLA class I and / or class II alleles to generate a combinatorial prediction model. The new antigen prediction device 100 can input HLA class I and / or class II alleles into the combinatorial prediction model to output the combinatorial prediction value of HLA class I and / or class II alleles. The new antigen prediction device 100 can use the immunogenicity prediction model to output the immunogenicity prediction value, use the combinatorial prediction model to output the combinatorial prediction value, and use the new antigen prediction model for the immunogenicity prediction value and the combinatorial prediction value to output data about the new antigen.

[0062] The first data input unit 111 inputs a peptide sequence extracted from a target cancer tissue or cancer cell, or inputs a selected peptide sequence from a plurality of peptides extracted from a target cancer tissue or cancer cell.

[0063] The first data input unit 111 may perform one-hot encoding on the input peptide sequence and then generate folding data. More specifically, the first data input unit 111 may determine whether the length of the input peptide sequence exceeds a predetermined unit length. The first data input unit 111 may fold a peptide sequence that exceeds the unit length and normalize the folded peptide sequence to the unit length.

[0064] Regarding folding, as shown in FIG3 , the third to eleventh sequence values ​​CD1 fixed in the 12-mer peptide sequence PD1 except P1, P2 and P12 can be processed into the 11-mer peptide sequence PD2 by folding. In order to reduce the eleventh sequence value in the 12-mer peptide sequence PD1, the third and fourth sequence values ​​of PD1 can be combined and converted to the third sequence value. More specifically, the third and fourth sequence values ​​of PD1 can be combined, or the third or fourth sequence value of PD1 can be converted to the third sequence value. The value obtained by combining the third sequence value and the fourth sequence value of PD1 by applying a predetermined weight value can be converted to the third sequence value.

[0065] The 4th sequence value and the 5th sequence value of PD1 may be combined to convert to the 4th sequence value. The 5th sequence value and the 6th sequence value of PD1 may be combined to convert to the 5th sequence value. The 6th sequence value and the 7th sequence value of PD1 may be combined to convert to the 6th sequence value. The 7th sequence value and the 8th sequence value of PD1 may be combined to convert to the 7th sequence value. The 8th sequence value and the 9th sequence value of PD1 may be combined to convert to the 8th sequence value. The 9th sequence value and the 10th sequence value of PD1 may be combined to convert to the 9th sequence value. The 10th sequence value and the 11th sequence value of PD1 may be combined to convert to the 10th sequence value.

[0066] The 11-mer peptide sequence PD2 can be converted into the 10-mer peptide sequence PD3 by folding. The 10-mer peptide sequence PD3 can be converted into the 9-mer (9mer) peptide sequence PD4 by folding. The 9-mer peptide sequence PD4 can be converted into the 8-mer peptide sequence PD5 by folding.

[0067] The second data input unit 112 inputs the sequence of HLA class I and / or class II to confirm whether the peptide is a new antigen. The second data input unit 112 may encode the sequence of HLA class I and / or class II using a single-hot method or the like.

[0068] The immunogenicity prediction unit 120 may output the immunogenicity prediction value of the peptide sequence based on the data of the folded peptide sequence.

[0069] The immunogenicity prediction unit 120 may obtain the immunogenicity value of one or more amino acid sequences contained in the folded peptide sequence from an external database, and the immunogenicity prediction value of the peptide sequence for the combination of immunogenicity values ​​may be output to the immunogenicity prediction model trained by machine learning.

[0070] The immunogenicity prediction model can be machine-learned using input data of one or more amino acid sequences contained in the folded peptide sequence and output data of values ​​corresponding to the immunogenicity of the peptide sequence. Alternatively, the immunogenicity prediction model can be machine-learned by using the immunogenicity values ​​of one or more amino acid sequences contained in the folded peptide sequence as input data and the values ​​corresponding to the immunogenicity of the peptide sequence as output data.

[0071] The immunogenicity prediction model can be specifically learned by supervised learning, but is not limited thereto, and can be learned by unsupervised learning or reinforcement learning. The immunogenicity prediction model can output the immunogenicity prediction value of the peptide sequence by learning the input data using one or more neural networks.

[0072] The immunogenicity prediction model can be implemented to output an immunogenicity prediction value corresponding to a peptide sequence based on learned data, even if one or more amino acid sequences contained in the folded peptide sequence do not have learning data. The immunogenicity prediction model can infer the immunogenicity prediction value of an unlearned peptide sequence by inputting the immunogenicity data of each amino acid sequence contained in the peptide sequence. The immunogenicity prediction model can infer the immunogenicity prediction value of an unlearned peptide sequence based on the immunogenicity value of each of the multiple amino acid sequences contained in the peptide sequence.

[0073] More specifically, the immunogenicity prediction model can fold the embedded and encoded data to separate the peptide sequence into amino acid sequence units, and convert the sequence units into values ​​according to the presence or absence of each amino acid sequence. For example, the peptide sequence can be converted into the value (0 or 1) of the first to nth amino acid sequence. The converted value can be the input data of the immunogenicity prediction model.

[0074] The immunogenicity prediction model is learned using input data corresponding to the folded peptide sequence and output data including the immunogenicity value of the peptide sequence, wherein the immunogenicity value of the peptide sequence can be determined as the immunogenicity value of one or more amino acid sequences contained in the peptide sequence. The immunogenicity value of one or more amino acid sequences contained in the peptide sequence or the immunogenicity value of the peptide sequence can be obtained from pre-stored data.

[0075] The immunogenicity value of the new peptide sequence can be additionally input into the immunogenicity prediction model and used for immunogenicity prediction model learning. By inputting new peptide sequences and their immunogenicity values, the immunogenicity prediction model can be continuously upgraded.

[0076] The combinability prediction unit 130 may input the peptide sequence obtained from the first data input unit 111 and the HLA class I and / or class II allele sequence obtained from the second data input unit 112, and output a combinability prediction value corresponding to the binding force between the peptide sequence and the HLA class I and / or class II allele sequence.

[0077] The combinability prediction unit 130 may output a combinability prediction value corresponding to the binding force between the folded peptide sequence and the heat-treated HLA class I and / or class II allele sequence.

[0078] The combinability prediction model can be used to output a combinability prediction value. The combinability prediction model can use data about folded peptide sequences and data about heat-treated HLA allele sequences as input data, extract feature values ​​corresponding to the input data, and output a combinability prediction value based on the feature values.

[0079] By converting one or more amino acid sequences contained in a peptide sequence or HLA allele sequence into respective corresponding values, a unique heat encoding method can be performed. In addition, according to an embodiment of the present invention, a peptide sequence can be uniquely encoded, and the uniquely encoded peptide sequence can be folded. Experimentally and empirically, values ​​can be predefined for all amino acid sequences contained in a peptide sequence. For example, multiple alignments can be performed to a plurality of peptide sequences by position, the peptide sequences can be arranged on the x-axis and the y-axis, and positions with high similarity and positions with low similarity can be extracted from the peptide sequence based on the similarity between the peptide sequence on the x-axis and the peptide sequence on the y-axis. At this point, the position can be determined based on standard similarity. In addition, a processing method for unifying the length of the peptide can be performed. An example of this processing method can be a folding method.

[0080] The folding method may refer to combining sequence values ​​of predetermined regions in a peptide sequence that has been subjected to a single heat treatment according to a predetermined rule to reduce the length of the peptide sequence.

[0081] The combinability prediction model can input data about peptide sequences processed by the folding method and data about HLA allele sequences to output a combinability prediction value.

[0082] The combinability prediction model can extract one or more feature values ​​that can be used as input data corresponding to the peptide sequence and the HLA allele sequence, and output the combinability between the peptide sequence and the HLA allele sequence. The type of feature value to be extracted can be determined based on the results of machine learning, which takes a combination of the combinability values ​​of the amino acid sequences contained in the peptide sequence as input data and takes a value corresponding to the combinability between the peptide sequence and the HLA allele sequence as output data. The combinability prediction model can extract key feature values ​​by learning the input data using one or more neural networks, and output a combinability prediction value based on the key feature values.

[0083] Even if the input peptide sequence has no learning data, the combinatorial prediction model can output the combinatorial prediction value of the new peptide sequence based on the data learned from other peptide sequences.

[0084] The combinability prediction model can fold the length of the peptide sequence according to a predetermined rule, and then output a combinability feature value corresponding to the folded peptide sequence. The combinability prediction model can encode and fold the peptide to output the combinability feature value of the peptide sequence.

[0085] The combinability prediction model can use values ​​corresponding to encoded HLA class I and / or class II allele sequences as input data and use combinability feature values ​​corresponding to HLA class I and / or class II allele sequences as output data.

[0086] The combinability prediction unit 130 may input data on a peptide sequence and HLA class I and / or class II alleles, and output a combinability prediction value corresponding to the binding force between the peptide sequence and the HLA class I and / or class II alleles.

[0087] In another embodiment, the combinatorial prediction unit 130 may output a combinatorial prediction value using a combinatorial prediction model. The combinatorial prediction model is learned by machine learning. As a result of the learning, the types of combinatorial characteristic values ​​of the peptide sequence and the combinatorial characteristic values ​​of the HLA class alleles may also be determined. In the combinatorial prediction model, the combinatorial prediction value used as output data may be a value of the binding force between the peptide sequence and the HLA class I and / or class II alleles obtained empirically or experimentally, or a value from an external database.

[0088] The neoantigen prediction unit 140 may output neoantigen data including the probability of the presence of a neoantigen in a peptide sequence, what the neoantigen is, etc. based on the immunogenicity prediction value and / or the combinability prediction value.

[0089] The neo-antigen prediction unit 140 may output neo-antigen data using a neo-antigen prediction model trained by inputting an immunogenicity prediction value and a combinability prediction value.

[0090] The new antigen prediction model can be learned by using the immunogenicity prediction value and / or the combinability prediction value as input data and the new antigen data of the peptide sequence of HLA class I and / or class II allele as output data, but is not limited thereto, and can be unsupervised learning or reinforcement learning. The new antigen data may include the presence or absence of the peptide sequence of the new antigen, the location of the new antigen, the specificity of the new antigen, the probability of the new antigen at each location, etc.

[0091] The immunogenicity prediction value and the combinability prediction value may include one or more values ​​of 0 or 1, but are not limited thereto and may include various defined values ​​between 0 and 1.

[0092] The neoantigen prediction device 100 according to an embodiment of the present invention can output neoantigen data from a peptide sequence derived from a target cancer tissue mutation, including the presence or absence of a peptide sequence of a neoantigen, a major position in the neoantigen, the specificity of the neoantigen, etc. Specifically, the significance of the neoantigen prediction device 100 is that it determines whether a peptide sequence is a neoantigen of an HLA class I and / or class II allele.

[0093] Figure 2 is an example diagram of a learning model 200 according to an embodiment of the present invention.

[0094] The learning model 200 may include a neoantigen prediction model M that inputs a peptide sequence i1 and an HLA class I and / or class II allele i2 to output immunogenicity data of the peptide.

[0095] The peptide sequence i1 can be input into the new antigen prediction model M after the single heat treatment P11 and the folding treatment P12. The peptide sequence after the single heat treatment is as follows Figure 4a As shown in FIG. 1 . The folding process P12 may convert the combined value into a first value calculated according to a predetermined rule, and apply the weight value to the calculated first value to calculate the final second value. Figure 4b As shown. The predetermined rule may be to add the combined values, but is not limited thereto, and may also be calculated as a weighted average, an average value, etc. In the learning model 200, the two values ​​of the peptide sequence i1 may be folded once until they become a unit length (see Figure 3a ). The final second value is obtained by applying the weight value to the first value, such as Figure 4c The weight value applied at this time may be a value proportional to the number of combined values ​​or the number of folding times, but is not limited thereto and may be determined by combining various values.

[0096] HLA class I and / or class II allele i2 can be input into the neoantigen prediction model M after single heat treatment P21.

[0097] The new antigen prediction model M can input a peptide sequence i1 treated with P11 and P12 and an HLA class I and / or class II allele i2 treated with P21 to output data on the immunogenicity of the peptide sequence i1. At this time, the length of the peptide sequence i1 can be folded into a predetermined unit length, such as 8-mer. The peptide sequence i1 can be subjected to a unit length folding process for regions other than the fixed region after being subjected to a single heat treatment. The unfolded fixed region may include the first, second, and last values, but is not limited thereto, and may include a critical region. The HLA class I allele i2 can be subjected to a single heat treatment based on a predetermined amino acid sequence.

[0098] The neoantigen prediction model M may be designed by including layers of neural networks NN 256, NN 128, and NN 64. As a result of learning by inputting a training data set into the neoantigen prediction model M, the layers of NN 256, NN 128, and NN 64 may be completed. Here, the number of neurons and the layer structure in the NN are not limited to the above examples, and may have different forms.

[0099] A neural network may be a set of algorithms that extracts various attribute information of input data using the results of statistical machine learning and identifies and / or determines objects in an image based on the extracted attribute information. A neural network may be implemented using software or an engine, etc., to execute a set of algorithms. A neural network implemented using software or an engine, etc. may be executed by a processor in a device or a processor on a server.

[0100] The new antigen prediction model M may include an immunogenicity prediction model that infers the immunogenicity of a peptide sequence and a combinability prediction model. The combinability prediction model may infer the combinability between a peptide sequence and an HLA class I and / or class II allele.

[0101] The neoantigen prediction model M according to the embodiment can output data about the neoantigen as an output value by using a model trained by inputting the output values ​​of the immunogenicity prediction model and the combinability prediction model between HLA class I and / or class II and peptides.

[0102] The learning model 200 may be implemented in software or hardware, or as a combination of software and hardware.

[0103] The learning model 200 may be provided in the neoantigen prediction device 100, or may be provided in a device separate from the neoantigen prediction device 100. The learning model 200 may be connected via a network or electrically connected to a plurality of neoantigen prediction devices. The learning model 200 may receive data required for learning from an external device. The learning model 200 may transmit the learned prediction model to the neoantigen prediction device to output the neoantigen data of the peptide sequence.

[0104] Figure 3a is an exemplary diagram of a peptide sequence processed by a method according to an embodiment of the present invention. Figure 3b An example diagram of input data processed by conventional methods.

[0105] The length of the peptide sequence PD1 after the single heat treatment exceeds the unit length of 20×12 and needs to be folded. The values ​​of the predetermined fixed regions P1, P2, and P12 in the peptide sequence PD1 can be fixed without folding. The fixed regions can be set differently for each peptide sequence. The fixed regions are determined based on the probability of folding in each peptide sequence and can be set to regions with a lower probability of folding. By fixing some regions and then folding them, the immunogenicity and combinability of peptide sequences can be predicted more accurately.

[0106] If the region CD1 other than the fixed region in the peptide sequence PD1 is folded once, the region CD1 is the same as PD2. The third to tenth values ​​CD2 are calculated in such a way that the third and fourth values ​​are folded and calculated as the third value. The length of the peptide sequence after folding once may become 20×11.

[0107] If the region CD2 other than the fixed region in the peptide sequence PD2 is folded once again, the region CD2 is the same as PD3. The third to ninth values ​​CD3 are calculated in such a way that the third and fourth values ​​after folding once are folded and calculated as the third value. The length of the peptide sequence after folding twice may become 20×10.

[0108] If the region CD3 other than the fixed region in the peptide sequence PD3 is folded once again, the region CD3 is the same as PD4. The third to eighth values ​​CD4 are calculated in such a manner that the third and fourth values ​​after folding twice are folded together and calculated as the third value. The length of the peptide sequence after folding three times can be changed to 20×9.

[0109] If the region CD4 other than the fixed region in the peptide sequence PD4 is folded once more, the region CD4 is the same as PD5. The third to seventh values ​​CD5 are calculated by folding the third and fourth values ​​after folding three times as the third value. The length of the peptide sequence after folding four times can be changed to 20×8.

[0110] exist Figure 3a In the present invention, the fixed regions are P1, P2, and P12, but in order to increase the probability of predicting immunogenicity, the fixed regions can be changed according to the characteristics of the target cancer, HLA class I, or peptide sequence.

[0111] The folding process can also be altered to increase the probability of predicting immunogenicity.

[0112] When applying the method of extracting the sequence with the largest diagonal sum after taking the inner product of the PPSM matrix of conventional peptide sequences and MHC data, there may be a problem of inevitably losing information on one of the two anchor points. In addition, when using the PSSM matrix, there is a problem of manually preprocessing the data whenever new reference data is updated. Figure 3b As shown, when the method of extracting the sequence with the largest diagonal sum after taking the inner product of the PPSM matrix of peptide sequence and MHC data is applied, it can be seen that the tenth value, the first value, and the first to third values ​​are lost (see OD).

[0113] Figure 4a is an exemplary diagram of a one-hot encoded peptide sequence according to an embodiment of the present invention.

[0114] Figure 4b is an exemplary diagram of a folded peptide sequence according to an embodiment of the present invention.

[0115] Figure 4c is an example diagram of a peptide sequence to which weight values ​​are applied according to an embodiment of the present invention.

[0116] exist Figure 4a In the peptide sequence P41, Figure 4b P42 can be generated by changing the remaining area D41 except the predetermined fixed area to the value D42 combined according to the predetermined rule. At this time, in order to combine the 12×20 matrix into the 8×20 matrix, it is necessary to combine five values ​​into one value. Figure 4b As shown, the third sequence value can be determined by combining the third to seventh sequence values ​​according to a predetermined rule. The predetermined rule can be one of a sum value, a weighted average value, and an average value. Optionally, after combining according to the predetermined rule, P43 can be generated by applying each weight value to each sequence value. Figure 4c As shown, processed data P43 and D43 can be generated by applying a multiple of 5 to each combined value of the peptide sequence. The multiple of 5 is just an example, and other multiples can be applied. Figures 4a to 4c In the example, the 12×20 peptide sequence is changed to an 8×20 peptide sequence with a unit length of 8, but peptide sequences of various sizes such as 11×20, 10×20, 13×20, etc. can be changed to a predetermined unit length.

[0117] Figure 5 is a flow chart of a method for predicting immunogenicity according to an embodiment of the present invention.

[0118] In S110, the neoantigen prediction apparatus 100 inputs a peptide sequence extracted from a target cancer tissue. The neoantigen prediction apparatus 100 may determine whether the length of the peptide sequence exceeds a predetermined unit length.

[0119] In S120, when the length of the peptide sequence exceeds a predetermined unit length, the new antigen prediction device 100 can set the remaining region in the peptide sequence except for the predetermined fixed region, and fold the value of the region exceeding the unit length of the remaining region to shorten the peptide sequence to the unit length (S130).

[0120] In S140 , the neoantigen prediction apparatus 100 may input a sequence of a complex of HLA class I alleles, and process the sequence of the complex using a one-hot processing method.

[0121] In S150, the neoantigen prediction apparatus 100 may extract feature values ​​from input data including sequence values ​​contained in the peptide sequence and output an immunogenicity prediction value of the feature values. At this time, the neoantigen prediction apparatus 100 may output an immunogenicity prediction value using a learned immunogenicity prediction model.

[0122] In S160 , the neoantigen prediction apparatus 100 may output a combinability prediction value by performing one-hot processing on data obtained via the folded peptide sequence and the HLA class I and / or class II allele sequence and inputting the processed sequence into a combinability prediction model.

[0123] In S170 , the neo-antigen prediction apparatus 100 may output neo-antigen data of peptide sequences of immunogenicity prediction values ​​and / or combinability prediction values ​​using the immunogenicity prediction model.

[0124] Figure 6 is a diagram comparing a neoantigen prediction model according to an embodiment of the present invention with a conventional neoantigen prediction model.

[0125] It was experimentally concluded that the AUC of the results predicted by inputting the data of the folded peptide sequence was 0.822, which was significantly better than the AUC value of 0.720 of the results predicted by inputting the data of the peptide sequence processed by other processing methods.

[0126] The new antigen prediction device 100 can output the immunogenicity prediction value of the peptide sequence according to a combination of one or more amino acid sequences contained in the folded peptide sequence and the immunogenicity value corresponding to the one or more amino acid sequences.

[0127] The immunogenicity prediction device 100 can obtain the immunogenicity value of one or more amino acid sequences contained in the peptide sequence from an external database, and can output the immunogenicity prediction value of the peptide sequence for the combination of immunogenicity values ​​to the immunogenicity prediction model trained by machine learning.

[0128] The immunogenicity prediction model can be machine-learned using input data of one or more amino acid sequences contained in the peptide sequence and output data of values ​​corresponding to the immunogenicity of the peptide sequence. Alternatively, the immunogenicity prediction model can be machine-learned by combining the immunogenicity values ​​of one or more amino acid sequences contained in the peptide sequence as input data with the values ​​corresponding to the immunogenicity of the peptide sequence as output data, specifically supervised learning, but not limited to this, and unsupervised learning or reinforcement learning can be performed. The immunogenicity prediction model can extract key feature values ​​by learning the input data with one or more neural networks to output immunogenicity prediction values. The immunogenicity prediction value may include one or more immunogenicity prediction values.

[0129] The immunogenicity prediction model can output an immunogenicity prediction value corresponding to a peptide sequence based on the learned data, even if one or more amino acid sequences included in the peptide sequence have no learned data.

[0130] The immunogenicity prediction model can perform an embedding process to separate the peptide sequence into amino acid sequence units and convert the sequence units into values ​​according to the presence or absence of each amino acid sequence. For example, the peptide sequence can be converted into values ​​(0 or 1) for the first to nth amino acid sequences. The converted values ​​can be the input data of the immunogenicity prediction model.

[0131] The immunogenicity prediction model is learned using input data corresponding to a plurality of peptide sequences and output data including immunogenicity values ​​of the peptide sequences, wherein the immunogenicity values ​​of the peptide sequences can be determined as the immunogenicity values ​​of one or more amino acid sequences contained in the peptide sequences. The immunogenicity values ​​of the peptide sequences can be obtained from pre-stored data. The immunogenicity values ​​of new peptide sequences can be additionally input into the immunogenicity prediction model and used to allow the immunogenicity prediction model to learn. The immunogenicity prediction model can be updated by new peptide sequences and their immunogenicity values.

[0132] The combinability prediction model can use the folded peptide sequence as input data and extract feature values ​​corresponding to the input data. The type of feature value extracted from the combinability prediction model can be determined by machine learning a combination of the combinability value of the amino acid sequence contained in the peptide sequence as input data and the value corresponding to the combinability of the peptide sequence as output data.

[0133] The combinability prediction model can input data about peptide sequences processed by the folding method and values ​​corresponding to encoded HLA class I and / or class II allele sequences to output a combinability prediction value.

[0134] The composability prediction model can extract key feature values ​​by learning input data using one or more neural networks, and output a composability prediction value based on the key feature values.

[0135] The combinability prediction model can output a combinability prediction value by outputting a combinability feature value of a peptide sequence based on learned data, even if there is no learning data of the peptide sequence.

[0136] The composability processing model can fold the peptide sequence to unify the length of the peptide sequence into a unit length, then calculate the amino acid usage characteristics of the binding / non-binding peptides of each length, and extract its corresponding peptide feature value. In another embodiment, the composability prediction model can convert the peptide sequence into a vector using one-hot encoding, and pass the vector through multiple neural networks to output a composability prediction value.

[0137] In an optional embodiment, the combinatorial prediction model includes one or more prediction models, each of which uses the α chain or β chain of the HLA class I allele or class II allele as input data. Each prediction model can output a combinatorial feature value corresponding to the HLA class I and / or class II allele. The combinatorial prediction model can extract key feature values ​​by learning the input data using one or more neural networks, and output a combinatorial prediction value between the peptide and the HLA allele corresponding to the key feature value.

[0138] The combinatorial prediction model can extract the α chain sequence of the HLA class I or class II allele from the peptide sequence, and input the α chain sequence of the HLA class I or class II allele into the HLAα processing model to extract the HLA-α combinatorial feature value. The combinatorial prediction model can extract the HLA-β combinatorial feature value by inputting the β chain sequence of the HLA class II allele into the HLAβ processing model.

[0139] The combinatorial prediction model can be a model trained by inputting peptide feature values ​​extracted from peptide sequences, HLA-α combinatorial feature values ​​extracted from α chain sequences of HLA class I or class II alleles, and HLA-β combinatorial feature values ​​extracted from β chain sequences of HLA class II alleles, and outputting a combinatorial prediction value corresponding to the binding force.

[0140] The combinatorial prediction model can use a peptide processing model to extract peptide feature values ​​from data processed to shorten the peptide sequence to a unit length. The combined prediction model can perform a single-heat treatment on the α chain sequence or β chain sequence of the HLA class I or class II allele, extract the HLA feature value corresponding to the processed chain sequence, and output a combinatorial prediction value through a combination of the peptide and the HLA feature value.

[0141] The neoantigen prediction apparatus 100 can output data on immunogenicity by combining the immunogenicity prediction value and the combinability prediction value of the peptide and HLA.

[0142] The new antigen prediction device 100 can output new antigen data including the probability of the existence of new antigens in peptide sequences, the positions of new antigens, the probability of the occurrence of new antigens by position, etc. based on at least one of the immunogenicity prediction values ​​and the combinability prediction values ​​of peptides and HLA.

[0143] The neoantigen prediction apparatus 100 may output neoantigen data using a neoantigen prediction model trained by inputting an immunogenicity prediction value and / or a combinability prediction value of a peptide and HLA.

[0144] The new antigen prediction model can be learned using the immunogenicity prediction values ​​and / or combinability prediction values ​​of peptides and HLA as input data, and using the new antigen data of peptide sequences of HLA class I and / or class II alleles as output data, but is not limited thereto, and can perform unsupervised learning or reinforcement learning.

[0145] The immunogenicity prediction value and / or the combinability prediction value of the peptide and HLA may include one or more values ​​of 0 or 1, but is not limited thereto, and may include various defined values.

[0146] The above-mentioned equipment can be implemented in a combination of hardware components, software components and / or hardware components and software components. For example, the equipment and components described in the embodiment can be implemented using, for example, one or more general-purpose computers or special-purpose computers (such as processors, controllers, arithmetic logic units (ALUs), digital signal processors, microcomputers, field programmable gate arrays (FPGAs), programmable logic units (PLUs), microprocessors or other devices that can execute and respond to instructions). The processing device can execute an operating system OS and one or more software applications executed on the operating system. In addition, the processing device can also access, store, manipulate, process and generate data in response to the execution of software. For ease of understanding, it can be described as using a processing device, but it can be seen by those skilled in the art that the processing device may include multiple processing elements and / or multiple types of processing elements. For example, the processing device may include multiple processors or a processor and a controller. In addition, other processing configurations, such as parallel processors are also possible.

[0147] Software may include computer programs, codes, instructions, or one or more combinations thereof, and may configure a processing device to operate as desired, or instruct a processing device, either individually or collectively. Software and / or data are interpreted by a processing device, or may be embodied, permanently or temporarily, in any type of machine, component, physical device, virtual device, computer storage medium or device, or signal wave to be transmitted, in order to provide instructions or data to a processing device. Software may be distributed on computer systems connected via a network, and may be stored or executed in a distributed manner. Software and data may be stored in one or more computer-readable recording media.

[0148] The method according to the embodiment can be implemented in the form of program instructions, and the program instructions can be executed by various computer means to be recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., alone or in combination. The program instructions recorded in the medium may be specially designed and configured for the embodiment, or they may be known and used by technicians in the field of computer software. Examples of computer-readable media include magnetic media, such as hard disks, floppy disks, and tapes, optical media, such as CD-ROMs and DVDs, magneto-optical media, such as optical disks, and hardware devices, such as ROMs, RAMs, and flash memories, which are specially configured to store and execute program instructions. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, etc., and machine language codes created by a compiler. The hardware device can be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

[0149] As described above, although the embodiments have been described through limited embodiments and drawings, those skilled in the art can make various modifications and variations based on the above description. For example, even if the technology is performed in a different order from the method, and / or the above-mentioned systems, structures, devices, circuits, etc. are coupled or combined in a different form from the method, or replaced or substituted by other components or equivalents, appropriate results can be achieved.

[0150] Accordingly, other implementations, other embodiments, and equivalents of the appended claims are within the scope of the following claims.

Claims

1. A method for predicting new antigens by processing peptides of various lengths through folding, the method comprising: The new antigen prediction device inputs a peptide sequence extracted from a target cancer tissue and determines whether the length of the peptide sequence exceeds a predetermined unit length; When the peptide sequence exceeds the unit length, determining a remaining region in the peptide sequence except for a predetermined fixed region, and processing the remaining region so that a value of the region exceeding the unit length is folded to shorten the peptide sequence to the unit length; The sequence of the HLA class I or class II allele complex is input by the new antigen prediction device; Extracting, by the neoantigen prediction device, a feature value from input data including a sequence value included in the processed peptide sequence, and outputting an immunogenicity prediction value of the feature value; extracting, by the neoantigen prediction device, peptide feature values ​​from the data processed to shorten the peptide sequence to the unit length; The neoantigen prediction device performs a heat treatment on the α chain sequence or the β chain sequence of the HLA class I or class II allele, and extracts the HLA feature value corresponding to the processed chain sequence; The new antigen prediction device outputs a combinability prediction value by combining the peptide feature value and the HLA feature value; and The neo-antigen prediction device determines the neo-antigen of the peptide sequence based on the immunogenicity prediction value or the combinability prediction value.

2. The method of claim 1 , wherein shortening the peptide to the unit length is processed by a folding process in which a first value at a first position and a second value at a second position in the remaining region of the peptide are folded to be expressed as the sum of the first value and the second value, and the length of the peptide is reduced by 1.

3. The method according to claim 1, wherein the fixed region of the peptide comprises a first value, a second value and a last value, and The values ​​contained in the fixed area are not folded.

4. The method of claim 2, wherein the folding process is repeated until the peptide reaches the unit length.

5. The method according to claim 1, wherein the outputting the combinability prediction value is performed by outputting the combinability prediction value using a combinability prediction model, and The combinatorial prediction model is a model trained in the following manner: inputting the peptide feature value extracted from the peptide sequence, the HLA-α feature value extracted from the α chain sequence of the HLA class I or class II allele, and the HLA-β feature value extracted from the β chain sequence of the HLA class II allele, and outputting the combinatorial prediction value corresponding to the binding force.

6. The method according to claim 5, wherein the combinatorial prediction model further comprises a peptide processing model for extracting the peptide feature value, an HLA α processing model for extracting the HLA-α feature value, and an HLA β processing model for extracting the HLA-β feature value, and The combinatorial processing models of the peptide, the HLAα and the HLAβ are encoders of different encoder forms, respectively.

7. The method according to claim 1, wherein the outputting the immunogenicity prediction value is performed by outputting the immunogenicity prediction value using an immunogenicity prediction model, and The immunogenicity prediction model is a model trained by inputting the feature value extracted from the peptide sequence and outputting the immunogenicity prediction value corresponding to the immunogenicity of the peptide sequence.

8. The method of claim 1, wherein the target cancer tissue comprises cells engineered to express a single HLA class I or class II allele.

9. The method of claim 1, wherein the target cancer tissue is obtained from a plurality of organisms or comprises mutant cells derived therefrom.

10. The method of claim 1, wherein the target cancer tissue comprises fresh or frozen tumor cells obtained from a plurality of organisms.

11. The method of claim 1, wherein the target cancer tissue comprises fresh or frozen tissue cells obtained from a plurality of organisms.

12. A computer program stored in a computer-readable recording medium for executing the method according to any one of claims 1 to 11 using a computer.

Images