T cell epitope high-throughput screening method and device based on deep learning framework

Abstract

The invention relates to the technical field of biological information prediction, and provides a T cell epitope high-throughput screening method and device based on a deep learning framework, and the method comprises the steps: extracting target input data according to an input T cell receptor sequence and an antigen sequence; wherein the target input data comprises a T cell receptor CDR3 beta sequence and at least one antigen epitope peptide fragment sequence; inputting the target input data into a pre-constructed prediction model to obtain a prediction result; and screening according to the prediction result to obtain a screening result. According to the method, a T cell receptor sequence and an antigen sequence which are not subjected to standardization treatment are allowed to be input, the CDR3 beta sequence can be efficiently extracted, then the binding activity prediction work is effectively carried out, the data preprocessing workload of an operator is reduced while the universality and the stability are improved, and the detection efficiency is improved. Meanwhile, the T cell receptor and epitope binding probability is predicted based on the deep neural network, and screening work is effectively completed.

Description

Technical Field

[0001] The present invention relates to the field of bioinformation prediction technology, and in particular to a method and device for high-throughput screening of T cell epitopes based on a deep learning framework. Background Art

[0002] The interaction between T cell receptor (TCR) and antigen epitope is the basis of immune response. In adaptive immune response, T cells recognize pathogen-derived peptides (i.e., epitopes) presented by major histocompatibility complex (MHC) molecules through their T cell receptors, thereby triggering cell-mediated immune response, activating relevant immune cells and clearing infected cells. Obtaining T cell antigens and analyzing the interaction between TCR and antigen epitopes will help us understand how the immune system recognizes and attacks abnormal cells such as infections and tumors, and how T cells participate in the development of human diseases, which will help develop more effective immunotherapies, such as TCR gene therapy, tumor immune checkpoint inhibitors, etc., and design more selective and affinity drugs. In addition, based on the interaction between TCR and antigen epitopes, disease diagnosis can be performed by detecting disease-specific TCR-antigen epitopes, and personalized immunotherapy can be achieved by identifying the interaction between patient-specific TCR and tumor-associated antigens.

[0003] However, the process of directly experimentally determining the interaction between TCR and epitope is both complicated and costly. To address this problem, a series of computational methods have been developed to predict and simulate the interaction between TCR and epitope. In the process of adaptive immune recognition, epitopes are presented to the cell surface by MHC molecules and recognized by TCR. The core binding region of TCR to epitope and MHC molecules is located between the complementary determining region 3 (CDR3β) of the TCRβ chain and the epitope. Although the interaction between TCR and epitope belongs to a special protein-peptide interaction, there are machine learning-based methods or molecular docking-based methods that can be used to predict or analyze specific contact sites for protein-peptide interactions, but these methods are not specifically designed for TCR-epitope interactions, and their application effects still need to be evaluated. In addition, there are also many computational tools that use various machine learning algorithms to predict whether a given CDR3β and epitope sequence pair can bind, but most of these tools can only predict whether the two bind, and fail to reveal the deeper interaction details between TCR and epitope.

[0004] Furthermore, in actual high-throughput screening applications, the input antigen sequence may be a longer original sequence or a complete protein sequence, and it is impossible to determine which part of the protein is the antigenic peptide that can produce interaction. Existing TCR and epitope interaction prediction tools only support the input of short peptide sequences within a specified length range and are not suitable for inputting original sequences or complete protein sequences.

[0005] In summary, the existing high-throughput screening of T cell epitopes has problems with poor technical versatility and stability. Summary of the invention

[0006] The present invention provides a method and device for high-throughput screening of T cell epitopes based on a deep learning framework, so as to solve the defects of poor technical versatility and stability in the prior art and realize high-throughput screening of T cell epitopes with good technical versatility and stability.

[0007] The present invention provides a high-throughput screening method for T cell epitopes based on a deep learning framework, comprising the following steps: Extracting target input data according to the input T cell receptor sequence and antigen sequence; wherein the target input data includes a T cell receptor CDR3β sequence and at least one antigen epitope peptide sequence; Inputting the target input data into a pre-built prediction model to obtain a prediction result; wherein the prediction result at least includes the predicted binding probability of the T cell receptor CDR3β sequence and the antigen epitope peptide sequence, and the prediction model is obtained by training a deep neural network using T cell receptor CDR3β sequence samples and antigen epitope peptide sequence samples; The screening results are obtained by screening according to the prediction results.

[0008] According to a T cell epitope high-throughput screening method provided by the present invention, target input data is extracted based on the input T cell receptor sequence and antigen sequence, specifically comprising: Filtering illegal characters on the T cell receptor sequence and the antigen sequence to obtain legal T cell receptor sequence and antigen sequence; When the legal T cell receptor sequence is a complete β sequence, the T cell receptor variable region sequence is numbered and annotated using a genome library comparison algorithm based on a hidden Markov model, and the T cell receptor CDR3β sequence is extracted according to the numbering and the annotation; When the legal antigen sequence length is greater than the length threshold, the antigen sequence is segmented into a series of antigen epitope peptide sequences according to the specified length.

[0009] According to a T cell epitope high-throughput screening method provided by the present invention, the prediction model includes a feature extractor, an interaction extractor, an epitope autoencoder, a binding prediction module and a residue interaction prediction module; Accordingly, the target input data is input into a pre-built prediction model to obtain a prediction result, which specifically includes: Extracting a first feature and a second feature based on the feature extractor according to the target input data; forming an interaction feature graph by dimensional expansion according to the first feature and the second feature; Obtaining global epitope information based on the interaction extractor according to the epitope feature vector; wherein the epitope feature vector is extracted based on the epitope autoencoder according to a preset epitope database; extracting interaction information of pairs of residues based on the interaction extractor according to the interaction feature graph; aggregating pairwise interactions based on the binding prediction module according to the interaction information and the global epitope information to obtain a predicted binding probability; Obtaining a residue-level interaction score heat map according to the interaction information; A prediction result is obtained according to the interaction score heat map and the predicted binding probability.

[0010] According to a T cell epitope high-throughput screening method provided by the present invention, screening results are obtained according to the prediction results, specifically comprising: The T cell receptor CDR3β sequence and the antigen epitope peptide sequence whose predicted binding probability is greater than a preset binding probability judgment threshold are screened and output as screening results.

[0011] According to a T cell epitope high-throughput screening method provided by the present invention, the feature extractor is constructed based on an embedding layer and a one-dimensional convolutional neural network; The residue interaction prediction module is constructed based on a two-dimensional convolutional neural network.

[0012] According to a T cell epitope high-throughput screening method provided by the present invention, the prediction result also includes the target input data corresponding to the predicted binding probability and the input T cell receptor sequence and antigen sequence corresponding to the target input data.

[0013] The present invention also provides a T cell epitope high-throughput screening device based on a deep learning framework, comprising the following modules: An input unit, used to extract target input data according to the input T cell receptor sequence and antigen sequence; wherein the target input data includes a T cell receptor CDR3β sequence and at least one antigen epitope peptide sequence; A prediction unit, used for inputting the target input data into a pre-built prediction model to obtain a prediction result; wherein the prediction result at least includes the predicted binding probability of the T cell receptor CDR3β sequence and the antigen epitope peptide sequence, and the prediction model is obtained by training the T cell receptor CDR3β sequence samples and the antigen epitope peptide sequence samples based on a deep neural network; A screening unit is used to screen and obtain screening results according to the prediction results.

[0014] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the high-throughput screening method for T cell epitopes as described above is implemented.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the high-throughput screening method for T cell epitopes as described in any one of the above is implemented.

[0016] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the high-throughput screening method for T cell epitopes as described in any one of the above is implemented.

[0017] The method and device for high-throughput screening of T cell epitopes based on a deep learning framework provided by the present invention obtain target input data by extracting according to the input T cell receptor sequence and antigen sequence; wherein the target input data includes a T cell receptor CDR3β sequence and at least one antigen epitope peptide sequence; the target input data is input into a pre-constructed prediction model to obtain a prediction result; wherein the prediction result at least includes the predicted binding probability of the T cell receptor CDR3β sequence and the antigen epitope peptide sequence, and the prediction model is obtained by training the T cell receptor CDR3β sequence sample and the antigen epitope peptide sequence sample based on a deep neural network; the screening result is obtained by screening according to the prediction result. The present invention allows the input of T cell receptor sequences and antigen sequences that have not been normalized, and can efficiently extract their CDR3β sequences, thereby effectively performing binding activity prediction work, while improving versatility and stability while reducing the operator's data preprocessing workload, and at the same time predicting the binding probability of T cell receptors and epitopes based on a deep neural network, effectively completing the screening work. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 This is one of the schematic diagrams of the process of the high-throughput screening method for T cell epitopes provided by the present invention.

[0020] Figure 2 This is the second schematic diagram of the process of the high-throughput screening method for T cell epitopes provided by the present invention.

[0021] Figure 3 It is a schematic diagram of the framework of the artificial intelligence pre-model of the T cell epitope high-throughput screening method provided by the present invention.

[0022] Figure 4 It is a schematic diagram of the interaction sites between the two in the TCR-antigen epitope complex crystal structure of one embodiment of the T cell epitope high-throughput screening method provided by the present invention.

[0023] Figure 5 It is a schematic diagram of the structure of the T cell epitope high-throughput screening device provided by the present invention.

[0024] Figure 6 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] Combine the following Figure 1-Figure 4 Describe the high-throughput screening method of T cell epitopes based on the deep learning framework of the present invention, Figure 1 FIG. 1 is one of the flow diagrams of the high-throughput screening method for T cell epitopes provided by the present invention, such as Figure 1 As shown, the method includes: Step 110: extracting target input data based on the input T cell receptor sequence and antigen sequence; wherein the target input data includes a T cell receptor CDR3β sequence and at least one antigen epitope peptide sequence.

[0027] It needs to be explained that the prior art has high requirements on the format of input data, and the processed TCR-CDR3β sequence and antigen peptide data must be input for subsequent model or algorithm processing. However, in actual high-throughput screening applications, the input antigen sequence may be a longer original sequence or a complete protein sequence, and the staff is required to determine which part of the protein is the antigen peptide that can produce interactions, which is inefficient and has high labor costs. The present invention does not limit the state of the input T cell receptor sequence and antigen sequence, and can use unprocessed T cell receptor sequences and antigen sequences of any length including antigen peptides that can produce interactions, and has good technical versatility and stability.

[0028] After the input T cell receptor sequence and antigen sequence are obtained, target input data including the T cell receptor CDR3β sequence and at least one antigen epitope peptide sequence are extracted therefrom. The present invention will further illustrate the extraction step in subsequent embodiments.

[0029] It needs to be explained that when the input antigen sequence is long, the input antigen sequence can be cut to obtain multiple antigen epitope peptide sequences, and then the prediction model can be used to screen antigens that can interact among multiple sequences. For the convenience of description, multiple antigen epitope peptide sequences can form a polypeptide library for screening by the prediction model.

[0030] Step 120: Input the target input data into a pre-constructed prediction model to obtain a prediction result; wherein the prediction result at least includes the predicted binding probability of the T cell receptor CDR3β sequence and the antigen epitope peptide sequence, and the prediction model is obtained by training a deep neural network using T cell receptor CDR3β sequence samples and antigen epitope peptide sequence samples.

[0031] Specifically, in step 120, the extracted T cell receptor CDR3β sequence and antigen epitope peptide sequence are input into the artificial intelligence prediction model for prediction, and a prediction result including at least the predicted binding probability of the T cell receptor CDR3β sequence and the antigen epitope peptide sequence is obtained.

[0032] It can be understood that by using the complete amino acid sequence of the T cell receptor or the sequence of the CDR3 region of the β chain of the T cell receptor as the input data of the T cell receptor, and using the amino acid sequence or peptide sequence of the antigen protein as the input data of the antigen epitope, the screening of the epitope for the T cell receptor can be quickly completed.

[0033] The obtained prediction results at least include the predicted binding probability of the T cell receptor CDR3β sequence and the antigen epitope peptide sequence.

[0034] In addition, the prediction model is obtained by training based on a deep neural network using T cell receptor CDR3β sequence samples and antigen epitope peptide sequence samples. The present invention does not limit the collective training process of the prediction model, and any reasonable training method can be used.

[0035] Step 130: Filter and obtain screening results according to the prediction results.

[0036] Specifically, in step 130, direct screening is performed based on the predicted binding probability in the prediction results and the sequence file of the epitope peptide segment that can be bound is output as the screening result. Using the amino acid sequence file in the screening result, the structure of the T cell receptor and epitope complex can be constructed in combination with computer structure simulation tools such as molecular model construction and protein docking.

[0037] Based on the T cell epitope high-throughput screening method provided by the present invention, even if the user directly inputs the complete TCR sequence, its CDR3β sequence can be efficiently extracted, thereby effectively predicting the binding activity, reducing the operator's data preprocessing workload. Compared with the prior art, the present invention is more suitable for large-scale data, even uncleaned data containing abnormal entries, to effectively complete the screening work.

[0038] The following further describes step 110. In some embodiments, extracting target input data based on the input T cell receptor sequence and antigen sequence specifically includes: Step 111: filtering illegal characters from the T cell receptor sequence and the antigen sequence to obtain legal T cell receptor sequences and antigen sequences; Step 112: when the legal T cell receptor sequence is a complete β sequence, the T cell receptor variable region sequence is numbered and annotated using a genome library comparison algorithm based on a hidden Markov model, and a T cell receptor CDR3β sequence is extracted according to the numbering and the annotation; Step 113: When the length of the legal antigen sequence is greater than the length threshold, the antigen sequence is segmented into a series of antigen epitope peptide sequences according to the specified length.

[0039] Specifically, after obtaining the input T cell receptor sequence and antigen sequence, before entering the artificial intelligence prediction model, in order to adapt to high-throughput screening tasks, data error checking and preprocessing steps were designed, which reduced the possibility of model errors and ensured stability and universality.

[0040] Furthermore, the validity check includes checking whether the input sequence is legal, and the preprocessing includes the extraction of the T cell receptor CDR3β sequence and the segmentation of the antigen epitope peptide sequence.

[0041] Furthermore, a legality check is performed in step 111. Specifically, it is detected whether there are illegal characters and other problems in the amino acid sequence string, and entries that meet the sequence data specification are filtered out for subsequent screening to obtain legal T cell receptor sequences and antigen sequences.

[0042] In step 112, since the present invention allows the input of the complete amino acid sequence of the T cell receptor, the sequence of the epitope binding core region of the T cell receptor, the β chain CDR3 region, is automatically extracted for the complete β sequence using a genome library alignment algorithm based on a hidden Markov model, so that the sequence information of the core region can be used in subsequent steps for the prediction of the interaction between TCR and epitopes.

[0043] The present invention provides a genome library comparison algorithm based on a hidden Markov model for automatic extraction of TCR-CDR3β. Even if the user directly inputs a complete TCR sequence, its CDR3β sequence can be efficiently extracted, thereby effectively predicting the binding activity and reducing the operator's data preprocessing workload.

[0044] In step 113, since the present invention allows the input of a longer antigen protein amino acid sequence, the antigen sequence greater than the length threshold is cut into antigen epitope peptide sequences of a specified length. Furthermore, the cut antigen epitope peptide sequences can be used to construct an antigen peptide sequence library (i.e., the aforementioned polypeptide library) for T cell antigen screening.

[0045] It should be noted that the present invention does not limit the length threshold, which can be set according to the usage scenario. The present invention also does not specifically limit the specified length, which can be set according to the situation of the prediction model or the needs of the user.

[0046] The embodiment of the present invention can process the original screening data more automatically, and can cut the input protein sequence according to the user's needs to generate a polypeptide library, and then use the prediction model to screen the antigens that can produce interactions in the generated polypeptide library.

[0047] For example, in one specific embodiment, Figure 2 As shown in the figure, the prediction model only needs to use the sequence of the complementary determining region 3 (CDR3β) of the TCRβ chain and the antigen epitope sequence with a length of 8 to 12 amino acids. To meet various input requirements, the preprocessing includes the following: If the input T cell receptor sequence is a complete β sequence, the ANARCI tool based on the genome library alignment algorithm of the hidden Markov model is used to number and annotate the T cell receptor variable region sequence, and then extract the CDR3β sequence of the T cell receptor as input; if the input antigen sequence is too long, the length (8 to 12 amino acids) can be specified, and the antigen sequence can be segmented into a series of antigen epitope peptide sequences that meet the length requirements in a sliding window segmentation manner, and then the peptide sequence library corresponding to the input antigen is constructed for screening.

[0048] Based on the above two points, coupled with a module for legitimacy detection, the present invention is more suitable for effectively completing the screening work for large-scale data, even uncleaned data containing abnormal entries.

[0049] The following further describes step 120. In some embodiments, the prediction model includes a feature extractor, an interaction extractor, an epitope autoencoder, a binding prediction module, and a residue interaction prediction module; Accordingly, the target input data is input into a pre-built prediction model to obtain a prediction result, which specifically includes: Step 121: extracting a first feature and a second feature based on the feature extractor according to the target input data; Step 122: forming an interaction feature graph by dimension expansion according to the first feature and the second feature; Step 123: obtaining global epitope information based on the interaction extractor according to the epitope feature vector; wherein the epitope feature vector is extracted based on the epitope autoencoder according to a preset epitope database; Step 124: extracting interaction information of paired residues based on the interaction extractor according to the interaction feature graph; Step 125: aggregating pairwise interactions based on the binding prediction module according to the interaction information and the global epitope information to obtain a predicted binding probability; Step 126: obtaining a residue-level interaction score heat map according to the interaction information; Step 127: Obtain a prediction result according to the interaction score heat map and the predicted binding probability.

[0050] Specifically, Figure 3 As shown in the figure, the structure of the model is that the feature extractor learns the sequence features of CDR3 and epitope to obtain the first feature and the second feature, and then forms an interaction feature map through dimension expansion, and the interaction extractor constructed by the two-dimensional convolutional neural network extracts the interaction information of paired residues. At the same time, the interaction extractor obtains the global epitope information from the epitope feature vector generated by the autoencoder, wherein the epitope feature vector is extracted based on the epitope autoencoder according to the preset epitope database. The binding prediction module aggregates all pairwise interactions to predict the binding probability, and the residue interaction prediction module is used to output the residue-level interaction matrix. According to the residue-level interaction matrix, the residue-level interaction score heat map can be drawn.

[0051] Furthermore, the results of the T cell receptor and epitope interaction prediction include the predicted binding probability of the two, as well as the specific information of the two at the residue interaction level.

[0052] Further, step 130 is further described below. In some embodiments, obtaining a screening result according to the prediction result screening specifically includes: The T cell receptor CDR3β sequence and the antigen epitope peptide sequence whose predicted binding probability is greater than a preset binding probability judgment threshold are screened and output as screening results.

[0053] Specifically, the screening is performed using a pre-set binding probability judgment threshold. During the screening process, not only can the binding probability of the T cell receptor and the epitope be predicted efficiently and accurately, thereby screening out the antigen epitope for the specified T cell interaction from the antigen peptide sequence library, but the contact score heat map of the two at the residue interaction level can also be output to analyze the residue-level interaction between the TCR and the epitope.

[0054] Further, based on the above embodiment, the feature extractor is constructed based on an embedding layer and a one-dimensional convolutional neural network; The residue interaction prediction module is constructed based on a two-dimensional convolutional neural network.

[0055] Specifically, the feature extractor is composed of an embedding layer and a one-dimensional convolutional neural network. The residue interaction prediction module is built based on a two-dimensional convolutional neural network. The interaction extractor contains two two-dimensional convolutional neural network modules, each of which contains a two-dimensional convolutional layer, a batch normalization layer, and a ReLU activation function. The epitope autoencoder contains two one-dimensional convolutional neural network modules, each of which contains a one-dimensional convolutional layer, a flattening layer, and a fully connected layer. The binding prediction module contains a maximum pooling layer and a fully connected layer with a sigmoid activation function.

[0056] More specifically, in some embodiments, the residue interaction prediction module in the deep learning prediction model provides the use of a two-dimensional convolutional neural network module comprising two output channels, which can predict the residue distance matrix and the contact matrix, and can provide specific information on the residue interaction level between the T cell receptor and the epitope.

[0057] Further, based on the above embodiments, in some embodiments, the prediction result also includes the target input data corresponding to the predicted binding probability and the input T cell receptor sequence and antigen sequence corresponding to the target input data.

[0058] Specifically, in order to facilitate screening, the prediction results can be further filled in based on the results output by the artificial intelligence prediction model, including the target input data corresponding to the predicted binding probability and the input T cell receptor sequence and antigen sequence corresponding to the target input data.

[0059] For example, in one embodiment, the input T cell receptor sequence and antigen sequence, as well as the pre-processed T cell receptor CDR3β sequence and antigen epitope peptide sequence are intuitively presented in a table form. At the same time, the predicted binding probability of each pair of T cell receptor and antigen epitope is output, as well as the interaction score heat map between the two at the residue level.

[0060] In order to further explain the T cell epitope high-throughput screening method provided by the present invention in detail, the following specific embodiments are provided. In the present embodiment, it is provided in the form of a computing component in the MaXFlow platform. By connecting the "read sequence file" component with the "TCR-antigen epitope interaction prediction" component to build a computing workflow, the present invention is applied very conveniently.

[0061] In this embodiment, the melanoma antigen-A4 (MAGE-A4) peptide-TCR complex is a key oncology target, which can be expressed in many tumors, including esophagus, head and neck, lung, ovary, bladder, breast, lymphoma and melanoma. Taking MAGE-A4 and its specific TCR as an example, the present invention is applied to predict the interaction between the two. In the calculation workflow, the first "read sequence file" component is set to input the sequence of the melanoma antigen-A4 protein with ID P43358 in the UniPort protein database, and the second "read sequence file" component is set to input the CDR3β sequence of the specific TCR in the structure file with ID 6TRO in the RCSB PDB protein structure database. The "cut length" parameter of the "TCR-antigen epitope interaction prediction" component is set to 10, and MAGE-A4 will be cut into a series of peptide segments with a length of 10 during the calculation process.

[0062] After the computational workflow is completed, the report of the "TCR-antigen epitope interaction prediction" component can be viewed. The component report includes that the CDR3β sequence of the input TCR has been successfully extracted, and the input melanoma antigen-A4 protein antigen sequence has been cut into a series of antigen peptide sequences of length 10 during the calculation process. The report also gives the predicted probability of interaction between each antigen peptide and the input TCR predicted by the deep learning model and the judgment of whether the antigen peptide can bind to the TCR, and also provides a heat map of the interaction score between the antigen peptide and the TCR at the residue level. In this embodiment, the peptide with the sequence "GVYDGREHTV" has the highest predicted probability of interaction, and it can be screened to determine that the peptide should be the antigen epitope of MAGE-A4. The MAGE-A4 antigen peptide bound to TCR in the crystal structure of the TCR-antigen epitope complex with PDB ID 6TRO in the RCSB PDB database is also "GVYDGREHTV", indicating that the antigen epitope peptide predicted by the present invention is consistent with experimental evidence.

[0063] Furthermore, in the interaction score heat map, it is shown that the arginine residues (R) and glutamic acid residues (E) of the MAGE-A4 antigen peptide are closer to the methionine residues (M) of the TCR, suggesting that these residues are the sites where the antigen peptide interacts with the TCR. Figure 4 ) is consistent with the prediction results of the present invention, and also provides experimental evidence for the effectiveness of the prediction results of the present invention.

[0064] The above embodiments reflect the function of constructing an antigen peptide sequence library and extracting the CDR3β sequence of TCR during data preprocessing of the present invention. At the same time, the present invention can also obtain calculation results that are consistent with experimental evidence in T cell epitope screening and prediction of epitope and TCR interaction sites.

[0065] The T cell epitope high-throughput screening method provided by the present invention obtains target input data by extracting according to the input T cell receptor sequence and antigen sequence; wherein the target input data includes a T cell receptor CDR3β sequence and at least one antigen epitope peptide sequence; the target input data is input into a pre-constructed prediction model to obtain a prediction result; wherein the prediction result at least includes the predicted binding probability of the T cell receptor CDR3β sequence and the antigen epitope peptide sequence, and the prediction model is obtained by training the T cell receptor CDR3β sequence sample and the antigen epitope peptide sequence sample based on a deep neural network; the screening result is obtained by screening according to the prediction result. The present invention allows the input of T cell receptor sequences and antigen sequences that have not been normalized, and can efficiently extract their CDR3β sequences, thereby effectively performing binding activity prediction work, while improving versatility and stability while reducing the operator's data preprocessing workload, and at the same time, based on the deep neural network, the prediction of the binding probability of the T cell receptor and the epitope is performed, and the screening work is effectively completed.

[0066] The T cell epitope high-throughput screening device provided by the present invention is described below. The T cell epitope high-throughput screening device described below and the T cell epitope high-throughput screening method described above can be referenced to each other. Figure 5 Schematic diagram of the structure of the T cell epitope high-throughput screening device provided by the present invention, such as Figure 5 As shown, the device includes the following modules: An input unit 510 is used to extract target input data according to the input T cell receptor sequence and antigen sequence; wherein the target input data includes a T cell receptor CDR3β sequence and at least one antigen epitope peptide sequence; A prediction unit 520 is used to input the target input data into a pre-built prediction model to obtain a prediction result; wherein the prediction result at least includes the predicted binding probability of the T cell receptor CDR3β sequence and the antigen epitope peptide sequence, and the prediction model is obtained by training the T cell receptor CDR3β sequence samples and the antigen epitope peptide sequence samples based on a deep neural network; The screening unit 530 is used to screen the prediction results to obtain screening results.

[0067] According to a T cell epitope high-throughput screening device provided by the present invention, target input data is extracted based on input T cell receptor sequences and antigen sequences, specifically comprising: Filtering illegal characters on the T cell receptor sequence and the antigen sequence to obtain legal T cell receptor sequence and antigen sequence; When the legal T cell receptor sequence is a complete β sequence, the T cell receptor variable region sequence is numbered and annotated using a genome library comparison algorithm based on a hidden Markov model, and the T cell receptor CDR3β sequence is extracted according to the numbering and the annotation; When the legal antigen sequence length is greater than the length threshold, the antigen sequence is segmented into a series of antigen epitope peptide sequences according to the specified length.

[0068] According to a T cell epitope high-throughput screening device provided by the present invention, the prediction model includes a feature extractor, an interaction extractor, an epitope autoencoder, a binding prediction module and a residue interaction prediction module; Accordingly, the target input data is input into a pre-built prediction model to obtain a prediction result, which specifically includes: Extracting a first feature and a second feature based on the feature extractor according to the target input data; forming an interaction feature graph by dimensional expansion according to the first feature and the second feature; Obtaining global epitope information based on the interaction extractor according to the epitope feature vector; wherein the epitope feature vector is extracted based on the epitope autoencoder according to a preset epitope database; extracting interaction information of pairs of residues based on the interaction extractor according to the interaction feature graph; aggregating pairwise interactions based on the binding prediction module according to the interaction information and the global epitope information to obtain a predicted binding probability; Obtaining a residue-level interaction score heat map according to the interaction information; A prediction result is obtained according to the interaction score heat map and the predicted binding probability.

[0069] According to a T cell epitope high-throughput screening device provided by the present invention, screening results are obtained according to the prediction results, specifically comprising: The T cell receptor CDR3β sequence and the antigen epitope peptide sequence whose predicted binding probability is greater than a preset binding probability judgment threshold are screened and output as screening results.

[0070] According to a T cell epitope high-throughput screening device provided by the present invention, the feature extractor is constructed based on an embedding layer and a one-dimensional convolutional neural network; The residue interaction prediction module is constructed based on a two-dimensional convolutional neural network.

[0071] According to a T cell epitope high-throughput screening device provided by the present invention, the prediction result also includes the target input data corresponding to the predicted binding probability and the input T cell receptor sequence and antigen sequence corresponding to the target input data.

[0072] The T cell epitope high-throughput screening device provided by the present invention obtains target input data by extracting according to the input T cell receptor sequence and antigen sequence; wherein the target input data includes a T cell receptor CDR3β sequence and at least one antigen epitope peptide sequence; the target input data is input into a pre-constructed prediction model to obtain a prediction result; wherein the prediction result at least includes the predicted binding probability of the T cell receptor CDR3β sequence and the antigen epitope peptide sequence, and the prediction model is obtained by training the T cell receptor CDR3β sequence sample and the antigen epitope peptide sequence sample based on a deep neural network; the screening result is obtained by screening according to the prediction result. The present invention allows the input of T cell receptor sequences and antigen sequences that have not been normalized, and can efficiently extract their CDR3β sequences, thereby effectively performing binding activity prediction work, while improving versatility and stability while reducing the operator's data preprocessing workload, and at the same time, based on the deep neural network, the prediction of the binding probability of the T cell receptor and the epitope is performed, effectively completing the screening work.

[0073] Figure 6 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 6As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the T cell epitope high-throughput screening method, the method comprising: extracting target input data according to the input T cell receptor sequence and antigen sequence; wherein the target input data comprises a T cell receptor CDR3β sequence and at least one antigen epitope peptide sequence; inputting the target input data into a pre-constructed prediction model to obtain a prediction result; wherein the prediction result at least comprises the predicted binding probability of the T cell receptor CDR3β sequence and the antigen epitope peptide sequence, the prediction model is obtained by training based on a deep neural network using a T cell receptor CDR3β sequence sample and an antigen epitope peptide sequence sample; and screening according to the prediction result to obtain a screening result.

[0074] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0075] On the other hand, the present invention also provides a computer program product, which includes a computer program, and the computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the T cell epitope high-throughput screening method provided by the above methods, the method comprising: extracting target input data based on the input T cell receptor sequence and antigen sequence; wherein the target input data includes a T cell receptor CDR3β sequence and at least one antigen epitope peptide sequence; inputting the target input data into a pre-constructed prediction model to obtain a prediction result; wherein the prediction result includes at least the predicted binding probability of the T cell receptor CDR3β sequence and the antigen epitope peptide sequence, and the prediction model is obtained by training based on a deep neural network using T cell receptor CDR3β sequence samples and antigen epitope peptide sequence samples; screening according to the prediction result to obtain a screening result.

[0076] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the T cell epitope high-throughput screening method provided by the above-mentioned methods, the method comprising: extracting target input data based on the input T cell receptor sequence and antigen sequence; wherein the target input data comprises a T cell receptor CDR3β sequence and at least one antigen epitope peptide sequence; inputting the target input data into a pre-constructed prediction model to obtain a prediction result; wherein the prediction result comprises at least a predicted binding probability of the T cell receptor CDR3β sequence and the antigen epitope peptide sequence, and the prediction model is obtained by training a deep neural network using T cell receptor CDR3β sequence samples and antigen epitope peptide sequence samples; and screening according to the prediction result to obtain a screening result.

[0077] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0078] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0079] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A high-throughput screening method for T cell epitopes based on a deep learning framework, characterized in that: include: Extracting target input data according to the input T cell receptor sequence and antigen sequence; wherein the target input data includes a T cell receptor CDR3β sequence and at least one antigen epitope peptide sequence; Inputting the target input data into a pre-built prediction model to obtain a prediction result; wherein the prediction result at least includes the predicted binding probability of the T cell receptor CDR3β sequence and the antigen epitope peptide sequence, and the prediction model is obtained by training a deep neural network using T cell receptor CDR3β sequence samples and antigen epitope peptide sequence samples; The screening results are obtained by screening according to the prediction results.

2. The method for high-throughput screening of T cell epitopes according to claim 1, characterized in that: The target input data is extracted based on the input T cell receptor sequence and antigen sequence, including: Filtering illegal characters on the T cell receptor sequence and the antigen sequence to obtain legal T cell receptor sequence and antigen sequence; When the legal T cell receptor sequence is a complete β sequence, the T cell receptor variable region sequence is numbered and annotated using a genome library comparison algorithm based on a hidden Markov model, and the T cell receptor CDR3β sequence is extracted according to the numbering and the annotation; When the legal antigen sequence length is greater than the length threshold, the antigen sequence is segmented into a series of antigen epitope peptide sequences according to the specified length.

3. The method for high-throughput screening of T cell epitopes according to claim 1, characterized in that: The prediction model includes a feature extractor, an interaction extractor, an epitope autoencoder, a binding prediction module and a residue interaction prediction module; Accordingly, the target input data is input into a pre-built prediction model to obtain a prediction result, which specifically includes: Extracting a first feature and a second feature based on the feature extractor according to the target input data; forming an interaction feature graph by dimensional expansion according to the first feature and the second feature; Obtaining global epitope information based on the interaction extractor according to the epitope feature vector; wherein the epitope feature vector is extracted based on the epitope autoencoder according to a preset epitope database; extracting interaction information of pairs of residues based on the interaction extractor according to the interaction feature graph; aggregating pairwise interactions based on the binding prediction module according to the interaction information and the global epitope information to obtain a predicted binding probability; Obtaining a residue-level interaction score heat map according to the interaction information; A prediction result is obtained according to the interaction score heat map and the predicted binding probability.

4. The method for high-throughput screening of T cell epitopes according to claim 1 or 3, characterized in that: The screening results are obtained by screening according to the prediction results, specifically including: The T cell receptor CDR3β sequence and the antigen epitope peptide sequence whose predicted binding probability is greater than a preset binding probability judgment threshold are screened and output as screening results.

5. The method for high-throughput screening of T cell epitopes according to claim 3, characterized in that: The feature extractor is constructed based on an embedding layer and a one-dimensional convolutional neural network; The residue interaction prediction module is constructed based on a two-dimensional convolutional neural network.

6. The method for high-throughput screening of T cell epitopes according to claim 3, characterized in that: The prediction result also includes the target input data corresponding to the predicted binding probability and the input T cell receptor sequence and antigen sequence corresponding to the target input data.

7. A T cell epitope high-throughput screening device based on a deep learning framework, characterized in that: include: An input unit, used to extract target input data according to the input T cell receptor sequence and antigen sequence; wherein the target input data includes a T cell receptor CDR3β sequence and at least one antigen epitope peptide sequence; A prediction unit, used for inputting the target input data into a pre-built prediction model to obtain a prediction result; wherein the prediction result at least includes the predicted binding probability of the T cell receptor CDR3β sequence and the antigen epitope peptide sequence, and the prediction model is obtained by training the T cell receptor CDR3β sequence samples and the antigen epitope peptide sequence samples based on a deep neural network; A screening unit is used to screen and obtain screening results according to the prediction results.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the high-throughput screening method for T cell epitopes according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for high-throughput screening of T cell epitopes according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for high-throughput screening of T cell epitopes according to any one of claims 1 to 6 is implemented.

Images