Tumor neoantigen mutation load detection method and device, equipment and storage medium
By performing multi-step detection and screening of tumor samples and normal samples, combined with HLA genotype correction, the problem of low detection accuracy in the existing technology is solved, the detection accuracy of neoantigen mutation load is improved, and the tumor immunotherapy is provided More accurate biomarkers.
Patent Information
- Application Number
- CN202411033236.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-30
- Publication Date
- 2024-11-15
AI Technical Summary
The prior art has low detection accuracy in evaluating tumor neoantigens mutation burden in tumor immunotherapy, especially inaccurate HLA genotyping and rough neoantigens screening methods, resulting in a high false positive rate and affecting the treatment effect.
By performing original variant detection, annotation, screening, HLA genotyping and neoantigen prediction on tumor samples and normal samples, screening with candidate peptides, neoantigen mutation load was calculated, candidate mutations were screened using multi-dimensional indicators and HLA genes were performed. type correction to improve detection accuracy.
It improves the detection accuracy of tumor neoantigens mutation burden, reduces the false positive rate, and enhances the accuracy of HLA genotyping results, thereby providing more accurate biomarkers for tumor immunotherapy.
Smart Images

Figure CN118969082A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of bioinformatics, and specifically relates to a method and device, equipment, and storage medium for detecting tumor neoantigen mutation load. Background Art
[0002] In recent years, tumor immunotherapy has become a breakthrough in the field of tumor treatment. In this field, immune checkpoint inhibitors (ICIs) are one of the most widely used methods. However, individual differences lead to differences in treatment effects. Therefore, before treatment, it is crucial to evaluate the molecular indicators related to the efficacy of ICIs in patients in order to provide patients with more effective treatment plans and prognosis. At present, the commonly used molecular indicators related to the efficacy of ICIs include tumor mutational burden (TMB) and microsatellite instability (MSI).
[0003] However, in practice, it has been found that some patients with low TMB levels may also benefit from ICIs treatment, which indicates that existing indicators cannot accurately screen out potential beneficiaries, so more precise indicators are needed. Tumor Neoantigens Burden (TNB) can directly reflect the translation of mutations, the patient's own specific human leukocyte antigen subtype (Human Leukocyte Antigen, HLA) and the true state of the immune system. Many studies have shown that TNB is associated with the efficacy and prognosis of tumor immunotherapy and is expected to become a key biomarker. Therefore, a method to accurately evaluate TNB is urgently needed.
[0004] With the development of next generation sequencing (NGS), researchers can now use bioinformatics methods to quickly analyze sequencing data to predict TNB. TNB prediction involves a number of steps, including somatic variant detection, somatic variant screening, variant annotation, HLA genotyping, affinity prediction, new antigen screening, TNB calculation, and more.
[0005] However, in clinical practice, existing methods still face many challenges. For example, inaccurate HLA genotyping limits the comprehensive screening of neoantigens, and the neoantigen screening methods are rough and have a high false positive rate, which leads to low detection accuracy of tumor neoantigen mutation load. Summary of the invention
[0006] The purpose of the present invention is to provide a method and device, equipment, and storage medium for detecting tumor neoantigen mutation load, which can improve the detection accuracy of tumor neoantigen mutation load.
[0007] The first aspect of the present invention discloses a method for detecting tumor neoantigen mutation load, comprising:
[0008] The tumor samples and their paired normal samples were respectively subjected to original mutation detection to obtain somatic mutations;
[0009] Annotate somatic variants to obtain annotated somatic variants;
[0010] Screen the annotated somatic variants to obtain candidate variants;
[0011] Translate the candidate variant into an amino acid sequence to obtain multiple variant peptides, and screen out variant peptides that meet the preset peptide length as candidate peptides;
[0012] Perform HLA genotyping on the DNA data of the tumor sample and its paired normal sample respectively, and calibrate the HLA genotyping results of the two to obtain the target HLA genotype;
[0013] Predicting new antigens based on the target HLA genotype and the candidate peptides to obtain multiple original new antigens;
[0014] Screening a plurality of the original neoantigens to obtain candidate neoantigens;
[0015] Based on the candidate neoantigens, the neoantigen mutation load is calculated.
[0016] In some embodiments, the annotated somatic variants are screened to obtain candidate variants, including:
[0017] Remove somatic variants with "strand_bias", "normal_artifact", "slippage", or "weak_evidence" labels in the FILTER column label, and only keep somatic variants with the "clustered_events" label;
[0018] From the somatic variants with the "clustered_events" label, remove somatic variants with a variant allele frequency less than 5% in tumor samples and a variant allele frequency greater than 5% in their paired normal samples; and filter out somatic variants with an allele frequency greater than or equal to 3% in the gnomAD database population, and then retain somatic variants with a coverage greater than 50 layers in tumor samples and greater than 20 layers in normal samples. Finally, only somatic variants with a variant type of non-synonymous mutation are retained as the final candidate variants obtained by screening.
[0019] In some embodiments, performing HLA genotyping on DNA data of a tumor sample and its paired normal sample respectively includes:
[0020] Construct the necessary database of HLA genotypes, labeled as HLA dictionary;
[0021] According to the HLA dictionary, HLA genotyping is performed on the DNA data of the tumor sample and its paired normal sample respectively.
[0022] In some embodiments, a database necessary for HLA genotypes is constructed, labeled as an HLA dictionary, including:
[0023] Download HLA allele sequences from the IPD-IMGT / HLA database and decompose the HLA allele sequences into exon sequences and intron sequences;
[0024] Add 150 "N" nucleotides at both ends of the exon sequence and the intron sequence;
[0025] According to the HLA interval coordinates, the HLA region data of the exon sequence and the intron sequence are extracted, the format of the HLA region data is adjusted according to the format requirements, and the data is saved as an HLA dictionary.
[0026] In some embodiments, according to the HLA dictionary, HLA genotyping is performed on the DNA data of the tumor sample and its paired normal sample respectively, including:
[0027] Compare the read sequences in the BAM files of the tumor sample and its paired normal sample with the HLA region data of all exon sequences and intron sequences in the HLA dictionary respectively;
[0028] After the alignment is completed, the read sequences that satisfy no mismatches in exons or allow up to two mismatches in introns are selected as matching sequences;
[0029] Assigning the matching sequences to corresponding HLA alleles, and weighting each exon sequence according to the number of matching sequences to obtain allele pairs corresponding to the tumor sample and its paired normal sample;
[0030] The scores of the allele pairs are calculated to obtain the HLA genotyping results of the tumor sample and its paired normal sample.
[0031] In some embodiments, screening a plurality of said original new antigens to obtain candidate new antigens comprises:
[0032] The affinity IC50 value and %Rank value of each of the original neoantigens were calculated, and the original neoantigens with IC50 ≤ 500 nM and %Rank ≤ 2 were screened as candidate neoantigens.
[0033] The second aspect of the present invention discloses a device for detecting tumor neoantigen mutation load, comprising:
[0034] A variation detection unit is used to perform original variation detection on tumor samples and their paired normal samples to obtain somatic variation;
[0035] An annotation unit, used to annotate somatic cell variations and obtain annotated somatic cell variations;
[0036] A first screening unit is used to screen the annotated somatic variants to obtain candidate variants;
[0037] The second screening unit is used to translate the candidate variant into an amino acid sequence to obtain a plurality of variant peptides, and screen out variant peptides that meet a preset peptide length as candidate peptides;
[0038] The HLA typing unit is used to perform HLA genotyping on the DNA data of the tumor sample and its paired normal sample respectively, and to calibrate the HLA genotyping results of the two to obtain the target HLA genotype;
[0039] A prediction unit, used to predict new antigens according to the target HLA genotype and the candidate peptide segment to obtain multiple original new antigens;
[0040] A third screening unit is used to screen the multiple original neoantigens to obtain candidate neoantigens;
[0041] A calculation unit is used to calculate the neoantigen mutation load based on the candidate neoantigen.
[0042] In some embodiments, the first screening unit is specifically used to remove somatic variants containing "strand_bias", "normal_artifact", "slippage" or "weak_evidence" labels in the FILTER column labels, and only retain somatic variants with "clustered_events" labels; and
[0043] From the somatic variants with the "clustered_events" label, remove somatic variants with a variant allele frequency less than 5% in tumor samples and a variant allele frequency greater than 5% in their paired normal samples; and filter out somatic variants with an allele frequency greater than or equal to 3% in the gnomAD database population, and then retain somatic variants with a coverage greater than 50 layers in tumor samples and greater than 20 layers in normal samples. Finally, only somatic variants with a variant type of non-synonymous mutation are retained as the final candidate variants obtained by screening.
[0044] The third aspect of the present invention discloses an electronic device, comprising a memory storing executable program code and a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the method for detecting tumor neoantigen mutation load disclosed in the first aspect.
[0045] The fourth aspect of the present invention discloses a computer-readable storage medium, which stores a computer program, wherein the computer program enables a computer to execute the method for detecting the tumor neoantigen mutation load disclosed in the first aspect.
[0046] The beneficial effects of the present invention are that somatic mutations are obtained by performing original mutation detection on tumor samples and normal samples respectively, candidate mutations are obtained by screening after annotation, and multiple variant peptides are obtained by translation into amino acid sequences, and candidate peptides are screened out; HLA genotyping is performed on tumor samples and normal samples respectively, and the typing results of the two are mutually corrected to obtain the target HLA genotype; candidate neoantigens are predicted in combination with candidate peptides, and the neoantigen mutation load is calculated, so that by performing HLA genotyping on tumor samples and normal samples at the same time, the influence of HLA loss of heterozygosity in tumor samples can be excluded, the accuracy of HLA genotyping results can be improved, and the accuracy of subsequent detection of tumor neoantigen mutation load can be improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] The accompanying drawings herein show specific examples of the technical solutions described in the present invention, and together with the specific implementation methods, constitute a part of the specification, and are used to explain the technical solutions, principles and effects of the present invention.
[0048] Unless otherwise specified or defined, the same reference numerals in different drawings represent the same or similar technical features, and the same or similar technical features may also be represented by different reference numerals.
[0049] Figure 1 It is a flow chart of a method for detecting tumor neoantigen mutation load disclosed in an embodiment of the present invention;
[0050] Figure 2It is a schematic diagram of a flow chart of a method for detecting tumor neoantigen mutation load disclosed in an embodiment of the present invention;
[0051] Figure 3 It is the IC50 and %Rank distribution diagram disclosed in the embodiment of the present invention;
[0052] Figure 4 is an example diagram of the correlation between TNB and TMB in 27 WES samples disclosed in an embodiment of the present invention;
[0053] Figure 5 is an example diagram of the correlation between TNB and TMB in 80 panel samples disclosed in an embodiment of the present invention;
[0054] Figure 6 is an example diagram of the correlation between TNB and TMB in 27 WES samples and simulated panel samples disclosed in an embodiment of the present invention;
[0055] Figure 7 This is an example diagram of key information of new antigens disclosed in an embodiment of the present invention;
[0056] Figure 8 The embodiment of the present invention is disclosed Figure 7 Supplementary Figure of
[0057] Fig. 9 is an example diagram of TNB results disclosed in an embodiment of the present invention;
[0058] Fig.10 It is a schematic diagram of the structure of a device for detecting tumor neoantigen mutation load disclosed in an embodiment of the present invention;
[0059] Fig.11 It is a structural schematic diagram of an electronic device disclosed in an embodiment of the present invention.
[0060] Description of reference numerals:
[0061] 101. mutation detection unit; 102. annotation unit; 103. first screening unit; 104. second screening unit; 105. HLA typing unit; 106. prediction unit; 107. third screening unit; 108. calculation unit; 111. memory; 112. processor. DETAILED DESCRIPTION
[0062] Unless otherwise specified or defined, all technical and scientific terms used herein have the same meaning as those generally understood by those skilled in the art. In the case of combining the technical solution of the present invention with realistic scenarios, all technical and scientific terms used herein may also have meanings corresponding to the purpose of implementing the technical solution of the present invention. The "first, second..." used herein is only used to distinguish the names and does not represent a specific quantity or order. The term "and / or" used herein includes any and all combinations of one or more related listed items.
[0063] It should be noted that when a component is considered to be "fixed to" another component, it can be directly fixed to the other component or there can be a central component; when an component is considered to be "connected to" another component, it can be directly connected to the other component or there can be a central component at the same time; when an component is considered to be "installed on" another component, it can be directly installed on the other component or there can be a central component at the same time. When an component is considered to be "set on" another component, it can be directly set on the other component or there can be a central component at the same time.
[0064] Unless otherwise specified or defined, the "said" and "the" used in this document refer to the technical features or technical contents mentioned or described before the corresponding position, and the technical features or technical contents may be the same as or similar to the technical features or technical contents mentioned therein. In addition, the terms "including" and "having" and any variations thereof used in this document are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.
[0065] The embodiment of the present invention discloses a method for detecting tumor neoantigen mutation load, which can be implemented by computer programming. The execution subject of the method can be an electronic device such as a computer, a laptop, a tablet computer, or a detection device for tumor neoantigen mutation load embedded in an electronic device, and the present invention is not limited to this. In order to facilitate the understanding of the present invention, the specific embodiments of the present invention will be described in more detail below with reference to the drawings of the specification.
[0066] like Figure 1 and Figure 2 As shown, the method for detecting tumor neoantigen mutation load includes the following steps 110 to 180:
[0067] 110. Perform original mutation detection on tumor samples and their paired normal samples to obtain somatic mutations.
[0068] Specifically, the BAM files of tumor samples and their paired normal samples are tested for somatic variations (including SNVs and INDELs) using the Genome Analysis Toolkit (GATK), and the commercial software Sentieon is used to improve the detection speed and accuracy. When a region with possible somatic variation is found, GATK reassembles the reads to generate candidate haplotypes of the variation, and uses the Pair-HMM algorithm to calculate the likelihood matrix, which represents the pairwise comparison results of each read and each haplotype. These likelihood values are then marginalized to obtain the allele likelihood for each potential variation site of the given reads. Finally, GATK will use a Bayesian model to evaluate the probability of somatic variation and sequencing errors. Ultimately, the detection results of the tumor samples and their paired normal samples are collectively referred to as detected somatic variations.
[0069] In order to improve the accuracy of variant detection, this step is implemented from three aspects:
[0070] 1) The more scientific hs37d5 reference genome is used. In addition to being consistent with the commonly used clinical gene detection version in China, hs37d5 also adds some carefully selected bait sequences based on hg19. These bait sequences can "induce" some reads during the variation detection process, preventing these reads from being mistakenly aligned to similar regions on other chromosomes, thereby reducing false positives.
[0071] 2) Somatic mutation sites in normal samples are also considered, and the contamination of normal samples in tumor samples and the contamination of tumor samples in normal samples are comprehensively considered to reduce the impact of mutual contamination between tumor samples and their paired normal samples on the accuracy of mutation detection and improve the accuracy of the process.
[0072] 3) At the same time, after large sample testing, the software default parameters have been optimized to improve the sensitivity of low-frequency variant detection, which helps to identify low-frequency variants more accurately.
[0073] Through the above implementation methods, the problems of cross-contamination between paired samples and difficulty in detecting low-frequency variations in existing somatic cell variation detection are solved. By strictly dealing with cross-contamination and low-frequency variations of paired samples in somatic cell variation detection, the accuracy of variation detection can be improved.
[0074] 120. Annotate somatic variants and obtain annotated somatic variants.
[0075] The gene variation annotation tool VEP was used to perform variation annotation for somatic variation, including adding annotation information such as HGNC gene name, sequence ontology category to which the transcript belongs, transcript support level, HGVS format, mutant protein sequence, transcribed protein sequence, and allele frequency in the gnomAD database.
[0076] In an embodiment of the present invention, when annotating transcripts for somatic variants, all transcripts are considered, and the most confident single transcript is selected through the "--pick" option of VEP. Since a single variant may be annotated with multiple genes, the same gene may be annotated with multiple transcripts and multiple effects of the same transcript, by prioritizing the canonical transcripts, protein-coding transcripts, and types that can lead to more serious consequences for each gene, for example, missense variants are more serious than intronic variants, these measures can reduce the time of the test run and simplify the complexity of subsequent new antigen screening.
[0077] 130. Screen the annotated somatic variants to obtain candidate variants.
[0078] Through multiple indicators, the annotated original variants are further screened to obtain high-quality candidate variants.
[0079] First, the variant sites determined as low-quality by the commercial software Sentieon were excluded, that is, the somatic variants with the labels "strand_bias", "normal_artifact", "slippage" or "weak_evidence" in the FILTER column labels were removed, and only the somatic variants with the label "clustered_events" were retained. Secondly, according to the variant allele frequency (VAF) of somatic variants, somatic variants with VAF less than 5% in tumor samples and VAF greater than 5% in their paired normal samples were removed from the somatic variants with the label "clustered_events". Next, somatic variants with gnomAD_AF greater than or equal to 3% were filtered out according to the population allele frequency (gnomAD_AF) of the gnomAD database. Then, somatic variants with coverage greater than 50 layers in tumor samples and greater than 20 layers in normal samples were retained. Finally, only somatic variants with non-synonymous mutations were retained as candidate variants obtained by the final screening. Among them, only the following non-synonymous mutations are retained: missense mutation, frame-shift insertion, frame-shift deletion, in-frame insertion, in-frame deletion, nonsense mutation, non-stop mutation and splice site mutation. In this way, the present invention can strictly control the quality of candidate mutations by covering a somatic variation screening method with multi-dimensional indicators. It solves the problem of single indicators in existing somatic variation screening methods and can further improve the accuracy of variation detection.
[0080] 140. The candidate variant is translated into an amino acid sequence to obtain multiple variant peptides, and the variant peptides that meet the preset peptide length are screened out as candidate peptides.
[0081] It should be noted that the length of the variant peptide segment will cause a single variant to produce multiple variant peptide segments, and different affinities are predicted in combination with HLA, which is directly related to the number of new antigens and affects the TNB calculation. The peptide segment bound to the class I HLA gene is usually between 8 and 15 amino acids, and the most common length of the peptide segment bound to the class II HLA gene is between 13 and 25 amino acids. However, for any specific HLA allele, there is a preferred peptide segment length. In the present invention, through flexible parameter control, the peptide segment length that meets the requirements can be selected to solve the preference of HLA alleles for peptide segment length. It can be set to select the peptide segment preferred by the class I HLA gene (that is, the preset peptide segment length is between 8 and 15 amino acids), or it can be set to select the peptide segment preferred by the class II HLA gene (that is, the preset peptide segment length is between 13 and 25 amino acids). It can also be set to select the class I and class II HLA genes at the same time to meet different TNB downstream application requirements. By providing flexible variation filtering index selection, it supports the setting of HLA categories and peptide segment length categories on demand.
[0082] 150. HLA genotyping is performed on the DNA data of the tumor sample and its paired normal sample respectively, and the HLA genotyping results of the two are mutually corrected to obtain the target HLA genotype.
[0083] That is, when the allele type of some sites in the HLA typing results of normal samples is homozygous, it is necessary to combine the HLA typing results of tumor samples to analyze the impact of HLA loss of heterozygosity. Finally, the target HLA genotype is obtained.
[0084] Specifically, in step 150, HLA genotyping is performed on the DNA data of the tumor sample and its paired normal sample respectively, which may include the following steps 1501 to 1502 (not shown):
[0085] 1501. Construct the necessary database of HLA genotypes, marked as HLA dictionary.
[0086] The HLA dictionary construction step includes the following steps S11 to S13 (not shown):
[0087] S11. Download the HLA allele sequence from the IPD-IMGT / HLA database and decompose the HLA allele sequence into exon sequence and intron sequence.
[0088] S12. Add 150 "N" nucleotides at both ends of the exon sequence and intron sequence; this length is designed to be greater than half of the maximum length of the NGS short read sequence (about 300 bases) to reduce the bias in the exon and intron regions during subsequent alignment.
[0089] S13. Extract HLA region data of exon sequences and intron sequences according to HLA interval coordinates, adjust the format of the HLA region data according to format requirements, and save it as an HLA dictionary.
[0090] To improve the accuracy of HLA typing, HLA allele sequences can be downloaded from the IPD-IMGT / HLA database regularly, and the HLA dictionary can be updated regularly according to the above steps.
[0091] 1502. According to the HLA dictionary, perform HLA genotyping on the DNA data of the tumor sample and its paired normal sample respectively.
[0092] In step 1502, HLA genotyping is performed according to the following steps S21 to S24 (not shown):
[0093] S21. Align the read sequences in the BAM files of the tumor sample and its paired normal sample with the HLA region data of all exon sequences and intron sequences in the HLA dictionary.
[0094] Among them, BAM files are obtained by whole exome sequencing (WES) and panel sequencing. During the alignment process, the Bowtie 2 algorithm is used to align the read sequences at the forward and reverse ends respectively.
[0095] S22. After the alignment is completed, the read sequences that meet one of the following two conditions are selected as matching sequences for subsequent analysis: a) no mismatches in exons, b) a maximum of two mismatches are allowed in introns.
[0096] S23. Assign matching sequences to corresponding HLA alleles, and weight each exon sequence according to the number of matching sequences to obtain the allele pairs corresponding to the tumor sample and its paired normal sample.
[0097] S24. Calculate the scores of the allele pairs and obtain the HLA genotyping results of the tumor sample and its paired normal sample.
[0098] In order to improve the accuracy of the HLA typing module, in the embodiment of the present invention, by using the DNA data of tumor samples and normal samples for HLA genotyping at the same time, the results of the two samples are mutually corrected, which greatly improves the accuracy of the HLA genotyping results. In addition, through parameter control, the class I and class II HLA genotyping can be turned on or off according to the needs, so as to meet the different calculation requirements such as the flexible selection of class I and class II HLA genotypes and the length of variant peptides when calculating downstream TNB, and have high adaptability and efficiency in different application scenarios.
[0099] 160. Neoantigens are predicted based on the target HLA genotype and candidate peptides to obtain multiple original neoantigens.
[0100] Among them, neoantigen prediction refers to neoantigen affinity prediction, which is used to evaluate the binding strength of candidate peptides to patient-specific class I and class II major histocompatibility complex (MHC) molecules. Human MHC genes are commonly referred to as HLA, and neoantigen affinity plays an important role in neoantigen screening. Neoantigen affinity prediction is a complex process involving multiple aspects, including the cleavage site of proteases on mutant proteins, the transport of peptides, the affinity of mutant peptides and MHC binding, and the presentation of binding complexes (pMHC). When making predictions, the effects of these processes need to be considered comprehensively. Using affinity data from in vitro and in vivo experiments, mass spectrometry elution ligand data, and a large amount of latest literature-related data, different types of algorithmic methods can be used to model and predict affinity. For example, some algorithms rely on linear regression algorithms, especially stable matrix methods such as SMM, SMMPMBEC, and Pickpocket. In addition, some algorithms use variants of neural networks, such as NetMHC 4.0, EDGE, and MHCflurry. The existing TNB testing process usually combines the results of multiple algorithms in order to comprehensively consider the impact of multiple factors. However, due to the overlap and redundancy between the results of multiple algorithms, there are huge decision-making challenges in aggregating, merging, filtering and selecting representatives for new antigen results. In addition, the use of multiple algorithms for operation, calculation and storage will also greatly increase costs.
[0101] The present invention screens out the best new antigen affinity prediction algorithm through validation data, and uses the gold standard data set of clinically validated immunogenicity in vivo and in vitro to systematically evaluate 11 algorithms such as pTuneos, pVACtools, MHCflurry, MHCnuggetsI, NetMHC, PickPocket, NetMHCpan4.1, NetMHCcons, SMMPMBEC, SMM and NetMHCpan4.0. Finally, MHCflurry was selected as the best algorithm for affinity prediction by comprehensively considering indicators such as storage cost, computing time, sensitivity and specificity. MHCflurry is a neural network algorithm that incorporates affinity data obtained from in vitro experiments, mass spectrometry elution ligand data and a large amount of latest literature related data. These data can assist the algorithm to more comprehensively cover many factors involved in new antigen affinity prediction. At the same time, the neural network algorithm can capture the nonlinear relationship between the binding affinity of the peptide sequence and the corresponding MHC molecule. The artificial neural network uses more training data and has higher accuracy than the linear regression prediction algorithm.
[0102] 170. Screen multiple original neoantigens to obtain candidate neoantigens.
[0103] In medicine, new antigen screening uses two indicators: affinity IC50 value and %Rank value. Among them, the IC50 value indicates that at a certain MHC concentration, 50% of the MHC is bound by the peptide molecules, and the concentration of the peptide molecules used. The unit of the IC50 value is nanomolar (nM). Generally, the smaller the IC50 value, the stronger the binding strength between the peptide and the HLA molecule. IC50≤50nM is usually regarded as the "strong affinity" threshold, IC50≤500nM is regarded as the "medium affinity" threshold, and IC50≤5000nM is regarded as the "low affinity" threshold.
[0104] In addition to the IC50 value of each peptide, the percentile ranking is generated by comparing the IC50 value of the peptide with a set of random peptides from the SWISSPROT database to obtain the %Rank value. A smaller %Rank value indicates a high affinity. Neoantigens with %Rank < 2 are generally regarded as candidate affinity neoantigens, %Rank ≥ 0.5 and %Rank ≤ 2 are regarded as weak affinity neoantigens, and %Rank < 0.5 are considered strong affinity neoantigens. In the neoantigen prediction process, the use of a "strong affinity" threshold (IC50 ≤ 50nM) or a "medium affinity" threshold (IC50 ≤ 500nM) or %Rank ≤ 2 may result in the missed detection of neoantigens. Even with higher IC50 and %Rank thresholds, the prediction performance is generally not improved.
[0105] The present invention uses clinically validated gold standard datasets of immunogenicity in vivo and in vitro, in interactive visualization charts (such as Figure 3 The IC50 and %Rank distribution graphs shown, Figure 3 Different colors represent different HLA genotypes, the horizontal axis is the IC50 value (Best MT Score), and the vertical axis is the %Rank value). The IC50 value (Best MT Score) and the %Rank value (Best MT Percentile) are dynamically selected. Finally, the threshold rules of IC50 and %Rank, the key parameters for new antigen filtering, are determined, that is, when IC50≤500nM and %Rank≤2, a balance is achieved between the sensitivity and specificity of the new antigen prediction results. Therefore, in step 170, the affinity IC50 value and %Rank value of each original new antigen are calculated, and the original new antigens with IC50≤500nM and %Rank≤2 are screened as candidate new antigens.
[0106] 180. Calculate the neoantigen mutation load based on the candidate neoantigens.
[0107] Considering the complex relationship between somatic variants and the short peptide neoantigen candidates they may represent. A single somatic variant may correspond to multiple candidate neoantigens for the following reasons: a homologous somatic variant can represent two different alleles, which may produce different peptide sequences if these alleles are in phase with nearby heterozygous proximal variants; multiple short peptides may be generated by moving the position of somatic variants in short sequences or using a variety of peptide lengths (e.g., 8-11 amino acids) for pMHC binding affinity prediction; the same peptide may bind to different neoantigens.
[0108] Therefore, optionally, before calculating the TNB value, the candidate neoantigens can be further filtered to obtain the final neoantigen for each variant. Specifically, all candidate neoantigens with the same chromosome, start position, end position, reference sequence, and variant epitopes are compared based on IC50 and %Rank, and the best peptide sequence is automatically selected as the final neoantigen for each variant.
[0109] Finally, the TNB value is calculated based on the number of final neoantigens and the tumor genome interval, and a neoantigen information report is generated, presenting the key information of each final neoantigen, including chromosome number, genome coordinates, reference sequence, variant sequence, HLA typing information, wild-type peptide length, mutant peptide amino acid sequence, IC50 value, %Rank, and wild-type peptide amino acid sequence. Among them, the ratio of the number of neoantigens to the size of the tumor genome interval is the calculation result of TNB.
[0110] In summary, the present invention is implemented to use a clinically validated gold standard data set to select a robust neoantigen prediction algorithm when calculating TNB, and to establish a scientific neoantigen screening system, so as to more accurately and comprehensively predict TNB, thereby improving the accuracy and efficiency of TNB prediction and providing more precise guidance and decision-making support for tumor immunotherapy.
[0111] Example 1
[0112] This example uses a large amount of internal data, open literature data, and purchased positive reference data to discuss in detail aspects such as validation data collection, variant detection, peptide screening, HLA typing, and TNB calculation, further demonstrating the accuracy of the present invention.
[0113] (1) Verification data collection
[0114] Whole exome sequencing BAM data of 27 tumor patients were collected, of which 3 samples had inter-batch duplication, 2 samples had intra-batch duplication, and 1 sample was the same sample diluted to different tumor purities according to tumor purity. The TMB results of all 27 samples were known. In addition, panel sequencing BAM data of 80 tumor patients were collected, and the TMB results of these 80 samples were also known. Four HLA-positive references were purchased, and the typing results of HLA class I genes (HLA-A, HLA-B, HLA-C) and HLA class II genes (HLA-DQB1, HLA-DRB1) of these samples were known. At the same time, whole exome sequencing BAM files of 8 tumor patients were downloaded from the European Nucleotide Archive (ENA). A total of 31 new antigens were detected in these samples, and their immunogenicity has been strictly verified by experiments.
[0115] (2) Mutation Detection
[0116] The commercial software Sentieon was used to detect somatic variants in 115 validation data, including whole exome sequencing data of 27 tumor patients, panel sequencing data of 80 tumor patients, and paired BAM files of whole exome sequencing data of 8 tumor patients from ENA. The hs37d5 reference genome was used in the study. This choice is not only consistent with the commonly used clinical gene detection version in China, but also the bait sequence on hs37d5 can "induce" some reads during the variation detection process, effectively preventing these reads from being misaligned to similar regions on other chromosomes, thereby significantly reducing the false positive rate. In addition, the study retained the somatic variant information in normal samples, effectively reducing the potential impact of mutual contamination between tumor samples and their paired normal samples on the accuracy of variation detection, and further improving the accuracy of variation detection.
[0117] Next, high-quality candidate variants were further screened using multiple metrics. First, sites identified as low-quality by Sentieon were excluded, i.e., variants with labels of "strand_bias", "normal_artifact", "slippage", or "weak_evidence" were removed from the VCF file, while only variants with the label of "clustered_events" were retained. Secondly, variants with a VAF of less than 5% in tumor samples and a VAF of greater than 5% in their paired normal samples were removed based on the variant allele frequency (VAF). Next, variants with a frequency greater than or equal to 3% were filtered out based on the population frequency (gnomAD_AF). Then, variants with a coverage greater than 50 layers in tumor samples and greater than 20 layers in normal samples were retained. Finally, only non-synonymous mutations were retained as the final candidate variant results.
[0118] (3) Peptide screening
[0119] The candidate variants in the 115 validation data were converted into amino acid sequences, and peptides with a length between 8 and 15 amino acids were selected based on the preference of class I HLA genes, while peptides with a length between 13 and 25 amino acids were selected based on the preference of class II HLA genes.
[0120] (4) HLA typing
[0121] The accuracy of the HLA typing module was verified in HLA-positive reference samples. Whole exome sequencing (WES) and panel sequencing were performed on 4 HLA-positive reference samples to obtain BAM data, and HLA class I and class II HLA genotyping was performed using HLA-HD software. In the WES and panel sequencing data, the class I and class II HLA genotyping results of the reference samples were completely consistent with the expected results, with an accuracy of 4 digits, which met clinical needs (see Tables 1 and 2 for details). Next, HLA genotyping was performed on 115 samples using the validated HLA typing algorithm.
[0122] Table 1: HLA typing results of HLA reference samples verified by WES sequencing
[0123]
[0124] Table 2: HLA typing results of HLA reference samples verified by panel sequencing
[0125]
[0126] (5) TNB calculation
[0127] The data of 8 ENA patients were used to predict neoantigens using 11 algorithms, including pTuneos, pVACtools, MHCflurry, MHCnuggetsI, NetMHC, PickPocket, NetMHCpan4.1, NetMHCcons, SMMPMBEC, SMM and NetMHCpan4.0. The accuracy of these 11 algorithms was evaluated based on the information of 31 experimentally verified neoantigens in these 8 samples. The results showed that in the MHCflurry algorithm, the sensitivity and specificity of neoantigen detection were well balanced (see Table 3). Based on this result, the MHCflurry algorithm was determined as the preferred algorithm for neoantigen prediction. At the same time, through interactive visualization charts ( Figure 3), dynamically determined the threshold rules of IC50 and %Rank, the key parameters for neoantigen filtering, i.e., IC50 ≤ 500nM and %Rank ≤ 2. Ultimately, the ratio of the number of neoantigens to the size of the tumor genome interval is the calculation result of TNB.
[0128] Table 3: Evaluation results of new antigen prediction algorithms
[0129]
[0130] Table 3 continued: Evaluation results of new antigen prediction algorithms
[0131]
[0132]
[0133] Note: In Table 3, TP, true positive. FN, false negative. FP, false positive. TN, true negative.
[0134] Then, the accuracy and robustness of the TNB detection algorithm were verified from four dimensions in KingMed’s internal data. First, in the intra-batch and inter-batch repeatability tests, the TNB results in the WES samples between three batches and within two batches were highly consistent (Table 4), verifying the consistency of the algorithm in different batches. Second, under low tumor purity conditions, the same WES sample was diluted in a ratio of 5:3:2:1 to simulate four different tumor purities. Regardless of the lowest or highest tumor purity, the TNB detection results were accurate (Table 5), further demonstrating the accuracy of the algorithm. Third, in 27 WES samples, TNB was highly correlated with TMB results (fit = 0.9911, Figure 4 ), in the 80 panel samples, the correlation is also obvious (fit = 0.5781, Figure 5 ), which is consistent with the conclusions in many papers, indirectly confirming the accuracy of the TNB detection algorithm. Finally, the TNB results of the panel simulated based on the results of 27 WES samples showed high consistency (fitting degree = 0.9302, Figure 6 ), verifying the robustness and applicability of the TNB detection algorithm in WES data and panel data. Based on the above verification results, the TNB detection algorithm showed excellent accuracy and reliability.
[0135] Finally, the customized output of neoantigen information and TNB results shows the key data of the final neoantigen. The specific content includes mutation information, HLA typing information, peptide information, IC50 value and %Rank, etc. See the attached figure for details ( Figure 7 and Figure 8 ).
[0136] Table 4: TNB consistency between replicate samples within and between batches
[0137]
[0138]
[0139] Table 5: TNB consistency after serial dilution of the same sample
[0140] Sample ID Tumor purity (original purity*dilution ratio) TNB dilution-1 50%*0.625 1.6 dilution-2 50%*0.375 1.2 dilution-3 50%*0.25 1.1 dilution-4 50%*0.125 1.3
[0141] like Fig.10 As shown, the embodiment of the present invention discloses a device for detecting tumor neoantigen mutation load, including a mutation detection unit 101, an annotation unit 102, a first screening unit 103, a second screening unit 104, an HLA typing unit 105, a prediction unit 106, a third screening unit 107, and a calculation unit 108, wherein:
[0142] A variation detection unit 101 is used to perform original variation detection on a tumor sample and its paired normal sample to obtain somatic variation;
[0143] An annotation unit 102, used to annotate somatic cell variations to obtain annotated somatic cell variations;
[0144] A first screening unit 103 is used to screen the annotated somatic variants to obtain candidate variants;
[0145] The second screening unit 104 is used to translate the candidate variation into an amino acid sequence to obtain a plurality of variant peptides, and screen out variant peptides that meet a preset peptide length as candidate peptides;
[0146] The HLA typing unit 105 is used to perform HLA genotyping on the DNA data of the tumor sample and its paired normal sample respectively, and to calibrate the HLA genotyping results of the two to obtain the target HLA genotype;
[0147] A prediction unit 106 is used to predict new antigens according to the target HLA genotype and the candidate peptide segments to obtain multiple original new antigens;
[0148] The third screening unit 107 is used to screen multiple original neoantigens to obtain candidate neoantigens;
[0149] The calculation unit 108 is used to calculate the neoantigen mutation load based on the candidate neoantigens.
[0150] Optionally, the first screening unit 103 is specifically used to eliminate somatic variants containing the labels "strand_bias", "normal_artifact", "slippage" or "weak_evidence" in the FILTER column labels, and retain only the somatic variants with the label "clustered_events"; and, from the somatic variants with the label "clustered_events", remove the somatic variants whose variant allele frequency in the tumor sample is less than 5% and whose variant allele frequency in its paired normal sample is greater than 5%; and, filter out the somatic variants whose allele frequency in the gnomAD database population is greater than or equal to 3%, and then retain the somatic variants with a coverage greater than 50 layers in the tumor sample and a coverage greater than 20 layers in the normal sample, and finally retain only the somatic variants whose variant type is non-synonymous mutation as the candidate variants finally screened.
[0151] Optionally, the HLA typing unit 105 includes the following subunits not shown:
[0152] The construction subunit is used to construct the necessary database of HLA genotypes, which is labeled as HLA dictionary;
[0153] The typing subunit is used to perform HLA genotyping on the DNA data of the tumor sample and its paired normal sample respectively according to the HLA dictionary.
[0154] Furthermore, the above-mentioned construction subunit is specifically used to download the HLA allele sequence from the IPD-IMGT / HLA database, decompose the HLA allele sequence into exon sequence and intron sequence; add 150 "N" nucleotides at both ends of the exon sequence and the intron sequence; extract the HLA region data of the exon sequence and the intron sequence according to the HLA interval coordinates, adjust the format of the HLA region data according to the format requirements, and save it as an HLA dictionary.
[0155] Furthermore, the above-mentioned typing subunit is specifically used to compare the read sequences in the BAM files of the tumor sample and its paired normal sample with the HLA region data of all exon sequences and intron sequences in the HLA dictionary respectively; after the comparison is completed, the read sequences that meet the requirements of no mismatch in the exon or a maximum of two mismatches in the intron are selected as matching sequences; and the matching sequences are assigned to the corresponding HLA alleles, and each exon sequence is weighted according to the number of matching sequences to obtain the allele pairs corresponding to the tumor sample and its paired normal sample respectively; the scores of the allele pairs are calculated to obtain the HLA genotyping results of the tumor sample and its paired normal sample.
[0156] Optionally, the third screening unit 107 is specifically used to calculate the affinity IC50 value and %Rank value of each original neoantigen, and screen the original neoantigens with IC50≤500nM and %Rank≤2 as candidate neoantigens.
[0157] like Fig.11 As shown, an embodiment of the present invention discloses an electronic device, including a memory 111 storing executable program codes and a processor 112 coupled to the memory 111;
[0158] The processor 112 calls the executable program code stored in the memory 111 to execute the method for detecting the tumor neoantigen mutation load described in the above embodiments.
[0159] An embodiment of the present invention further discloses a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the method for detecting the tumor neoantigen mutation load described in the above embodiments.
[0160] The purpose of the above embodiments is to exemplarily reproduce and deduce the technical solution of the present invention, and to fully describe the technical solution, purpose and effect of the present invention. Its purpose is to make the public understand the disclosed content of the present invention more thoroughly and comprehensively, and it does not limit the scope of protection of the present invention.
[0161] The above embodiments are not exhaustive enumerations of the present invention, and there may be multiple other implementations not listed. Any replacement and improvement made without violating the concept of the present invention shall fall within the protection scope of the present invention.
Claims
1. A method for detecting tumor neoantigen mutation load, characterized in that: include: The tumor samples and their paired normal samples were respectively subjected to original mutation detection to obtain somatic mutations; Annotate somatic variants to obtain annotated somatic variants; Screen the annotated somatic variants to obtain candidate variants; Translate the candidate variant into an amino acid sequence to obtain multiple variant peptides, and screen out variant peptides that meet the preset peptide length as candidate peptides; Perform HLA genotyping on the DNA data of the tumor sample and its paired normal sample respectively, and calibrate the HLA genotyping results of the two to obtain the target HLA genotype; Predicting new antigens based on the target HLA genotype and the candidate peptides to obtain multiple original new antigens; Screening a plurality of the original neoantigens to obtain candidate neoantigens; Based on the candidate neoantigens, the neoantigen mutation load is calculated.
2. The method for detecting tumor neoantigen mutation load according to claim 1, characterized in that: The annotated somatic variants were screened to obtain candidate variants, including: Remove somatic variants whose FILTER column labels contain "strand_bias", "normal_artifact", "slippage", or "weak_evidence" labels, and only keep somatic variants with the "clustered_events" label; From the somatic variants with the "clustered_events" label, remove somatic variants with variant allele frequencies less than 5% in tumor samples and greater than 5% in their paired normal samples; and filter out somatic variants with allele frequencies greater than or equal to 3% in the gnomAD database population, then retain somatic variants with coverage greater than 50 layers in tumor samples and greater than 20 layers in normal samples, and finally retain only somatic variants with variant types of non-synonymous mutations as the final candidate variants obtained by screening.
3. The method for detecting tumor neoantigen mutation load according to claim 1 or 2, characterized in that: HLA genotyping was performed on the DNA data of tumor samples and their paired normal samples, including: Construct the necessary database of HLA genotypes, labeled as HLA dictionary; According to the HLA dictionary, HLA genotyping is performed on the DNA data of the tumor sample and its paired normal sample respectively.
4. The method for detecting tumor neoantigen mutation load according to claim 3, characterized in that: Construct the necessary database of HLA genotypes, labeled as HLA dictionary, including: Download HLA allele sequences from the IPD-IMGT / HLA database and decompose the HLA allele sequences into exon sequences and intron sequences; Add 150 "N" nucleotides at both ends of the exon sequence and the intron sequence; According to the HLA interval coordinates, the HLA region data of the exon sequence and the intron sequence are extracted, the format of the HLA region data is adjusted according to the format requirements, and the data is saved as an HLA dictionary.
5. The method for detecting tumor neoantigen mutation load according to claim 4, characterized in that: According to the HLA dictionary, HLA genotyping is performed on the DNA data of the tumor sample and its paired normal sample, including: Compare the read sequences in the BAM files of the tumor sample and its paired normal sample with the HLA region data of all exon sequences and intron sequences in the HLA dictionary respectively; After the alignment is completed, the read sequences that satisfy no mismatches in exons or allow up to two mismatches in introns are selected as matching sequences; Assigning the matching sequences to corresponding HLA alleles, and weighting each exon sequence according to the number of matching sequences to obtain allele pairs corresponding to the tumor sample and its paired normal sample; The scores of the allele pairs are calculated to obtain the HLA genotyping results of the tumor sample and its paired normal sample.
6. The method for detecting tumor neoantigen mutation load according to claim 1, characterized in that: Screening a plurality of the original new antigens to obtain candidate new antigens comprises: The affinity IC50 value and %Rank value of each of the original neoantigens were calculated, and the original neoantigens with IC50 ≤ 500 nM and %Rank ≤ 2 were screened as candidate neoantigens.
7. A device for detecting tumor neoantigen mutation load, characterized in that: include: A variation detection unit is used to perform original variation detection on tumor samples and their paired normal samples to obtain somatic variation; An annotation unit, used to annotate somatic cell variations and obtain annotated somatic cell variations; A first screening unit is used to screen the annotated somatic variants to obtain candidate variants; The second screening unit is used to translate the candidate variant into an amino acid sequence to obtain a plurality of variant peptides, and screen out variant peptides that meet a preset peptide length as candidate peptides; The HLA typing unit is used to perform HLA genotyping on the DNA data of the tumor sample and its paired normal sample respectively, and to calibrate the HLA genotyping results of the two to obtain the target HLA genotype; A prediction unit, used to predict new antigens according to the target HLA genotype and the candidate peptide segment to obtain multiple original new antigens; A third screening unit is used to screen the multiple original neoantigens to obtain candidate neoantigens; A calculation unit is used to calculate the neoantigen mutation load based on the candidate neoantigen.
8. The device for detecting tumor neoantigen mutation load according to claim 7, characterized in that: The first screening unit is specifically used to remove somatic variants containing "strand_bias", "normal_artifact", "slippage" or "weak_evidence" labels in the FILTER column label, and only retain somatic variants with "clustered_events" label; and, From the somatic variants with the "clustered_events" label, remove somatic variants with variant allele frequencies less than 5% in tumor samples and greater than 5% in their paired normal samples; and filter out somatic variants with allele frequencies greater than or equal to 3% in the gnomAD database population, then retain somatic variants with coverage greater than 50 layers in tumor samples and greater than 20 layers in normal samples, and finally retain only somatic variants with variant types of non-synonymous mutations as the final candidate variants obtained by screening.
9. An electronic device, characterized in that: It includes a memory storing executable program code and a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the method for detecting the tumor neoantigen mutation load according to any one of claims 1 to 6.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program enables a computer to execute the method for detecting the tumor neoantigen mutation load according to any one of claims 1 to 6.
Citation Information
Cited By
Method for optimizing molecular marker algorithm based on machine learning and immune escape mechanism
CN119479814A
Method for optimizing molecular marker algorithms based on machine learning and immune escape mechanisms
CN119479814B
Tumor neoantigen screening method and device, electronic equipment and storage medium
CN122531507A
