A method for automatic typing of HLA based on third generation sequencing

Abstract

The application relates to the field of information technology and discloses a method for automatically analyzing and typing HLA based on third-generation sequencing; first, the HLA types of first sequencing read data of different data types of a to-be-tested sample are preliminarily judged, then second sequencing read data is selected, then the identified second sequencing read data of different HLA types is compared with corresponding HLA reference sequences for verification and screening, so that the HLA types existing in the to-be-tested sample are determined; the method can effectively process long read length data generated based on third-generation sequencing technology, reduces the calculation amount of accurate comparison, can realize accurate analysis and report generation of typical third-generation HLA data within 15 minutes, and meets the demand of rapid HLA detection and analysis; meanwhile, the method can also analyze short read length data obtained by second-generation sequencing, and has good data compatibility.

Description

Technical Field

[0001] This invention relates to the field of information technology, and more specifically, to a method for automated HLA genotyping based on third-generation sequencing. Background Technology

[0002] The HLA identification method based on high-throughput sequencing data mainly involves experimental steps such as sample pretreatment, nucleic acid extraction, amplification, and sequencing library construction. A high-throughput sequencer is used to determine the amplified nucleic acid sequences in the sample. After obtaining the nucleic acid sequence data, software for sequence assembly, clustering, and alignment is used to compare and analyze these sequences with the HLA sequence database. Certain alignment result filtering conditions are applied to obtain reliable results. The number of sequences belonging to each HLA subtype in the sample's nucleic acid sequence data and their proportion in the total sequence count are then calculated. Finally, the presence of HLA subtypes in the detected sample is determined.

[0003] Due to the large number of HLA genotyping types and the substantial amount of data, the main technical challenge in rapid detection and analysis of clinical samples lies in achieving rapid assembly, accurate alignment, and haplotype analysis of input data, while ensuring accuracy and minimizing reliance on hardware performance. Therefore, to improve the throughput, resolution, timeliness, and accuracy of clinical HLA genotyping testing, it is essential to establish rapid alignment and analysis methods and systems for HLA genotyping nanopore sequencing data.

[0004] HLA genotyping based on nanopore sequencing data is gradually replacing first- and second-generation sequencing as the mainstream sequencing method for HLA genotyping due to its advantages such as long read lengths, high throughput, and fast genotyping speed. Currently, researchers have developed various genotyping methods based on second-generation sequencing technology. For example, Chinese invention patent CN103221551A discloses an HLA genotype-SNP linkage database, its construction method, and an HLA genotyping method. This method selects HLA locus sequences as reference sequences, identifies SNP sites, obtains the linkage relationships of HLA to the reference sequences to construct a database, and finally performs HLA genotyping based on the SNP linkage relationships of different types. Chinese invention patent application CN109477143A discloses a human leukocyte antigen (HLA) genotyping method that uses multiple sequence alignment (MSA) of known HLA allele reference sequences to determine the HLA allele reference sequences. Finally, it selects the reference sequence closest to the individual to complete HLA genotyping. This method produces more accurate results for HLA types I and II. Chinese invention patent application CN113409890 A discloses an HLA genotyping method based on second-generation sequencing data. This method constructs an HLA reference gene sequence database, generates sequence coordinate mapping vectors, performs quality control and filtering on the raw data, and then uses a clustering algorithm to separate heterozygous and homozygous base positions to obtain haploid sequences, ultimately yielding the genotyping result. International patent WO2022 / 116456 provides a method and device for analyzing second-generation sequencing data of HLA chromosomal region heterozygous deletions, automating the detection of whether HLA chromosomal region heterozygous deletions occur in samples. Currently, commonly used second-generation sequence alignment software includes bwa, bowtie 2, SOAP, and BLAST. Chinese invention patent application CN201810191663 discloses an HLA genotyping method based on a third-generation sequencing platform. Specifically, after the PCR product passes the test, third-generation sequencing is performed to obtain the raw data. Long-sequence alignment is then performed with the reference gene sequence, and sequencing errors are corrected. Haplotype sequences are obtained through phase separation, and then genotyping is performed.

[0005] Although various HLA genotyping methods exist, several problems remain: 1. Most current methods do not support HLA genotyping using nanopore long-read sequencing. 2. Many methods do not support high-resolution HLA genotyping. 3. HLA genes exhibit high polymorphism, high sequence similarity, and numerous repetitive sequences, making alignment difficult and prone to errors, thus compromising the accuracy of genotyping results. 4. Using threshold methods to determine heterozygous bases is inflexible and prone to errors in complex situations. Therefore, how to achieve automated HLA genotyping analysis based on nanopore sequencing data remains an unsolved problem. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies and provide a method for automatic HLA genotyping based on third-generation sequencing, thereby solving the problem of how to quickly and accurately determine the HLA genotype in a sample based on third-generation sequencing data.

[0007] First, two different algorithms are used to make a preliminary judgment on the HLA type of the first sequencing read data of different data types of the sample to be tested. Then, the second sequencing read data is selected. The second sequencing read data of different HLA types are then compared and verified with the corresponding HLA reference sequences to determine the HLA type present in the sample to be tested.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solution: a method for automated HLA genotyping based on third-generation sequencing, the method comprising the following steps:

[0009] Step S1: The HLA gene to be tested is split according to the sample sequence to obtain the raw data.

[0010] Step S2: The raw data obtained in step S1 is subjected to quality assessment.

[0011] Step S3: Preprocess the data from Step S2 that has completed the quality assessment, including removing reads with an average data quality lower than Q15, removing primer sequences designed from the HLA genes on both sides of the reads, and retaining reads with a length between 1000bp and 30000bp.

[0012] Step S4: Use the minimap2 algorithm to align the preprocessed data obtained in step S3 to the HLA-related reference genome, filter out sequences that are not aligned to the HLA-related reference genome, further filter out reads with an average alignment data quality lower than Q30, then output the data, and calculate the coverage depth of HLA genes in the target region based on the obtained data.

[0013] Step S5: Perform sequence filtering on the data output in step S4, retaining sequences with lengths between -1000bp and +1000bp of the HLA amplicon design region length, and downsample the filtered data according to a coverage depth of 2000 to obtain the preliminary filtered data.

[0014] Step S6: Using samtools and Megalodon algorithm, mutation detection is performed on the preliminary filtered data obtained in step S5 based on the HLA-related reference genome. Two mutation site detection results are obtained separately. Then, the two mutation site detection results are merged, and mutation sites with a base depth of less than 30 are filtered out. Then, haplotype splitting is performed based on the mutation detection results to obtain split data.

[0015] Step S7: Perform cluster analysis on the split data obtained in step S6, set the required minimum read length, and cluster reads longer than this length to generate different cluster sequences. The length of each cluster sequence segment is not less than 1000, and the number of sequences in each cluster sequence is not less than 30 reads. The data after cluster analysis is obtained.

[0016] Step S8: Perform genome assembly on the cluster analysis data obtained in step S7;

[0017] Step S9: Perform sequencing error correction on the sequence obtained in step S8 to obtain a corrected consistent sequence;

[0018] Step S10: Extract the CDS region of the corrected sequence and compare it with the reference IMGT database using blastn software to obtain the typing result;

[0019] Step S11: If the total number of sequences in the typing results is less than 300, they are filtered out; otherwise, the typing results are quality checked.

[0020] Step S12: Perform data quality control on the data obtained after genotyping quality inspection in step S11. This involves removing bases, primer dimer sequences, repetitive sequences, and adapter sequences whose average data quality is below Q15. Then, compare the reference sequences from the IPD-IMGT / HLA database with the haplotype consistency sequences obtained in step S10 to obtain the HLA allele analysis results and complete the analysis.

[0021] Preferably, step S1 is as follows: the HLA gene to be tested is added to the sequencing chip, the ONT sequencing software Minknow is started to convert the electrical signal generated by nanopore sequencing into unclassified data, and then the raw data is obtained by automatically splitting the sample barcode sequence.

[0022] Preferably, the quality assessment in step S2 includes average read length, average data quality, median sequence length, median sequence quality, number of sequences, N50 read length, total number of bases, data quality distribution, and sequence length distribution.

[0023] Preferably, the HLA-related reference genomes mentioned in step S3 include HLA-A, HLA-B, HLA-C, HLA-DQA1, HLA-DQB1, HLA-DPA1, HLA-DPB1, HLA-DRB1, HLA-DRB3, HLA-DRB4, and HLA-DRB5.

[0024] Preferably, the haplotype separation based on HLA gene type described in step S5 specifically includes:

[0025] Heterozygous mutations with an average data quality (QUAL) > 15 and a single nucleotide variant depth greater than 25% of the average depth were extracted.

[0026] Extract homozygous mutations with an average data quality (QUAL) > 15 and a SNV frequency greater than 90%.

[0027] If the number of heterozygous mutations exceeds 25% of the total number of mutations, the bases corresponding to the heterozygous mutation positions in the read sequence are extracted to form a new mutant combination sequence; if the number of heterozygous mutations is less than 25%, the bases at the homozygous mutation positions are randomly extracted to form a new mutant combination sequence until its length exceeds 25% of the total number of mutations.

[0028] If there is no heterozygous SNV, skip step S7 and proceed directly to step S10.

[0029] Preferably, step S8 is as follows: Select sequences from the first two clusters of each amplicon region in step S7, with a length between -1000bp and +1000bp from the amplicon, downsample to 500X coverage, and perform genome assembly using Flye and Canu software. Select the minimum length of the amplicon length -1000bp or 10000bp as the minimum assembly overlap region for assembly.

[0030] Preferably, the quality control of the typing results in step S11 includes amplification disequilibrium evaluation, linkage disequilibrium evaluation, homozygous typing heterozygous analysis, serological conversion, rare typing differentiation, new typing identification, and visual comparison.

[0031] This invention addresses numerous problems in HLA genotyping using nanopore sequencing by optimizing algorithms, integrating multi-source data, introducing automated processes, and employing dynamic threshold judgment, achieving the following beneficial effects:

[0032] The method of this invention can effectively process long read data generated by third-generation sequencing technology, reduce the computational load of accurate alignment, and achieve accurate analysis and report generation of typical third-generation HLA data within 15 minutes, meeting the needs of rapid HLA detection and analysis. At the same time, the method can also analyze short read data obtained by second-generation sequencing, and has good data compatibility.

[0033] This invention supports high-resolution HLA genotyping using nanopore long-read sequencing; it not only improves alignment accuracy and reduces erroneous alignments, but also flexibly handles heterozygous base identification to adapt to complex situations; it enhances the timeliness and throughput of clinical applications to meet the needs of large-scale screening; and it provides comprehensive genotyping result reports to assist clinical decision-making. Attached Figure Description

[0034] Figure 1 A flowchart of a method for automatically analyzing HLA typing;

[0035] Figure 2 This is an example of the output for quality assessment in a method for automated analysis of HLA typing.

[0036] Figure 3 Minimap2 uses an index to derive a chain for location;

[0037] Figure 4 Flowchart for the minimap2 algorithm implementation;

[0038] Figure 5 A flowchart illustrating data quality control in a method for automated HLA typing analysis. Detailed Implementation

[0039] The present invention will be further described below with reference to embodiments, but these embodiments are not intended to limit the scope of the invention.

[0040] A method for automated HLA genotyping based on third-generation sequencing, the specific steps of which are as follows:

[0041] Step S1, Data Acquisition and Splitting

[0042] The HLA gene to be tested is loaded into the sequencing chip, the ONT sequencing software Minknow is launched, the dna_r10.4.1_e8.2_400bps_hac@v4.3.0 model is selected, the electrical signals generated by nanopore sequencing are converted into unclassified data fastq, and then the raw data is obtained by automatically splitting the sample barcode sequence.

[0043] Step S2, Quality Assessment

[0044] The raw sequencing data were evaluated using the open-source software nanostat, including: average read length, average data quality, median sequence length, median sequence quality, number of sequences, N50 read length, total number of bases, data quality distribution, and sequence length distribution. If the evaluation results showed that the quality of the data was lower than expected (Q30 > 80%, minimum sequencing depth required 20X, average sequencing depth 1000X, sequence splicing length or average read length 3000-8000bp), especially if the data quality distribution was uneven or the sequence length distribution was abnormal, then library reconstruction or sequencing was required.

[0045] Step S3, Data Preprocessing

[0046] The data that have completed the quality assessment are preprocessed, including removing reads with an average data quality lower than Q15, removing primer sequences designed from the HLA genes on both sides of the reads, and retaining reads with a length between 1000bp and 30000bp.

[0047] Step S4, Data Comparison

[0048] The minimap2 algorithm was used to align the preprocessed data to an HLA-related reference genome. The map-ont parameter was set, using a standard minimizer as the seed. Specifically, a minimizer index table was built based on the reference genome: the minimizer index table of the genome sequence was stored in a hash table (the minimizer is the seed with the lowest hash value within a sequence, used to index the reference genome); for each sequence to be aligned (i.e., nanopore sequencing reads), all minimizers for the sequence were found, and an anchor table was built based on the position of x in minimizer[x-w+1,x]; the anchors were connected according to the optimal chain score to locate them on the reference genome; then, dynamic programming was used to amplify non-seed regions. Figure 3As shown, a minimizer index table is established based on the reference genome. Each read has a sorted anchor table after alignment with the minimizer. A minimizer may find many matching segments in the reads, so anchors that meet the requirements are connected to form chains based on scores, thereby locating the reads to the reference genome. Sequences that do not align to the HLA-related reference genome are filtered out, and reads with an average alignment quality lower than Q30 are further filtered out. Based on the obtained data, the coverage depth of the target HLA gene region is calculated, including: gene name, start point, end point, number of sequences, percentage of sequences passing the screening (screening principle: sequence length reaches 80% of the target region length), number of bases covered, average coverage, average depth, average base quality, and average alignment quality.

[0049] Table 1: Output list of coverage depth for the target area

[0050]

[0051] Step S5, collect the sampling sequence

[0052] The data obtained in step S4 is subjected to sequence filtering, retaining sequences with lengths between -1000bp and +1000bp of the HLA amplicon design region. The filtered data is then downsampled using seqkit according to a coverage depth of 2000 to obtain the preliminarily filtered BAM file.

[0053] Step S6, Mutation Detection

[0054] Using an HLA-related reference genome, mutation detection was performed on the preliminarily filtered BAM files obtained in step S5 using the samtools and Megalodon algorithms. The mutation site detection results from the two algorithms were merged, and mutation sites with a base depth less than 30 were filtered out. The preliminarily filtered BAM files obtained in step S5 were then haplotype-splitting based on the mutation site detection results. The processing method for each HLA gene was as follows:

[0055] #1. Extract heterozygous mutations with an average data quality (QUAL) > 15 and a single nucleotide variant depth greater than 25% of the average depth.

[0056] #2. Extract homozygous mutations with QUAL>15 and SNV frequency greater than 90%.

[0057] #3. If the number of heterozygous mutations exceeds 25% of the total number of mutations, the bases corresponding to the heterozygous mutation positions in the read sequence are extracted to form a new mutant combination sequence. If it is less than 25%, the bases at the homozygous mutation positions are randomly extracted to form a new mutant combination sequence until its length exceeds 25% of the total number of mutations.

[0058] #4. How to generate a BAM file that is directly aligned with the CDS region without performing S7 analysis and without hybrid SNVs.

[0059] Step S7, Cluster Analysis

[0060] Cluster analysis is performed on the mutant combination sequences obtained from the data in #1, #2, and #3 from the previous step. A minimum read length is set, and reads longer than this length are clustered to generate different clusters. Each cluster should have a sequence fragment length of at least 1000 bytes and contain at least 30 reads. This process can also be considered for phasing (genotyping) using Longshot. Longshot can phase SNVs (single nucleotide variants), insertions, deletions, MNPs (multiple adjacent SNVs), and "complex" variants.

[0061] Step S8, Genome Assembly

[0062] In step S7, the sequences from the first two clusters of each amplicon region were selected, and the sequences with lengths ranging from -1000 bp to +1000 bp were downsampled to 500X coverage. Genome assembly was performed using Flye and Canu software, and the smallest assembly overlap region was selected as the minimum length between -1000 bp and 10000 bp of the amplicon length.

[0063] Step S9, Sequencing Error Correction

[0064] Because nanopore sequencing has a high error rate, these errors typically manifest as insertion, deletion, and substitution errors, making the sequences obtained directly from nanopore sequencing potentially inaccurate. Therefore, to improve the accuracy of sequencing data, additional error correction measures must be taken. These measures utilize reference genome alignment, read overlap information, and machine learning models to correct erroneous sequences and improve the accuracy of the final assembly. The Nanopolish software was used to correct sequencing errors in the sequences obtained in the previous step, resulting in corrected, consistent sequences.

[0065] Step S10: Comparison of typing results

[0066] The CDS region of the corrected sequence was extracted and aligned using Blasten software and the reference IMGT database. The HLA type was checked for a four-part format, such as DQB1*05:02:01:01, and the first three parts were extracted and returned. Incomplete alignment results were filtered out, prioritizing the extraction of alignments with the fewest mismatches and gaps, and adding tags. Results with identical three-part domains were merged, and the number of merged results was calculated. The result with the most merged results was extracted and stored in the cluster alignment results for a single locus, thus obtaining the HLA typing result for each gene haplotype.

[0067] The #blastn processing logic is as follows:

[0068] #1. Perform a full comparison and filtering of the initial blastn files.

[0069] #2. Extract data based on the priority of (0,0), (1,0), and (0,1). If no result exists for (0,0), the script will automatically extract results for (1,0) and (0,1).

[0070] #Based on the number of merged results, extract the one with the most results as the optimal solution for the current cluster and present it to the next level. All merged results will be stored in Blast / clsuter_no / sampleid_Cluster_X_blastresult_Filter.xls;

[0071] #If there are no (1, 0) or (0, 1) results, the script will extract and present suboptimal solutions based on the ascending combination of mismatch and gap, and the results will still be stored in bam / barcodeX_Cluster_X_blastresult_Filter.xls

[0072] #3. The optimal solution for each cluster is presented to the amplicon layer, and the data is stored in amplicon / amplicon_blastresults.xls.

[0073] #4. The results of each amplicon will be included in the HLA typing results of the barcode, arranged in the order of cluster1 / 2 / 3 / 4 / 5, row by row.

[0074] # Combined with the clustering statistics results from cluster_stat.txt, the data is stored in Blast / sampleid_HLA_Result.xls

[0075] #5. Perform the final data merging step on sampleid_HLA_Result.xls, merging results with identical morphology and tags, and adding the proportions of the corresponding clusters to obtain the final result.

[0076] #Data is stored in Blast / sampleid_HLA_result.txt

[0077] Step S11, Quality Inspection of Typing Results

[0078] If the total number of sequences in the genotyping results is less than 300, the results will be automatically filtered out. Then, quality control of the genotyping results is performed, including: amplification disequilibrium assessment, linkage disequilibrium assessment, homozygous genotyping heterozygous analysis, serological conversion, rare genotyping differentiation, novel genotyping identification, and visual comparison. This is written in R language, and includes merging duplicate results, adding the sequence count and cluster number, and selecting the cluster with the highest sequence count. The filtering index is based on the top two clusters, with the second-ranked cluster having at least 1 / 5 the sequence count of the first-ranked cluster. Limitations are made according to HLA biological theory, such as: DRB345 absent -1 -2, DRB345 can exist in a maximum of two forms simultaneously, and those with only one cluster are directly classified as diploid. CWD frequency is used to determine rare genotypes. This comprehensive assessment of the accuracy and reliability of the genotyping results ensures their applicability for clinical diagnosis and research.

[0079] Step S12, Data Quality Control

[0080] By removing bases with an average numerical quality below Q15, primer dimer sequences, repetitive sequences, adapter sequences (tag sequences, primer sequences, sequencing primer binding sequences), and reference sequences from the current version of the IPD-IMGT / HLA database, the sequencing sequences are aligned with the reference database. After alignment, data processing is performed to obtain HLA allele analysis results, completing the secondary data analysis. Each sample sequencing outputs 250M of data, and the process includes primary and secondary data analysis.

[0081] HLA genotyping data quality control requirements include: Q30 > 80%, minimum sequencing depth of 20X, average sequencing depth of 1000X, and sequence splice length or average read length of 3000-8000 bp. Sequencing homogeneity requires reads to be evenly distributed across the detection region; sequencing coverage requires 100% coverage of the target region for each gene locus, especially complete coverage of key exon regions at the required sequencing depth. Allelic balance refers to the ratio of two bases at polymorphic allele positions, distinguishing between homozygous and heterozygous sites; when the ratio of two bases is close, it is considered allelic balance; when it is below an acceptable threshold (set to 1 / 3), it is considered allelic imbalance. Currently, 50% is used as the cutoff to determine mismatches (further experiments are needed to define parameters and grayscale regions).

[0082] HLA gene testing bioinformatics quality control standards include sample quality control and intra-assay positive, negative, and blank control sample (NTC) quality control. For detailed procedures, please refer to [link to documentation]. Figure 5 .

[0083] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. However, the above description is merely a specific embodiment of the present invention, and the technical features of the present invention are not limited thereto. Any other embodiments derived by those skilled in the art without departing from the technical solution of the present invention should be covered within the patent scope of the present invention.

Claims

1. A method for automated HLA genotyping based on third-generation sequencing, characterized in that, The method includes the following steps: Step S1: The HLA gene to be tested is split according to the sample sequence to obtain the raw data. Step S2: The raw data obtained in step S1 is subjected to quality assessment. Step S3: Preprocess the data from Step S2 that has completed the quality assessment, including removing reads with an average data quality lower than Q15, removing primer sequences designed from the HLA genes on both sides of the reads, and retaining reads with a length between 1000bp and 30000bp. Step S4: Use the minimap2 algorithm to align the preprocessed data obtained in step S3 to the HLA-related reference genome, filter out sequences that are not aligned to the HLA-related reference genome, further filter out reads with an average alignment data quality lower than Q30, then output the data, and calculate the coverage depth of HLA genes in the target region based on the obtained data. Step S5: Perform sequence filtering on the data output in step S4, retaining sequences with a length between -1000bp and +1000bp in the HLA amplicon design region, and downsample the filtered data according to a coverage depth of 2000 to obtain the preliminary filtered data. Step S6: Using samtools and Megalodon algorithm, mutation detection is performed on the preliminary filtered data obtained in step S5 based on the HLA-related reference genome. Two mutation site detection results are obtained separately. Then, the two mutation site detection results are merged, and mutation sites with a base depth of less than 30 are filtered out. Then, haplotype splitting is performed based on the mutation detection results to obtain split data. Step S7: Perform cluster analysis on the split data obtained in step S6, set the required minimum read length, and cluster reads longer than this length to generate different cluster sequences. The length of each cluster sequence segment is not less than 1000, and the number of sequences in each cluster sequence is not less than 30 reads. The data after cluster analysis is obtained. Step S8: Perform genome assembly on the cluster analysis data obtained in step S7; Step S9: Perform sequencing error correction on the sequence obtained in step S8 to obtain a corrected consistent sequence; Step S10: Extract the CDS region of the corrected sequence and compare it with the reference IMGT database using blastn software to obtain the typing result; Step S11: If the total number of sequences in the typing results is less than 300, they are filtered out; otherwise, the typing results are quality checked. Step S12: Perform data quality control on the data obtained after genotyping quality inspection in step S11. Remove bases, primer dimer sequences, repetitive sequences, and adapter sequences with an average data quality lower than Q15. Then, use the reference sequence from the IPD-IMGT / HLA database to compare with the haplotype consistency sequence obtained in step S10 to obtain the HLA allele analysis results and complete the analysis.

2. The method for automated HLA genotyping based on third-generation sequencing according to claim 1, characterized in that, Step S1 is as follows: Add the HLA gene to be tested to the sequencing chip, start the ONT sequencing software Minknow to convert the electrical signal generated by nanopore sequencing into unclassified data, and then automatically split it according to the sample barcode sequence to obtain the raw data.

3. The method for automated HLA genotyping based on third-generation sequencing according to claim 1, characterized in that, The quality assessment described in step S2 includes average read length, average data quality, median sequence length, median sequence quality, number of sequences, N50 read length, total number of bases, data quality distribution, and sequence length distribution.

4. The method for automated HLA genotyping based on third-generation sequencing according to claim 1, characterized in that, The HLA-related reference genomes mentioned in step S4 include HLA-A, HLA-B, HLA-C, HLA-DQA1, HLA-DQB1, HLA-DPA1, HLA-DPB1, HLA-DRB1, HLA-DRB3, HLA-DRB4, and HLA-DRB5.

5. The method for automated HLA genotyping based on third-generation sequencing according to claim 1, characterized in that, Step S6 involves haplotype separation based on HLA gene type, specifically as follows: Heterozygous mutations with an average data quality (QUAL) > 15 and a single nucleotide variant depth greater than 25% of the average depth were extracted. Extract homozygous mutations with a mean data quality (QUAL) > 15 and a SNV frequency greater than 90%. If the number of heterozygous mutations exceeds 25% of the total number of mutations, the bases corresponding to the heterozygous mutation positions in the read sequence are extracted to form a new mutant combination sequence; if the number of heterozygous mutations is less than 25%, the bases at the homozygous mutation positions are randomly extracted to form a new mutant combination sequence until its length exceeds 25% of the total number of mutations. If there is no heterozygous SNV, skip step S7 and proceed directly to step S10.

6. The method for automated HLA genotyping based on third-generation sequencing according to claim 1, characterized in that, The specific steps of step S8 are as follows: In each amplicon region, sequences with lengths ranging from -1000 bp to +1000 bp were selected from the first two clusters. These sequences were downsampled to 500X coverage, and genome assembly was performed using Flye and Canu software. The smallest assembly overlap region was selected as the minimum length between the amplicon length and 10000 bp.

7. The method for automated HLA genotyping based on third-generation sequencing according to claim 1, characterized in that, The quality control of the typing results in step S11 includes amplification disequilibrium evaluation, linkage disequilibrium evaluation, homozygous analysis and heterozygous analysis, serological conversion, rare typing differentiation, new typing identification and visual comparison.

Images