CAR fixed point integration mode quantitative analysis method based on long read length sequencing

Abstract

The invention discloses a CAR fixed-point integration mode quantitative analysis method based on long-read-long sequencing. The method comprises the following steps: establishing a reference sequence containing upstream and downstream of a target site, acquiring a fragment covering an integration region by utilizing long fragment PCR, and performing third-generation sequencing; hA integrity, position consistency and copy structure characteristics of reads are identified through a structure analysis algorithm, sequencing reads are divided into an HDR type and an NHEJ type, and an integration mode proportion is calculated according to the two types; according to the method, structural analysis and typing judgment are performed on long-read-long sequencing data, so that automatic classification and proportion statistics of an integration mode are realized, the accurate repair rate of an editing event can be quantitatively evaluated, and technical support is provided for quality control and safety evaluation of gene therapy products.

Description

Technical Field

[0001] This invention relates to the field of bioinformatics, and in particular to a quantitative analysis method for CAR site-directed integration patterns based on long-read sequencing. Background Technology

[0002] Chimeric antigen receptor (CAR) T-cell therapy integrates exogenous CAR gene sequences into specific sites in the host genome, thereby endowing cells with the ability to specifically recognize and kill tumor cells. Current site-specific integration typically relies on nucleases such as CRISPR / Cas9 to induce DNA double-strand breaks (DSBs), followed by insertion via intracellular repair mechanisms.

[0003] In the DSB repair pathway, homologous targeted repair (HDR) relies on a donor template with homologous arms (HA) to achieve precise insertion, while non-homologous end joining (NHEJ) completes repair by directly connecting the fractured ends. The two methods differ significantly in efficiency and outcome: NHEJ competitively suppresses HDR and is prone to structural alterations such as insertions / deletions. This results in HDR and NHEJ often coexisting during actual CAR-T cell construction, directly impacting the consistency and safety of the product.

[0004] Traditional next-generation sequencing (NGS) typically has read lengths of less than 300 bp, making it difficult to simultaneously cover upstream homologous arms, inserts, and downstream sequences, thus limiting the resolution of integrated structures. Long-read sequencing (such as ONT and PacBio HiFi) can span kb-level regions, significantly improving the resolution of structural variants (SVs) and breakpoints, and has been validated in various tumor and population studies. At the computational level, long-read alignments often employ tools such as Minimap2 to retain primary, secondary, and supplementary alignment information for resolving complex splicing relationships.

[0005] In summary, due to the limitations of short read coverage and resolution, the high homology between HA and the genome, and the insufficient utilization of long reads and extended reference sequences in existing procedures, there is still a lack of a stable scheme that can distinguish and quantify HDR and NHEJ integration patterns in scenarios containing HA. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides an analytical method capable of quantitatively assessing the accuracy of repair rates for edited events.

[0007] To achieve the above objectives, this invention provides a method for quantitative analysis of CAR site-specific integration patterns based on long-read sequencing, comprising: establishing an original reference sequence containing upstream and downstream of the target site within the CAR; obtaining HA sequences covering the integration region using long-fragment PCR and performing third-generation sequencing to obtain long-read data covering the integration region; preprocessing the long-read data to retain long-read data containing CAR signals; screening the preprocessed long-read data to obtain sequencing reads containing HA sequences; parsing the sequencing reads containing HA sequences to determine the HA integrity, positional consistency, and copy structure characteristics of the sequencing reads; determining the integration type of the sequencing reads based on the parsing results, and performing quantitative statistics based on the proportion of each type of sequencing read; wherein the integration types include HDR integration and NHEJ integration.

[0008] Optionally, the long read data is preprocessed to retain long read data containing CAR signals, including: using quality control software to perform overall quality statistics on the long read data and calculate the average Phred quality of each sequencing read; retaining sequencing reads with an average Phred quality greater than or equal to a first preset value and a read length greater than or equal to a second preset value; performing barcode recognition and CAR primer screening on the retained sequencing reads, eliminating non-specific amplified sequences, and matching the given forward / reverse primers on the positive strand and its reverse complement. If either match is found, it is determined to contain a CAR signal and enters the HA detection screening.

[0009] Optionally, the preprocessed long read data can be screened to obtain sequencing reads containing HA sequences, including: using the CAR non-human primer binding region sequence as a screening marker to perform the first round of screening on the quality-controlled sequencing reads and retaining the sequencing reads that can match the CAR primer binding region sequence; and using a sliding window scanning strategy to dynamically analyze the sequencing reads after the first round of screening to obtain the final sequencing reads.

[0010] Optionally, a sliding window scanning strategy is used to dynamically analyze the sequencing reads after the first round of screening to obtain the final sequencing reads. This includes: sequentially dividing each sequencing read into fixed window lengths while sliding forward with a fixed step size to maintain an overlap of a preset length between adjacent windows; then further subdividing each fixed window into continuous k-mer fragments of the preset length, counting the number of matches between the k-mer fragments and the HA reference sequence, and determining that the window is a candidate region with high similarity to the HA sequence if the number of matches is greater than or equal to a third preset value; if there are multiple adjacent candidate regions, they are merged into continuous candidate fragments; sequencing reads containing HA sequence fragments are retained, while fragments not containing HA sequences are discarded to form cleaned sequencing data.

[0011] Optionally, the sequencing reads containing HA sequences are parsed to determine the HA integrity, positional consistency, and copy structure characteristics of the sequencing reads, including: using a long read alignment algorithm to parse the sequencing reads containing HA sequences to obtain their structural composition; and determining the HA integrity, positional consistency, and copy structure characteristics based on the structural composition.

[0012] Optionally, the structural composition of sequencing reads containing HA sequences is obtained by parsing them using a long read alignment algorithm, including: constructing an extended reference genome sequence, which includes the original sequence of the target genome and artificially assembled exogenous fragments; performing dual reference alignment using a long read alignment algorithm, and retaining the primary alignment, secondary alignment, and supplementary alignment information; extracting the CIGAR, MD, and SA tags of each sequencing read based on the dual reference alignment results, and parsing its structural composition.

[0013] Optionally, the structural components include genomic fragments, HA fragments, and CAR fragments.

[0014] Optionally, the integrity, positional consistency, and copy structure characteristics of the HA are determined based on the structural composition, including: determining boundary crossing, judging whether the sequencing read simultaneously hits the genomic fragment and the HA fragment; if a cross-boundary structure exists, it indicates that an integration event has occurred; judging HA integrity, calculating the difference between the alignment length of the HA region and the designed length; if the deviation is less than or equal to a fourth preset value, it is considered complete; judging HA continuity, if the HA region is separated by soft shearing or breakage in CIGAR, it is considered discontinuous; judging HA duplication, if multiple HA_up hits are detected in the SA tag, it is judged as HA duplication or misalignment; judging positional deviation, comparing the positional offset of the HA region and the upstream genome in the calibration coordinates; if the deviation is greater than a fifth preset value, it is classified as misaligned integration.

[0015] Optionally, the HDR-type integration simultaneously satisfies structural continuity, HA integrity, fixed-point consistency, and no abnormal repetition.

[0016] Optionally, the NHEJ-type integration satisfies any of the following conditions: HA is missing or incomplete; the upstream segment coordinate deviation is greater than the fifth preset value or there is an insertion / missing event; or there are two or more HA_up or HA_down segments.

[0017] The advantages of this invention over the prior art are: by using long-read sequencing data covering the target integration region, structural analysis of the reads is performed, characteristic judgment criteria for homologous arm regions are established, and by combining artificially expanded reference genome sequences, different integration modes (HDR or NHEJ) of CAR fragments are identified through alignment information, and quantitative statistics are performed based on the proportion of each type of read, thereby achieving accurate classification and quantitative evaluation of integration events. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the CAR structure, CRISPR-Cas9 gene editing system, and on-target integration results provided in an embodiment of the present invention;

[0019] Figure 2 This is a flowchart of a method for quantitative analysis of CAR site-directed integration patterns based on long-read sequencing, provided by an embodiment of the present invention;

[0020] Figure 3 This is a schematic diagram of two different integrated event reads provided in an embodiment of the present invention;

[0021] Figure 4 This is a schematic diagram illustrating the statistical situation of HDR vs NHEJ integrated mode provided in an embodiment of the present invention;

[0022] Figure 5 This is a schematic diagram showing the distribution of breakpoint positions in the HDR integration mode according to an embodiment of the present invention;

[0023] Figure 6 This is a schematic diagram of the coverage of the target integration site of the HBB gene provided in an embodiment of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0025] For ease of description, the relevant terms appearing in this embodiment are given a general explanation:

[0026] FASTQ: FASTQ is a text file format used to store biological sequence data and their corresponding base quality scores.

[0027] Reads: Short sequence fragments obtained in sequencing technology, used for subsequent data analysis.

[0028] read_id: A unique identifier for a sequencing read.

[0029] Barcode: A sample-specific short DNA sequence tag used to distinguish samples or molecules from different sources in the same sequencing reaction.

[0030] k-mer / 20-mer: Fixed substrings of length k; used for initial screening of HA sequences using a sliding window.

[0031] Smith-Waterman is a dynamic programming local alignment algorithm used to find the best local similar regions between two sequences, and is particularly suitable for aligning partially similar sequences.

[0032] contig (extended reference contig): an artificially assembled reference sequence (left HA + CAR + right HA), which is combined with hg38 as a joint reference.

[0033] Primary alignment: The alignment record with the highest aligner score.

[0034] Secondary alignment: Substitute hits (non-fragmentation) of the same segment, flag 0x100.

[0035] AS (Alignment Score): Alignment score, often used to compare the quality of primary / secondary tests.

[0036] MAPQ (Mapping Quality): Matching quality, reflecting the uniqueness / confidence of the match (the higher the value, the more reliable).

[0037] HDR (Homologous Directed Repair): refers to the mechanism by which cells precisely repair DNA double-strand breaks using template DNA (homologous arms) with the same or similar sequences. In this invention, HDR means that a foreign CAR fragment is accurately inserted into the target site through a homologous sequence.

[0038] NHEJ (Non-Homologous End Joining): This refers to a repair mechanism in which cells directly join broken DNA ends without relying on a template. This method is prone to insertions, deletions, or misalignments. In this invention, NHEJ indicates that CAR fragments are randomly spliced ​​or incompletely inserted in a non-template manner.

[0039] Example 1

[0040] The main objective of this embodiment is to verify the feasibility and accuracy of the quantitative analysis method for CAR fragment site-specific integration patterns based on long-read sequencing proposed in this invention. By performing long-fragment PCR amplification and third-generation sequencing on CRISPR / Cas9-edited cell samples, and utilizing extended reference sequence alignment, structural analysis, and classification algorithms, two types of integration events—homology-directed repair (HDR) and non-homologous end joining (NHEJ)—were systematically identified, and their proportions were quantitatively calculated. The CAR structure, CRISPR-Cas9 gene editing system, and on-target integration results are shown below. Figure 1 As shown.

[0041] Reference Figure 2 This embodiment provides a method for quantitative analysis of CAR site-directed integration patterns based on long-read sequencing, including the following steps:

[0042] S10: Establish the original reference sequence containing the upstream and downstream of the target site within the CAR, use long fragment PCR to obtain the HA sequence covering the integration region, and perform third-generation sequencing to obtain long read data covering the integration region.

[0043] It should be noted that since the length of the HA sequence is usually 300~1500bp, it is necessary to extend it backward when amplifying the HA sequence covering the target integration region.

[0044] Specifically, in this embodiment, long-fragment PCR primers were designed in the non-homologous region inside the CAR to directionally amplify a fragment of about 500 bp covering the target integration site. Then, the amplified products were subjected to third-generation sequencing to obtain long-read data that spanned the integration region. The data obtained are shown in Table 1.

[0045] Sample Name Original data volume Q20 ratio Q30 ratio Shortest reading length Average read length Longest read length GC content (%) Sample 1 1.83G 71.5% 51.7% 59 733.5 483589 50.7

[0046] Table 1: Summary of Data Quality After Data Deployment

[0047] S20: Preprocess the long read data to retain the long read data containing the CAR signal.

[0048] Specifically, using the original FASTQ data from the edited samples obtained from third-generation sequencing as input, the overall quality was first analyzed using Chopper, and read filtering was performed accordingly: the average Phred quality value was calculated for each read based on the entire sequence, and only reads with an average quality ≥ Q10 and a read length ≥ 45 bp were retained; to avoid weakening breakpoints and micro-homologous signals, no end-pruning or internal trimming was performed in this stage. Subsequently, the retained reads were barcoded and CAR primers were screened, and non-specific amplified sequences were eliminated. The reads were matched with the given forward / reverse primers on the positive strand and its reverse complement. Any match was determined to contain a CAR signal and entered the homologous arm (HA) detection screening. A total of 834,763 reads were obtained, and the results are shown in Table 4.

[0049] S30: Filter the preprocessed long read data to obtain sequencing reads containing HA sequences.

[0050] Specifically, after performing basic quality control and preprocessing on the raw sequencing data, the CAR non-human primer binding region sequence is used as a screening marker to perform the first round of screening on the quality-controlled sequencing reads. Reads that match the CAR primer binding region sequence are retained, thereby excluding irrelevant sequences generated by non-specific amplification and initially identifying the read population that may contain the target fragment. Next, to accurately identify whether the reads contain HA sequence fragments, this embodiment uses a sliding window scanning strategy to dynamically analyze the reads after the first round of screening, specifically through sliding window scanning and candidate fragment screening (including CAR sequences and 300HA fragments).

[0051] To robustly screen fragments containing HA (homological sequences) in a highly homologous background, a sliding scan was performed on each CAR-positive read with a window length of 100 bp and a step size of 80 bp (windows with a terminal length of less than 100 bp were excluded from the evaluation), and each window was divided into 5 non-overlapping 20-mers. Non-overlapping 20-mer indices were pre-constructed based on the left and right HA sequences and their respective inverse complementarities were added, while the source side label (L or R) was retained.

[0052] For any window, if its 20-mer has an exact match number with the HA index ≥ 3 (equivalent to ≥ 60 bp of complete homologous coverage), then the window is marked as a hit and the L (left) / R (right) hit count is recorded. On the same read, adjacent consecutive hit windows (with a fixed 20 bp overlap due to the step size) are merged into a candidate segment (if at least one miss window appears in the middle, they are not merged). The start and end coordinates, span length, number of hit windows and total k-mer support of the candidate segments are summarized, and the HA side preference judgment is given based on the L / R hit advantage.

[0053] The aforementioned candidate fragments serve as input for subsequent local re-alignment and precise breakpoint localization, thereby achieving high-accuracy initial screening and evidence aggregation for CAR integration sites under homologous interference conditions. Examples of the results are shown in Table 2.

[0054] Table 2: Examples of reads filtered by kmer

[0055] Read the segment number Segment start position End position of the segment Homogeneous arm direction Number of windows k-mer support number pass Reading segment 1 1200 1460 L 3 11 yes Reading segment number 2 2105 2345 R 2 7 yes Reading segment number 3 540 720 L 2 6 yes

[0056] S40: Parse the sequencing read containing the HA sequence to determine the HA integrity, positional consistency, and copy structure characteristics of the sequencing read.

[0057] Specifically, the first step is to construct an extended reference genome sequence, which includes the original sequence of the target genome and artificially assembled exogenous fragments of "HA + CAR + HA".

[0058] Then, a long read length alignment algorithm is used to perform dual-reference alignment, and the primary alignment, secondary alignment, and supplementary alignment information are retained.

[0059] Finally, the CIGAR, MD, and SA tags of each read were extracted, and their structural composition (genomic fragment, HA fragment, CAR fragment) was analyzed.

[0060] In this embodiment, after candidate fragments are obtained through sliding window screening, a Smith-Waterman local alignment is performed on each candidate fragment and its corresponding HA sequence, using a fixed matching score, mismatch penalty, and insertion / deletion penalty. Considering that the HA sequence in some samples is not fully inserted, to avoid missed detections due to deletions, this step uses coverage and consistency as admission criteria rather than "full-length coverage" as a hard condition.

[0061] Alignment length ≥ 60 bp: As long as the alignment region reaches 60 bp, the fragment is determined to contain the HA sequence;

[0062] Consistency ≥ 85%: Ensures that at least 85% of the bases in the aligned portion are completely matched, reducing noise interference.

[0063] Reads shorter than 60 bp or with consistency below the threshold are discarded and not included in subsequent analysis.

[0064] This threshold optimization can take into account some missing insertion events, improve the sensitivity of the algorithm, and maintain high specificity. Examples of the results are shown in Table 3, and the reads selected through each step are shown in Table 4.

[0065] Table 3: Examples of reads compared via SW

[0066] Read the segment number Segment start position End position of the segment Comparison direction Comparison score Coverage Comparison length Similarity Whether to pass the filter Reading segment 1 1200 1460 fwd 312 168 190 0.902 yes Reading segment number 2 2105 2345 rev 245 132 155 0.8735 yes Reading segment number 3 540 720 fwd 118 84 110 0.7882 yes

[0067] Table 4: Reads filtered by barcode, CAR, and HA

[0068] Sample Name Valid data after quality control Number of segments filtered by barcode Read count filtered by CAR sequence Number of reads filtered by homologous arms (HA) Sample 1 1184663 834763 681943 550097

[0069] Structural analysis: Joint analysis of the alignment coordinates of the same reads in the original and extended references determines the following characteristics:

[0070] Boundary crossing: Determine whether reads simultaneously hit genomic fragments and HA fragments. If a cross-boundary structure exists, it indicates that an integration event has occurred.

[0071] HA integrity: Calculate the difference between the comparison length and the design length of the HA region. If the deviation is ≤50 bp, it is considered complete.

[0072] HA continuity: If the HA region is separated by soft shear (S) or fracture (N) in CIGAR, it is considered discontinuous.

[0073] HA Repeatability: If multiple HA_up hits are detected in the SA tag (in the same direction or in opposite directions), it is judged as HA repeatability or misalignment.

[0074] Positional bias: Compare the positional shift of the HA region and upstream genome in hg38 coordinates. If the bias is >100bp, it is classified as misaligned integration.

[0075] S50: Determine the integration type of the sequencing reads based on the analysis results, and perform quantitative statistics based on the proportion of each type of sequencing read.

[0076] Specifically, the sequencing read is determined to be either HDR-type integration or NHEJ-type integration based on the analysis results. A schematic diagram of the two different integration events is shown below. Figure 3 As shown.

[0077] The basis for judgment is as follows:

[0078] HDR integration must simultaneously meet the following requirements:

[0079] Continuous structure: It exhibits a "upstream of the genome → HA_up → CAR" structure;

[0080] HA integrity: The length of the HA region is consistent with the design value (300±50 bp);

[0081] Site-specific consistency: The HA and upstream genomic fragments deviate from the reference genome coordinates by ≤ 100 bp;

[0082] No abnormal duplication: No multiple copies or inversions of HA segments were detected.

[0083] NHEJ type integration satisfies any of the following conditions:

[0084] HA deletion or incompleteness (length < 150 bp or discontinuous alignment);

[0085] Upstream segment coordinate deviation > 100 bp or insertion / deletion event exists;

[0086] There are more than two HA_up or HA_down segments (repeated integration).

[0087] The judgment results of this embodiment are shown in Table 5.

[0088] Table 5 Examples of events for determining read ownership integration

[0089] read_id Sequencing read numbering reads1 reads2 primary_ref The reference sequence where the main alignment is located hg38 hg38 secondary_ref Suboptimal alignment in the extended reference sequence HA_CAR_HA HA_CAR_HA boundary_cross Does it cross the genome-HA boundary (Yes / No)? Yes Yes HA_length_diff The difference (bp) between the measured HA length and the design length. 39 39 HA_continuity Is the HA region contiguous (Complete / Split)? Complete Complete HA_repeat Whether duplicate or reverse HA sequences were detected no yes coord_deviation Upstream genome coordinate bias (bp) 19 13 integration_type Judgment type (HDR / NHEJ / Unclassified) HDR NHEJ

[0090] Finally, based on the quantitative statistical results, output the required classification results, such as the number of HDR type reads (N_HDR); the number of NHEJ type reads (N_NHEJ); HDR ratio = N_HDR / (N_HDR + N_NHEJ) × 100%.

[0091] In this embodiment, a visualization module can also be generated, such as an HDR / NHEJ scale (e.g. Figure 4 As shown); Breakpoint distribution diagrams for various integrated events (such as...) Figure 5 (As shown); Schematic diagram of structural alignment (showing genome-HA-CAR continuity)

[0092] Example 2

[0093] To verify the universality and robustness of the method of this invention, this embodiment applies the integration pattern determination process proposed in Example 1 to various gene site editing samples, including typical immune-related genes such as HBB, PDCD1, and CCR5. These genes play important functions in T cell, antigen presentation, and immune regulation, and are common targeted integration or knockout sites. In this embodiment, the coverage of the target integration site of the HBB gene is as follows: Figure 6 As shown.

[0094] Extended reference sequence construction

[0095] For each target gene locus, the corresponding homologous arm (HA) and exogenous CAR fragment are spliced ​​together to form an extended reference sequence. All extended reference sequences are added to a unified index file and used in conjunction with the standard human reference genome (GRCh38) for subsequent alignment analysis.

[0096] Comparison and classification process

[0097] The same long read length alignment algorithm and parsing strategy as in Example 1 are used to retain primary, secondary and supplementary alignment information.

[0098] For each sample, the system automatically extracts the alignment feature information (CIGAR, MD, SA labels, etc.) of the reads and performs integrated pattern classification based on the following features:

[0099] Does it cross the boundary between the genome and HA?

[0100] HA integrity and continuity;

[0101] The deviation between the location coordinates and the target location;

[0102] Does HA appear as duplicates, missing values, or misalignments?

[0103] The HDR and NHEJ type events were identified based on the decision matrix, and the number and proportion of the two types of events were counted. The results are shown in Table 6.

[0104] Table 6. Integration pattern recognition results in samples edited at different gene loci

[0105] gene locus HDR percentage (%) NHEJ percentage (%) HBB 72.4 27.6 PDCD1 69.8 30.2 CCR5 75.1 24.9

[0106] The results showed that this method could accurately identify integration patterns under different genetic backgrounds.

[0107] As can be seen from the above embodiments, the present invention achieves accurate classification and quantitative evaluation of integration patterns by performing structural analysis and genotyping on long-read sequencing data.

[0108] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

[0109] The above are merely optional embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for quantitative analysis of CAR site-directed integration patterns based on long-read sequencing, characterized in that, include: A raw reference sequence containing the upstream and downstream of the target site was established inside the CAR. The HA sequence covering the integration region was obtained by long fragment PCR and third-generation sequencing was performed to obtain long read data covering the integration region. The long read data is preprocessed to retain the long read data containing the CAR signal; The preprocessed long-read data were filtered to obtain sequencing reads containing HA sequences; Sequencing reads containing HA sequences are parsed to determine the HA integrity, positional consistency, and copy structure characteristics of the sequencing reads; The integration type of the sequencing reads is determined based on the analysis results, and quantitative statistics are performed based on the proportion of each type of sequencing read. The integration types include HDR integration and NHEJ integration.

2. The quantitative analysis method for CAR-based fixed-point integration mode according to claim 1, characterized in that, Preprocessing the long read data to retain the long read data containing the CAR signal includes: The overall quality of the long read data was statistically analyzed using quality control software, and the average Phred quality of each sequencing read was calculated. Sequencing reads with an average Phred quality greater than or equal to the first preset value and a read length greater than or equal to the second preset value are retained. The retained sequencing reads are barcode identified and CAR primers are screened. Non-specific amplified sequences are eliminated and matched against the given forward / reverse primers on the positive strand and its reverse complement. If either match is found, it is determined to contain a CAR signal and enters the HA detection screening.

3. The quantitative analysis method for CAR-based fixed-point integration mode according to claim 2, characterized in that, The preprocessed long-read data was filtered to obtain sequencing reads containing HA sequences, including: Using the CAR non-human primer binding region sequence as a screening marker, the sequencing reads that have passed quality control are screened in the first round, and the sequencing reads that can match the CAR primer binding region sequence are retained. A sliding window scanning strategy was used to dynamically analyze the sequencing reads after the first round of screening to obtain the final sequencing reads.

4. The quantitative analysis method for CAR-based fixed-point integration mode according to claim 3, characterized in that, A sliding window scanning strategy was used to dynamically analyze the sequencing reads after the first round of screening to obtain the final sequencing reads, including: Each sequencing read is sequentially divided into segments with a fixed window length, while sliding forward with a fixed step size to maintain a preset overlap between adjacent windows. Each fixed window is then further subdivided into continuous k-mer segments of the preset length. The number of matches between the k-mer segments and the HA reference sequence is counted. If the number of matches is greater than or equal to a third preset value, the window is determined to be a candidate region with high similarity to the HA sequence. If there are multiple adjacent candidate regions, they are merged into continuous candidate segments. Sequencing reads containing HA sequence fragments are retained, while fragments not containing HA sequences are discarded, resulting in cleaned sequencing data.

5. The quantitative analysis method for CAR-based fixed-point integration mode according to claim 1, characterized in that, The sequencing reads containing HA sequences are parsed to determine the HA integrity, positional consistency, and copy structure characteristics of the sequencing reads, including: The structural composition of sequencing reads containing HA sequences was obtained by parsing them using a long read length alignment algorithm. The integrity location consistency and copy structure characteristics of HA are determined based on the structural composition.

6. The quantitative analysis method for CAR-based fixed-point integration mode according to claim 5, characterized in that, The structural composition of sequencing reads containing HA sequences was obtained by parsing them using a long read alignment algorithm, including: Construct an extended reference genome sequence, which includes the original sequence of the target genome and artificially assembled exogenous fragments; A long read length alignment algorithm is used to perform dual-reference alignment, and the primary alignment, secondary alignment and supplementary alignment information are retained; Based on the results of the dual-reference alignment, the CIGAR, MD, and SA tags of each sequencing read were extracted, and their structural composition was analyzed.

7. The quantitative analysis method for CAR-based fixed-point integration mode according to claim 6, characterized in that, The structure consists of a genome fragment, an HA fragment, and a CAR fragment.

8. The quantitative analysis method for CAR-based fixed-point integration mode according to claim 5, characterized in that, Determining the HA integrity location consistency and copy structure characteristics based on the aforementioned structural composition includes: Determine the boundary crossing to determine whether the sequencing read simultaneously hits the genomic segment and the HA segment. If a cross-boundary structure exists, it indicates that an integration event has occurred. To determine the integrity of the HA, calculate the difference between the comparison length and the design length of the HA region. If the deviation is less than or equal to the fourth preset value, it is considered complete. To determine the continuity of HA, if the HA region is separated by soft shear or fracture in CIGAR, it is considered discontinuous. To determine HA repeatability, if multiple HA_up hits are detected in the SA tag, it is determined that the HA is repeated or misaligned. To determine the positional deviation, compare the positional shift of the HA region and the upstream genome under the calibrated coordinates. If the deviation is greater than the fifth preset value, it is classified as misaligned integration.

9. The quantitative analysis method for CAR-based fixed-point integration mode according to claim 1, characterized in that, The HDR-type integration simultaneously satisfies structural continuity, HA integrity, fixed-point consistency, and no abnormal repetition.

10. The quantitative analysis method for CAR-based fixed-point integration mode according to claim 1, characterized in that, The NHEJ-type integration satisfies any of the following conditions: HA is missing or incomplete; The upstream segment coordinate deviation is greater than the fifth preset value or an insertion / deletion event exists; There are more than two HA_up or HA_down segments.

Images