Method for accurate breakpoint identification of copy number variations and application thereof

Abstract

The application discloses a method for accurately identifying a copy number variation (CNV) breakpoint and application thereof. The method comprises the following steps: expanding a globally identified candidate CNV region, performing window merging on the expanded CNV region, and identifying a CNV accurate breakpoint based on the read depth of the merged window in the expanded CNV region. The application can ensure the accuracy of the breakpoint identification in the local CNV identification process by dynamically expanding the candidate CNV region. The machine learning unsupervised clustering method is innovatively used to cluster and merge adjacent fragments according to the fragment characteristics, so that the accurate CNV fragment can be finally identified, the identification error is lower than that of the global CNV identification, the identification accuracy can be adjusted according to the local CNV identification window merging parameter, and the sensitivity is higher and the operation is simpler.

Description

Technical Field

[0001] This invention belongs to the field of biotechnology and relates to a method for accurate identification of copy number variation breakpoints and its application. Background Technology

[0002] Copy number variations (CNVs) are structural variations in the human genome, associated with many complex human diseases such as autism, Parkinson's disease, schizophrenia, and cancer. They are a key focus of preimplantation genetic testing (PGT-A), non-invasive prenatal testing (NIPT), and miscarriage tissue genetic testing. CNVs are defined as deletions or duplications of DNA fragments between 1 and 5 Mb, accounting for 12-16% of the entire human genome. Traditionally, CNVs have been identified using cytogenetic techniques such as karyotype analysis and fluorescence in situ hybridization (FISH), array comparative genomic hybridization, or single nucleotide polymorphism array methods. However, these methods are not ideal due to hybridization noise, limited genome coverage, and low resolution. Unlike traditional methods, next-generation sequencing (NGS) technology produces short reads with higher resolution, offering a potential advantage for accurate detection of CNV regions.

[0003] CNV detection based on NGS sequencing can be categorized into four strategies: PEM (Paired End Mapping), SR (Splitting Reads), Read Coverage Depth (RD), and De novo Sequence Assembly (AS). In clinical applications, the RD strategy is commonly used for CNV detection. The basic idea is that regions with increased copy numbers will achieve higher read depths compared to normal regions, while regions with lost copy numbers will have lower read depths. Specifically, the first step in RD-based CNV detection is to align the sequencing sequence to a reference genome. Then, the average read count within a fixed window (bin) is calculated to obtain the read depth (RD) signal, which is normalized based on the GC content within the bin. After eliminating GC bias, segmentation is performed. The ultimate goal of this method is to cluster adjacent bins with similar RD signals into the same segment. Popular segmentation algorithms include Cyclic Binary Segmentation (CBS) (Venkatraman and Olshen, 2007), Mean-shift (Comaniciu and Meer, 2002), Hidden Markov Models (HMM), and LASSO regression.

[0004] In clinical data, due to low sequence coverage and misalignment, many existing methods for CNV identification based on the RD strategy have the following problems: (1) Signals with small fluctuations in copy number loss / duplication are easily smoothed out during global segmentation, resulting in low sensitivity. (2) Compared with larger CNVs, small CNVs show inconspicuous signals and are easily smoothed out by large fragments. (3) Existing methods rely on the size of the global merging window when identifying global CNVs, and the identified breakpoint location may have a large error compared with the actual breakpoint location.

[0005] In summary, current CNV identification methods suffer from several drawbacks, including the easy neglect of small fluctuations in copy number loss / duplication, poor sensitivity, and significant discrepancies between the identified breakpoint location and the actual breakpoint location. Developing a highly sensitive and accurate method for CNV breakpoint identification has become a pressing issue in the field of biotechnology. Summary of the Invention

[0006] To address the shortcomings of existing technologies and practical needs, this invention provides a method for accurate identification of copy number variation breakpoints and its application. This method can accurately identify CNV breakpoints with a lower identification error than global CNV identification. The identification accuracy can be adjusted based on the merging parameters of the local CNV identification window.

[0007] To achieve this objective, the present invention adopts the following technical solution:

[0008] In a first aspect, the present invention provides a method for accurate CNV breakpoint identification, the method comprising:

[0009] The candidate CNV region for global recognition is expanded, and the expanded CNV region is windowed. Based on the reading depth of the merged window within the expanded CNV region, the precise breakpoint of the CNV is identified.

[0010] This invention dynamically expands candidate CNV regions, ensuring the accuracy of breakpoint identification during local CNV identification. It innovatively employs an unsupervised clustering method based on machine learning to cluster and merge adjacent segments according to segment features, ultimately identifying accurate CNV segments with a lower identification error than global CNV identification. Its identification accuracy can be adjusted according to the merging parameters of the local CNV identification window, resulting in higher sensitivity and simpler operation.

[0011] Preferably, the extended global identification of candidate CNV regions includes the following steps:

[0012] (1) When a candidate CNV region is smaller than a specified size, the size of the candidate CNV region is expanded to the left and right of the original CNV region to form a new CNV region, denoted as region.new ;

[0013] (2) When the candidate CNV region is larger than the specified size, take 0.5-1.5Mb of area upstream and downstream of the original CNV region to form a new left region, denoted as region. left A new right region is formed by taking 0.5-1.5Mb upstream and downstream of the original CNV region, and is denoted as region. right .

[0014] The point values ​​in the above 0.5-1.5Mb range can specifically be 0.5Mb, 0.6Mb, 0.8Mb, 1Mb, 1.2Mb, 1.4Mb, 1.5Mb, etc.

[0015] Preferably, the window merging includes the following steps:

[0016] (1) The expanded region is merged according to the specified merge window size and the step window size to form the RD of the sample to be analyzed after GC correction under the merge window and the RD of the reference set after GC correction, wherein the RD represents the reading depth.

[0017] (2) Calculate the RR value and logRR value under the merged window, as shown in the formula:

[0018] RR = NormReads sample / Median ref

[0019] logRR = log2(NormReads) sample / Median ref )

[0020] Where RR represents the ratio of the corrected Reads of the processed sample to the reference sample within the window, and NormReads represents the NormReads of the sample to be analyzed within the window. sample GC-corrected, normalized values; Median ref This represents the median value of multiple reference sample reads within the window after GC correction and normalization.

[0021] Preferably, the identification of precise CNV breakpoints includes: using an algorithm to segment the logRR obtained after window merging, obtaining the RR value of each segment, clustering based on the obtained segment RR values, merging the clustering results, obtaining new CNV regions from the merged results, and outputting new CNV region information based on the new CNV breakpoint positions.

[0022] Preferably, the algorithm includes a cyclic binary partitioning algorithm.

[0023] Preferably, the merging criteria are:

[0024] (1) When two adjacent segments belong to the same cluster, merge the two adjacent segments into one segment;

[0025] (2) Calculate the RR value of the merged fragments in (1);

[0026] (3) Repeat steps (1) and (2) until no adjacent segments belong to the same cluster in the clustering results.

[0027] Preferably, the clustering algorithm includes the kmeans algorithm.

[0028] Preferably, the criteria for obtaining the new CNV region are:

[0029] (1) When the candidate CNV region is smaller than the specified size, select the region with the largest absolute value of logRR from the merged result as the final CNV region;

[0030] (2) When the candidate CNV region is larger than the specified size, the starting point of the region with the largest absolute value of logRR in the left region is selected as the starting point of the new CNV, and the ending point of the region with the largest absolute value of logRR in the right region is selected as the ending point of the new CNV.

[0031] In a second aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, the computer program being used to perform the following steps:

[0032] (1) Expand the candidate CNV region for global identification:

[0033] When a candidate CNV region is smaller than a specified size, the candidate CNV region is expanded to the left and right of the original CNV region to form a new CNV region, denoted as region. new If the candidate CNV region is larger than the specified size, a new left region is formed by taking 1Mb of space upstream and downstream of the original CNV region. left A new right region is formed by taking 1Mb of space upstream and downstream of the original CNV region, denoted as region. right ;

[0034] (2) Merge the windows of the expanded CNV region:

[0035] The expanded regions are merged according to the specified merge window size and step window size to form the RD of the sample to be analyzed after GC correction under the merge window, and the RD of the reference set after GC correction, wherein the RD represents the read depth.

[0036] Calculate the RR value and logRR value under the merged window, as shown in the formula:

[0037] RR = NormReads sample / Median ref

[0038] logRR = log2(NormReads) sample / Median ref )

[0039] Where RR represents the ratio of the corrected Reads of the processed sample to the reference sample within the window, and NormReads represents the NormReads of the sample to be analyzed within the window. sample GC-corrected, normalized values; Median ref This represents the median value of multiple reference sample reads within the window after GC correction and normalization.

[0040] (3) Use the algorithm to segment the logRR obtained after window merging, obtain the RR value of each segment, perform clustering based on the obtained segment RR values, merge the clustering results, obtain new CNV regions in the merged results, and output new CNV region information according to the new CNV breakpoint position. The algorithm includes a cyclic binary segmentation algorithm.

[0041] Thirdly, the present invention provides a method for precise CNV breakpoint identification as described in the first aspect or an application of a computer-readable storage medium as described in the second aspect in genome detection.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] This invention dynamically expands candidate CNV regions, ensuring the accuracy of breakpoint identification during local CNV identification. It innovatively employs an unsupervised clustering method based on machine learning to cluster and merge adjacent segments according to segment features, ultimately identifying accurate CNV segments with a lower identification error than global CNV identification. Its identification accuracy can be adjusted according to the merging parameters of the local CNV identification window, resulting in higher sensitivity and simpler operation. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of the technical principle of the present invention;

[0045] Figure 2AThis is a statistical result of the average positional error between the simulated breakpoints and the global and local CNV breakpoints identified based on simulated data distribution.

[0046] Figure 2B This is a statistical result of the average positional error between the simulated breakpoints and the global and local CNV breakpoints identified based on simulated data distribution. Detailed Implementation

[0047] To further illustrate the technical means and effects of this invention, the following description, in conjunction with embodiments and accompanying drawings, provides a further explanation of the invention. It is understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it.

[0048] Where specific techniques or conditions are not specified in the examples, they shall be performed in accordance with the techniques or conditions described in the literature in this field, or in accordance with the product instructions. Reagents or instruments whose manufacturers are not specified are all conventional products that can be purchased through legitimate channels.

[0049] Example 1

[0050] This embodiment performs precise CNV breakpoint identification on simulated samples. The specific steps are as follows:

[0051] 1. Sample DNA library construction and sequencing

[0052] In this embodiment, the sample used for testing was peripheral blood. The CNV-seq library preparation method was performed according to the product instructions provided by Suzhou Becon Medical Devices Co., Ltd., and mainly included the following steps: (1) genomic DNA extraction; (2) addition of fragmentation enzyme; (3) adapter ligation; and (4) library amplification. Sequencing was performed using the DA500 high-throughput sequencing platform of Suzhou Becon Medical Devices Co., Ltd. The sequencing type was single-end sequencing 100, and the number of sequences was 10Mb.

[0053] 2. Sequence alignment

[0054] The cut data was compared with the human reference genome (UCSC hg19) using bwa, and the resulting BAM files were sorted. The BamDuplicates module was used to remove duplicates.

[0055] 3. Window division

[0056] (1) Divide the human genome into non-overlapping regions of 10kb fragment size and count the GC content of each window;

[0057] (2) Count the number of sequencing sequences (RD) in each window of the bam file obtained in 2 under the 10kb window in 3.1;

[0058] (3) Merge files 3.1 and 3.2 to obtain the GC content and RD for each window.

[0059] 4. CNV Simulation

[0060] Based on the RD files obtained in each window, CNV simulation is performed on the RD files to simulate CNVs with different CNV splicing rates (30%, 60%, 80%) and CNV sizes (500kb, 1Mb, 4Mb, 10Mb).

[0061] 5. Reference Library Construction

[0062] (1) Obtain RD information in a 10kb window from 100 negative samples from both males and females;

[0063] (2) Perform GC correction on the RD of each sample;

[0064] (3) Standardize the RD after GC correction for each sample; the formula is as follows:

[0065] RD normal =RD / median(RD)

[0066] (4) Calculate the RD of 100 samples in each window normal The median value is used as the denominator in the final calculation of logRR. The formula is as follows:

[0067] Median ref =median(RD) normal )

[0068] 6. Global CNV Recognition

[0069] (1) Window merging: merge windows in 200kb increments, moving 100kb at a time.

[0070] (2) Perform GC correction and standardization on the RD under the merged window of the samples to be detected to obtain NormReads. sample

[0071] (3) Calculate RR, log2RR; the formula is as follows:

[0072] RR = NormReads sample / Median ref

[0073] log2RR = log2(NormReads) sample / Median ref )

[0074] (4) The genome is segmented using CBS based on logRR under the merge window, where CBS uses the segment function in the DNAcopy package in R.

[0075] (5) Calculate seg.means and mosaic for each segment after CBS segmentation; select segments with mosaic > 30 as candidate CNV regions. The formulas for calculating seg.means and mosaic are as follows;

[0076] seg.means = sum(log2RR) / n;

[0077] Where n represents the number of 10kb windows in the CNV region.

[0078] mosaic = 100 * |2 - 2 * 2^seg.mean|

[0079] 7. CNV area expansion:

[0080] (1) When the candidate CNV region is less than 1Mb, the size of the candidate CNV region is expanded to the left and right of the original CNV region to form a new CNV region, denoted as region. new ;

[0081] (2) When the candidate CNV region is greater than 1Mb, take 1Mb of the upstream and downstream regions from the left end of the original CNV region to form a new left region, denoted as region. left A new right region is formed by taking 1Mb of space upstream and downstream of the original CNV region, denoted as region. right .

[0082] 8. Expanded CNV region window sliding merge:

[0083] (1) To ensure that the breakpoint error is smaller, when merging the windows again, we select 50kb and step with a step size of 10kb to finally form the RD after GC correction of the sample to be analyzed under the merged window and the RD after GC correction of the reference set.

[0084] (2) Calculate the RR value and logRR value under the merged window, as shown in the formula:

[0085] RR = NormReads sample / Median ref

[0086] log2RR = log2(NormReads) sample / Median ref )

[0087] NormReads sample This represents the RD value of the sample to be analyzed within this window after GC correction and normalization; Median ref This represents the median value of multiple reference samples RD within the window after GC correction and normalization.

[0088] 9. Precise breakpoint identification

[0089] (1) Use the CBS (circular binary segmentation) algorithm to segment the log2RR obtained from module 3 to obtain the RR value of each segment;

[0090] (2) Cluster the fragment RR values ​​obtained in 4.1 based on kmeans, and merge the clustering results. The merging process follows the following principles:

[0091] ① When two adjacent segments belong to the same k-means cluster, merge the two adjacent segments into one segment;

[0092] ② Calculate the RR value of the merged fragments from step ①;

[0093] ③ Repeat steps ① and ② until no adjacent segments belong to the same cluster in the clustering results;

[0094] (3) In the result of merging in 4.2, obtain the new CNV region, specifically following the principles below;

[0095] ① When the candidate CNV region is less than 1Mb, select the region with the largest absolute value of log2RR from the results in 4.2 as the final CNV region.

[0096] ② When the candidate CNV region is greater than 1Mb; the starting point of the region with the largest absolute value of log2RR in the left region is selected as the starting point of the new CNV; the ending point of the region with the largest absolute value of log2RR in the right region is selected as the ending point of the new CNV.

[0097] (4) Based on the results of 4.3, obtain the new CNV breakpoint location and output the new CNV region information.

[0098] 10. Compare the positional errors of the breakpoints between the globally and locally identified CNV regions and the simulated CNV regions, such as... Figure 2A and Figure 2BAs shown, the error between the breakpoint location and the model location of most locally identified breakpoints is lower than that of global identification; and when the CNV splicing rate is greater than 80%, the error of most identified breakpoint locations is less than 10kb. The method of this invention can accurately identify CNV breakpoints, and its identification error is lower than that of global CNV identification.

[0099] Example 2

[0100] This embodiment performs precise CNV breakpoint identification on real samples. The specific steps are as follows:

[0101] 1. Routine library construction: Select 3 peripheral blood samples for DNA library construction and sequencing, following the same steps as 1-9 in Example 1; the difference is that real samples do not require the CNV simulation process in Example 1.

[0102] 2. Phbol-seq Library Construction: The phbol-seq library construction method was performed according to the product instructions provided by Suzhou Beikang Medical Devices Co., Ltd., and mainly included the following steps: (1) Genomic DNA extraction and dilution; (2) Addition of fragmentation enzyme; (3) Hybridization with magnetic beads with molecular tags; (4) Enzyme release; (5) Adapter ligation; (6) Library amplification. Sequencing was performed using the DA5000 high-throughput sequencing platform of Suzhou Beikang Medical Devices Co., Ltd. The sequencing type was paired-end sequencing 100, and the number of sequences was approximately 80 Gb.

[0103] 3. Analysis of phbol-seq results

[0104] (1) Whole genome data analysis

[0105] Based on the characteristics of Phbol-seq tagging, barcode molecular tag sequences and fixed sequences were removed from the FastQ files of read2 using a barcode table, and the barcode information was then marked on the readname of the FastQ file. Low-quality sequences were filtered out from the isolated FastQ files. The BWA software (v0.7.17-r1188) was used to align the FastQ files to the reference genome (hg19), generating a BAM file. The aligned sequences underwent PCR deduplication and BQSR calibration (GATKv4.1.4.0-local). GATK software was used to perform SNV analysis on the BQSR-calibrated BAM file, generating a VCF file.

[0106] (2) CNV Analysis

[0107] The number of unique matches in each window region is compared between the test sample and the control sample. The Log2RR value is calculated, and the location of the broken window in the sample is found using the CBS (Circular Binary Segmentation) algorithm. The windows are then merged, and the copy number of each region is calculated. Based on the copy number results, it is determined whether each region is normal and whether there are duplicates, deletions, or chimeric duplicates or deletions.

[0108] (3) CNV Precise Fracture Point Analysis

[0109] Based on the CNV analysis results, aligned sequences S1 within a 10kb range near the CNV breakpoint are extracted. The cases where paired sequences align to both ends of the CNV breakpoint are calculated and used as sequence fragments S2 supporting CNV duplication and deletion events. The S2 sequence fragments are then aggregated to calculate the soft-cut sequence fragments S3. The chromosome location where the same soft cut occurs in S3 is the precise CNV breakpoint location. If there are fewer than two sequences in S3 supporting the same soft cut location, the sequence location in S3 closest to the CNV breakpoint is selected as the precise CNV breakpoint (in this case, the precise breakpoint error does not exceed 300-500bp).

[0110] (4) Precise breakpoint analysis of CNV based on long fragment sequencing

[0111] Since the breakpoints in some samples occur in the repeat regions of chromosomes, the location of the CNV breakpoint cannot be directly found by comparing the differences between the paired ends in the S1 sequence of step (3). Therefore, a long DNA fragment sequence (with the same DNA molecular tag) that spans both ends of the CNV breakpoint is calculated as the sequence fragment S2 supporting the CNV duplication and deletion events. The S2 sequence fragments are combined to calculate the sequence fragment S3 that has undergone soft cutting. The chromosome position where the same soft cutting occurs in S3 is the location of the precise CNV breakpoint. If there are fewer than two sequences in S3 that support the same soft cutting position, the sequence position in S3 closest to the CNV breakpoint is selected as the precise CNV breakpoint (at this time, the error of the precise breakpoint does not exceed 1kp).

[0112] 4. The positional errors of the CNV regions identified globally and locally are compared with those identified by phbol-seq. The results are shown in Table 1.

[0113] Table 1

[0114]

[0115] phbol-seq start: phbol-seq identifies the starting position (10kb) of CNV;

[0116] phbol-seq end: phbol-seq identifies the end position of the CNV (10kb);

[0117] Identity start: The starting position (10kb) for regular database creation and CNV identification;

[0118] Identity end: The termination position (10kb) for regular database creation and CNV identification;

[0119] Start distance: The difference between the starting position of CNV identification using conventional database construction and the starting position of CNV identification using phbol-seq;

[0120] End distance: The error between the ending position of CNV identification using conventional database construction and the ending position of CNV identification using phbol-seq;

[0121] Method: The conventional methods used for CNV identification during database construction (global / local).

[0122] As shown in Table 1, when compared with phbol-seq as the gold standard, the breakpoints identified locally are closer to those identified by phbol-seq than those identified globally. This indicates that the method of the present invention can accurately identify CNV breakpoints, and its identification error is lower than that of global CNV identification.

[0123] In summary, the method of the present invention can accurately identify CNV breakpoints with a lower identification error than that of global CNV identification, and its identification accuracy can be adjusted according to the merging parameters of the local CNV identification window.

[0124] The applicant declares that the detailed method of the present invention is illustrated by the above embodiments, but the present invention is not limited to the above detailed method, that is, it does not mean that the present invention must rely on the above detailed method to be implemented. Those skilled in the art should understand that any improvements to the present invention, equivalent substitutions of the raw materials of the product of the present invention, addition of auxiliary components, selection of specific methods, etc., all fall within the protection scope and disclosure scope of the present invention.

Claims

1. A method for precise CNV breakpoint identification, characterized in that, The method includes: Expand the candidate CNV region for global recognition, merge the windows of the expanded CNV region, and identify the precise breakpoint of CNV based on the reading depth of the merged windows within the expanded CNV region. The extended global identification of candidate CNV regions includes the following steps: (1) When the candidate CNV region is smaller than the specified size, the size of the candidate CNV region is expanded to the left and right of the original CNV region to form a new CNV region, denoted as region. new ; (2) When the candidate CNV region is larger than the specified size, take 0.5-1.5 Mb of area upstream and downstream of the original CNV region to form a new left region, denoted as region. left A new right region, denoted as region, is formed by taking an area 0.5-1.5 Mb upstream and downstream of the original CNV region. right ; The window merging includes the following steps: (1) The expanded regions are merged according to the specified merge window size and the step window size to form the RD of the sample to be analyzed after GC correction under the merge window and the RD of the reference set after GC correction, wherein the RD represents the reading depth; (2) Calculate the RR value and logRR value under the merged window, as shown in the formula: RR = NormReads sample / Median ref logRR = log2(NormReads sample / Median ref ) Where RR represents the ratio of the corrected Reads of the processed sample to the reference sample within the window, and NormReads represents the NormReads of the sample to be analyzed within the window. sample GC-corrected, normalized values; Median ref This represents the median value of multiple reference sample reads within the window after GC correction and normalization. The identification of precise CNV breakpoints includes: using an algorithm to segment the logRR obtained after window merging, obtaining the RR value of each segment, clustering based on the obtained segment RR values, merging the clustering results, obtaining new CNV regions from the merged results, and outputting new CNV region information based on the new CNV breakpoint positions.

2. The method according to claim 1, characterized in that, The algorithm includes a cyclic binary partitioning algorithm.

3. The method according to claim 1, characterized in that, The criteria for the merger are as follows: (1) When two adjacent segments belong to the same cluster, merge the two adjacent segments into one segment; (2) Calculate the RR value of the merged fragments in (1); (3) Repeat steps (1) and (2) until no adjacent segments belong to the same cluster in the clustering results.

4. The method according to claim 3, characterized in that, The clustering algorithm includes the kmeans algorithm.

5. The method according to claim 1, characterized in that, The criteria for obtaining the new CNV region are as follows: (1) When the candidate CNV region is smaller than the specified size, select the region with the largest absolute value of logRR from the merged results as the final CNV region; (2) When the candidate CNV region is larger than the specified size, the starting point of the region with the largest absolute value of logRR in the left region is selected as the starting point of the new CNV, and the ending point of the region with the largest absolute value of logRR in the right region is selected as the ending point of the new CNV.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that performs the following steps: (1) Expand the candidate CNV region for global identification: When a candidate CNV region is smaller than a specified size, the candidate CNV region is expanded to the left and right of the original CNV region to form a new CNV region, denoted as region. new If the candidate CNV region is larger than the specified size, a new left region is formed by taking an area 0.5-1.5 Mb upstream and downstream of the original CNV region, denoted as region. left A new right region, denoted as region, is formed by taking an area 0.5-1.5 Mb upstream and downstream of the original CNV region. right ; (2) Merge the windows of the expanded CNV region: The expanded regions are merged according to the specified merge window size and step window size to form the RD of the sample to be analyzed after GC correction under the merge window, and the RD of the reference set after GC correction, wherein the RD represents the read depth. Calculate the RR value and logRR value under the merged window, as shown in the formula: RR = NormReads sample / Median ref logRR = log2(NormReads sample / Median ref ) Where RR represents the ratio of the corrected Reads of the processed sample to the reference sample within the window, and NormReads represents the NormReads of the sample to be analyzed within the window. sample GC-corrected, normalized values; Median ref This represents the median value of multiple reference sample reads within the window after GC correction and normalization. (3) Use the algorithm to segment the logRR obtained after window merging, obtain the RR value of each segment, perform clustering based on the obtained segment RR value, merge the clustering results, obtain the new CNV region in the merged result, and output the new CNV region information according to the new CNV breakpoint position. The algorithm includes the cyclic binary segmentation algorithm.

7. The method for precise CNV breakpoint identification according to any one of claims 1-5 or the application of the computer-readable storage medium according to claim 6 in genome detection.

Images