A method and system for intelligent analysis of flow product CNV based on CNV-SEQ

Abstract

The application relates to the technical field of bioinformatics analysis, in particular to a CNV-intelligent-analysis method and system based on CNV-SEQ; the method comprises the following steps: acquiring alignment reads of a to-be-tested sample through a genome positioning algorithm, performing segmentation processing based on the alignment reads, and acquiring a copy number state section; the copy number state sections in the same chromosome are sorted in ascending order according to starting coordinates to obtain a section list; the copy number state sections in the section list that meet preset conditions are merged to obtain a merged section; the merged section is analyzed, and a report document is output. The application can improve the efficiency of intelligent analysis and interpretation of the flow product CNV.

Description

Technical Field

[0001] This application relates to the field of bioinformatics analysis technology, and in particular to a CNV-SEQ-based intelligent analysis method and system for CNV of flow products. Background Technology

[0002] Chromosomal abnormalities are a significant genetic factor leading to abnormal embryonic development and spontaneous abortion. These include abnormal chromosome number, large chromosome deletions or duplications, and copy number variations (CNVs). Traditional chromosome karyotype analysis methods have long been used for chromosomal abnormality detection, capable of analyzing chromosome number and larger structural abnormalities. However, due to limitations in detection resolution at the chromosome banding level, they struggle to effectively identify smaller copy number variations, thus failing to meet the needs of analyzing minute chromosomal structural abnormalities in the genetic testing of aborted fetuses.

[0003] With the development of next-generation sequencing (NGS) technology, copy number variation sequencing (CNV-SEQ) based on low-depth whole-genome sequencing is increasingly being applied to the genetic analysis of abortion products. CNV-SEQ sequences the genome of the sample to be tested, divides the reference genome into multiple fixed-length data windows (bins), counts the number of sequences within each window, and determines whether there is an increase or decrease in copy number in a corresponding region based on the correlation between sequencing depth and the copy number of chromosomal regions. This technology can cover a large range of genomic regions and detect chromosomal copy number variations through sequencing data analysis.

[0004] For example, patent application CN202510441873.4 discloses a method, device, equipment and storage medium for detecting copy number variations based on low-depth whole genome sequencing. It mainly improves the accuracy of CNV detection under low-depth sequencing conditions by optimizing the selection of reference samples and the baseline construction method. It mainly optimizes the reference baseline construction step in the CNV detection process.

[0005] In addition, patent application CN202311830756.4 discloses a method and system for detecting low-depth whole-genome CNVs. It processes sequencing data and uses data block aggregation to achieve CNV detection, focusing mainly on the data processing process of the CNV detection algorithm itself.

[0006] The existing CNV-SEQ analysis workflow typically includes steps such as sequencing data quality control, genome alignment, sequencing depth normalization, reference baseline construction, CNV region identification, result annotation, and filtering. In the CNV region identification process, statistically variable-based segmentation algorithms, such as Circular Binary Segmentation (CBS), are commonly used to divide consecutive data windows with similar copy number states into CNV fragments.

[0007] However, due to factors such as random fluctuations in sequencing data, variations in sequencing depth in local regions, and the fact that segmentation algorithms determine breakpoints based on statistical differences, the same continuous region of abnormal chromosome copy number may be divided into multiple interconnected CNV fragments, leading to fragmentation of the detection results. For these fragmented results, existing workflows typically require analysts to manually judge and merge the fragments based on chromosome location, fragment length, chimerism ratio, and database information. This not only increases the workload of manual analysis but is also easily influenced by the analyst's experience, limiting the consistency and automation of CNV analysis results and resulting in low efficiency in interpreting the results. Summary of the Invention

[0008] To address the issue of low efficiency in chromosome analysis and interpretation in related technologies, this application provides a CNV-SEQ-based intelligent CNV analysis method and system for flow products.

[0009] Firstly, this application provides a CNV-SEQ-based intelligent analysis method for flow products, employing the following technical solution: A CNV-SEQ-based intelligent CNV analysis method for flow products includes: obtaining alignment reads of the sample to be tested through a genome localization algorithm, performing segmentation processing based on the alignment reads, and obtaining copy number status regions; The copy number status segments within the same chromosome are sorted in ascending order of their starting coordinates to obtain a segment list. Copy number status segments in the segment list that meet preset conditions are merged to obtain merged segments. Analyze the merged sections and output a report document.

[0010] By organizing copy number state segments within the same chromosome according to genomic coordinates and automatically merging segments that meet preset conditions, continuous copy number variations severed by pseudo-breakpoints can be restored, reducing redundant segments that need to be interpreted one by one. Furthermore, the merged results can be directly used for analysis and report generation, reducing the manual transcription and judgment steps between bioinformatics analysis and clinical reporting. Compared to existing CNV-SEQ analysis that directly outputs statistical segmentation results, this method improves interpretation efficiency.

[0011] Optionally, merged segments in the segment list that meet preset conditions to obtain merged segments, including: Identify the anomaly types of copy number state segments and calculate the degree of chimerism of the copy number state segments for each anomaly; The preset conditions include preconditions and postconditions. In response to the fact that the copy number status segments in the segment list meet the preconditions, the copy number status segments are merged using the merging strategy corresponding to the preconditions. Otherwise, determine whether the copy number status segments in the segment list that meet the preset conditions meet the postconditions. If the postconditions are met, merge the copy number status segments using the merging strategy corresponding to the postconditions.

[0012] First, identify segment types such as missing, duplicate, and normal. Calculate the degree of chimerism and merge only segments with abnormal copy number states. At the same time, select different merging strategies in the order of judging the preconditions first and then judging the postconditions if they are not met. This can ensure that high-proportion anomalies and locally adjacent anomalies are handled appropriately, reduce the situation where normal segments or different anomaly types are mistakenly merged, and reduce the invalid calculations caused by continuing to perform local traversal when the preconditions have been met.

[0013] Optionally, the precondition is: obtaining the total length of copy number state segments of the same mutation type in the segment list, wherein the total length is greater than a preset proportion of the total length of all abnormal copy number state segments; The merging strategy corresponding to the preconditions is: All copy number state segments under the same chromosome are merged into a single merged segment, with its starting point being the smallest starting point of all copy number state segments and its ending point being the largest ending point of the copy number state segments.

[0014] The cumulative length ratio of copy number state segments of the same type of abnormality is used as the preliminary judgment basis, and the overall range is restored with the minimum starting point and the maximum ending point to quickly form the merging result; the number of segments of large-scale chromosomal abnormalities and iteration overhead are reduced, so that the output format is more consistent with the actual structural characteristics of whole arm or whole chromosome level abnormalities.

[0015] Optionally, Postconditions include: The continuous merging condition and the gap merging condition are determined as follows: when the coordinates of two adjacent copy number state segments are continuous, it is determined whether the continuous merging condition is met; when the coordinates of two adjacent copy number state segments are not continuous, it is determined whether the gap merging condition is met. If either the continuous merging condition or the gap merging condition is met, the subsequent condition is determined to be met. The merging strategy corresponding to the postcondition is as follows: determine any adjacent copy number state segments, and merge the two adjacent copy number state segments in response to the satisfaction of the postcondition.

[0016] Simultaneously constraining the anomaly type, chimerism level, absolute difference in chimerism, and centromere position or physical spacing can bridge local sequencing blind spots and short-distance pseudo-breakpoints while preventing the incorrect connection of biologically inconsistent segments.

[0017] Optionally, based on the compared read segment, a segmentation process is performed to obtain the copy number status segment, including: Based on the preset binning window, the number of alignment reads in the test sample within each binning window is counted to obtain the counting vector; The count vector is corrected based on the baseline matrix constructed from negative control samples; For any bin window, calculate the logarithmic ratio of the corrected count value of each bin window to the mean count value of the corresponding bin in the baseline matrix, and obtain the logarithmic ratio sequence. Perform cyclic binary segmentation on the logarithmic ratio sequence to obtain the copy number state segment.

[0018] By using a negative control baseline to correct the bin count vector and using the logarithmic ratio of the corrected value to the baseline mean to characterize the copy number shift, the interference of GC content, sequencing depth, library batch and repetitive region preference on segment identification can be reduced, providing a comparable continuous numerical sequence for cyclic binary segmentation.

[0019] Optionally, identifying the anomaly type of the copy number status segment includes: for any copy number status segment, in response to the mean of the logarithmic ratios of the multiple binning windows corresponding to the copy number status segment being less than a preset missing threshold, confirming that the anomaly type of the copy number status segment is missing. If the mean of the logarithmic ratios of multiple bin windows corresponding to the copy number status segment is greater than the preset duplicate determination upper limit, the abnormal type of the copy number status segment is confirmed as duplicate. Otherwise, confirm that the copy number status segment is normal.

[0020] Compared to relying on analysts to observe signal waveforms and judge missing or duplicate signals based on experience, this application quickly classifies the segment status uniformly based on the relationship between the mean logarithmic ratio of multiple bin windows within a segment and the missing or duplicate thresholds. This improves efficiency while reducing the number of normally fluctuating segments entering the subsequent abnormal merging process.

[0021] Optionally, the degree of chimerism of the copy number state segments of each anomaly is calculated, including: For copy number state segments with the anomaly type of missing, the formula for calculating their chimerism is: ; For a copy number state segment with the anomaly type of duplicate, the formula for calculating its degree of nesting is: ; In the formula, , The degree of chimerism corresponds to the missing and duplicate segments respectively, and the value range is limited to [0%, 100%]. Calculation results outside this range are truncated according to the boundary value. Indicates the ploidy of the chromosomes in the sample being tested; This is the mean of the logarithmic ratios of each bin window corresponding to the copy number status segment.

[0022] By combining the segment abnormality type and the chromosome ploidy of the sample to be tested, the log ratio is converted into chimerism, which can reflect the proportion of abnormal cells. This provides a quantitative indicator with unified dimensions for missing and duplicate segments, and provides a basis for subsequent chimerism threshold screening, segment similarity comparison and merging judgment.

[0023] Optionally, the log ratio sequence is subjected to cyclic binary segmentation, including: pairwise matching of bin windows in the sub-intervals on both sides of any candidate breakpoint to obtain matching pair windows, and weighted summation of the difference in log ratio between the two bin windows in the matching pair window to obtain the difference value; In each round of segmentation iteration, the candidate breakpoint with the largest difference value and the permutation test significance level meets the preset threshold is selected for actual segmentation until the preset maximum number of iterations is reached or the interval length does not meet the minimum number of bins.

[0024] Weighting the difference in the logarithmic ratio between the windows on both sides of the candidate breakpoint makes the local step signal closer to the candidate breakpoint contribute more, while the interference from the noise window at a distance is suppressed, thus helping to reduce false breakpoints and oversegmentation.

[0025] Optionally, it also includes: performing multi-source database comparison on the copy number state segment to obtain pathogenicity and gene function annotation information.

[0026] In traditional CNV detection of aborted fetuses, doctors need to manually copy the coordinates of each variant segment and sequentially log into various genome browsers or databases for tedious manual searches and comparisons. This method automates this process by comparing against databases. Each output segment comes with detailed annotations, thereby improving the efficiency of interpretation and analysis.

[0027] Secondly, this application provides a CNV-SEQ-based intelligent analysis system for flow products, employing the following technical solution: A CNV-SEQ-based intelligent CNV analysis system for flow products includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the CNV-SEQ-based intelligent CNV analysis method for flow products as described above.

[0028] The above-mentioned CNV-SEQ-based flow product CNV intelligent analysis method is used to generate a computer program, which is then stored in a memory for loading and execution by a processor. This allows for the creation of a system based on the memory and processor, making it convenient to use.

[0029] This application has the following technical advantages: By organizing the same chromosomal region according to coordinates, first identifying deletions and duplications and quantifying chimerism, and then using the preconditions of high-proportion abnormalities and the postconditions of adjacent abnormalities to form a hierarchical decision, the redundant regions that need to be interpreted one by one are reduced, the burden of manual review is reduced, and the interpretation and processing efficiency is improved. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of a CNV-SEQ-based intelligent analysis method for flow products based on CNV-SEQ.

[0031] Figure 2 This is a logical diagram illustrating the merging judgment in a CNV-SEQ-based intelligent analysis method for flow products according to an embodiment of this application. Detailed Implementation

[0032] This application discloses a CNV-SEQ-based intelligent analysis method for flow products.

[0033] Reference Figure 1 A CNV-SEQ-based intelligent analysis method for flow products includes steps S1-S4.

[0034] In this embodiment, the specific implementation process of the method includes the following steps.

[0035] S1: Obtain the alignment reads of the sample to be tested through a genome localization algorithm, and perform segmentation processing based on the alignment reads to obtain the copy number status segment.

[0036] According to the first preset period (set to 4 hours in this embodiment), the directory is traversed through the fixed path in the storage array to check whether there is a sequencing task information file in a preset structured format (the file is named with the chip number and contains fields such as library number, adapter sequence and business type).

[0037] Once the file is detected, its built-in fields are read and parsed. Only when the value of a specific business type field matches the copy number variation detection business identifier, the data splitting status under the corresponding offline path is further monitored.

[0038] Then, according to the second preset period (set to 5 minutes in this embodiment), the generation status of the preset split completion marker file in the target folder and the change in the number of data files in that path are polled; if it is determined that the number of files remains constant in multiple consecutive polling periods, it is determined that the original sequencing data has been completely split and is ready.

[0039] Once the raw sequencing data has been completely split and ready, the analysis directory for this batch is generated according to the preset batch coding rules (for example, a composite coding method combining business type identifier, year, and sequence number can be used), and the raw data is archived in this directory. Then, the subsequent analysis process is automatically delivered, and a batch delivery success notification message is sent to the preset maintenance personnel.

[0040] It is understandable that raw sequencing data often contains sequencing adapter sequences and low-quality reads.

[0041] Therefore, the raw sequencing data is preprocessed.

[0042] Specifically, the raw sequencing data undergoes sequence quality control to remove reads carrying adapter sequences, low-quality bases, and unresolved nucleotide ratios exceeding limits, resulting in clean read data. Subsequently, the clean reads are mapped to a reference genome for sequence alignment, yielding an alignment result file. Further, the alignment results are sorted according to genomic coordinates. Finally, duplicate reads are marked on the sorted alignment results to avoid duplicate counting interference introduced by library amplification bias. Through the above quality control, alignment, sorting, and deduplication processes, an alignment file is obtained that eliminates physical noise and library preparation biases during sequencing. This alignment file serves as the sample data for downstream copy number quantification, providing a uniform and comparable characteristic basis.

[0043] After the genomic DNA of the sample to be tested is extracted, fragmented and sequenced in high throughput, the set of aligned reads is obtained by genomic localization algorithm (such as read alignment algorithm), which is a conventional technique in this field and will not be described in detail here.

[0044] After obtaining the comparison read segment, a segmentation process is performed based on the comparison read segment to obtain the copy number status segment.

[0045] A baseline matrix was pre-constructed before the segmentation process: Comparative data from no fewer than 100 negative control samples were collected in advance. The negative control samples met the following inclusion criteria: 1. The sample was confirmed to have no abnormal chromosomal segments by existing clinical testing methods (including but not limited to karyotype analysis or chromosome microarray analysis).

[0046] 2. The number of effective aligned reads after sequencing data is aligned to the reference genome (GRCh37 / hg19 version is used in this embodiment) is not less than the preset depth threshold (equivalent to 0.1 times genome coverage in this embodiment).

[0047] 3. Ensure that the sample library preparation batch and sequencing platform model are consistent with the sample to be tested, so as to reduce the impact of batch effects on baseline consistency.

[0048] The alignment reads of the above negative control samples were binned and counted using a 10kb window to obtain the original counting matrix of sample number × bin number. The matrix is ​​then normalized for each row (i.e., each sample) based on its total number of reads, resulting in a baseline matrix composed of relative coverage. The mean and standard deviation of coverage for each binning window across all negative control samples are then calculated to form a baseline reference vector. and Both are persistently stored along with the original count matrix of the negative control sample, serving as a reference for subsequent normalization calculations of the test samples.

[0049] The binning count of the comparison read segment of the sample to be tested is also performed using a 10kb window to obtain the count vector. ( (For binning window numbering), for counting vectors A correction is performed to obtain the corrected count vector. Each count value in the count vector represents the number of aligned reads in the corresponding binning window.

[0050] In this embodiment, the correction process includes GC content correction and principal component projection correction.

[0051] Specifically: GC content correction: Perform GC content (the proportion of guanine (G) and cytosine (C) in a specific DNA fragment) correction: Divide all binning windows into several intervals according to the GC content of their reference sequence (in this embodiment, the division is done in 1% increments), calculate the average coverage of the sample to be tested in each GC interval, and sum the read counts of all bins contained in each GC interval, and divide by the total number of bins in the corresponding interval.

[0052] Subsequently, based on the average coverage of all effective bins across the entire genome, the count values ​​of each bin window were subjected to inter-binary linear scaling correction. Specifically, the scaling correction was performed by using the ratio of the average coverage of the entire genome to the average coverage of a given GC interval as the scaling factor for that GC interval, and then multiplying the original count value of each bin by the scaling factor of its respective GC interval to obtain the GC-corrected count vector. .

[0053] Principal component projection correction: for baseline matrix (Negative control samples × binning window) Perform singular value decomposition, retaining the top performers whose cumulative explained variance reaches the 90% threshold. Each principal component serves as a covariate subspace (in the historical statistics of this embodiment, ...). The value is usually between 5 and 8.

[0054] The sample to be tested After projecting onto the covariate subspace, its projected component within that subspace is subtracted to remove the influence of systematic covariates such as batch effects, uneven sequencing depth, and repetitive region bias, resulting in the corrected count vector. .

[0055] Based on the corrected count vector, calculate the logarithmic ratio and standard score of the test sample within the binning window.

[0056] The formula for calculating the logarithmic ratio of the test samples within any binning window is: ; The formula for calculating the standard score of a sample within any binning window is: ; In the formula, Indicates the first The logarithmic ratio of the samples to be tested within each binning window; Indicates the first The projection-corrected count values ​​of the samples to be tested within each binning window; This indicates that all negative control samples were on the [number]th [day]. Average coverage within each binning window; This indicates that all negative control samples were on the [number]th [day]. The standard deviation of coverage within each bin window.

[0057] for A binning window with a variance less than the preset minimum variance threshold (set to 1% of the mean coverage of the binning window on all negative control samples in this embodiment) is determined to be a low-confidence repetitive region. Its standard score is set to null and it will not participate in the breakpoint determination in subsequent segmentation, but its log ratio is still retained for manual review and reference.

[0058] The logarithmic ratios of each bin form a logarithmic ratio sequence, and the logarithmic ratio sequence is used as the basis for further analysis. Using the input as input, perform cyclic binary partitioning to obtain the copy number state segment.

[0059] The logarithmic ratios of each bin form a logarithmic ratio sequence, and the logarithmic ratio sequence is used as the basis for further analysis. As input, a cyclic binary segmentation process is performed, with the segmentation difference value, weight function, and stopping criterion defined as follows: For any candidate breakpoint, the sequence is divided into two sub-intervals. Binning windows are extracted from the two sub-intervals for pairing calculation. The physical distance weight is applied to the logarithmic ratio difference between each pairing window according to the following formula, and then the weighted difference value of the candidate breakpoint is obtained by summing: ; In the formula: Indicates at candidate breakpoints The weighted difference value at the location; and These represent candidate breakpoints. The resulting left and right subintervals; and Traversing the left sub-interval respectively and right subinterval The current bin window number participating in the pairing; and These represent the binning windows respectively. and The logarithmic ratio; Based on physical distance The calculated physical distance weights (i.e., those including the attenuation constant) (The result of the weight function calculation).

[0060] ; In the formula, Based on physical distance The calculated physical distance weights; The physical distance between any two bin windows that belong to the two sub-intervals and are currently participating in the pairing statistics; A preset distance attenuation constant (valued at 500kb in this embodiment) is used to reduce the interference of long-distance noise windows on breakpoint determination.

[0061] In each round of segmentation, the absolute value of the weighted difference is selected, and the significance level is determined by permutation test. (The number of permutations is set to 1000) The candidate breakpoints are actually segmented, and the above judgment process is recursively executed on the two sub-intervals after segmentation until either of the following stopping conditions is met: First, the number of binning windows contained in the candidate sub-interval is less than the preset minimum number of bins (set to 5 windows in this embodiment, corresponding to a physical span of 50kb); Second, the significance level of the weighted difference value does not reach Third, the number of recursive segmentations of the same interval reaches the preset maximum number of iterations (set to 20 times in this embodiment).

[0062] After the above iterative segmentation, a preliminary sequence of copy number state segments is obtained, presented in binning windows. Each copy number state segment sequence is accompanied by the mean of the log ratio and the mean of the standard score within that segment.

[0063] S2: Perform multi-source database comparison on the copy number state segment to obtain pathogenicity and gene function annotation information.

[0064] The three pathogenicity assessment databases pre-deployed on local services are queried in sequence to compare the genomic coordinate range of the target segment (i.e. the segment whose copy number status is currently to be compared) with the reference coordinate range of the reference genome region in the database.

[0065] Each database pre-stores several known reference genomic regions, as well as clinical evidence records that are mapped one-to-one with each reference genomic region.

[0066] In this embodiment, three databases are deployed. The names, versions, and reference coordinate systems of the three databases are specified as follows: One is an automated scoring tool built based on the ACMG / ClinGen 2019 / 2020 guidelines for interpreting chromosome microarray variations (locally deployed version, with reference genome coordinates GRCh37 / hg19).

[0067] The second is a multi-source annotation integration tool for structural variations (locally deployed version, which also uses GRCh37 / hg19 as the reference coordinate and integrates downstream data sources such as OMIM, DECIPHER, ClinVar and ClinGen dose-sensitive regions).

[0068] The third is a classification and assessment tool based on the ClinGen dose-sensitive region map (local deployment version, with the same reference coordinates GRCh37 / hg19).

[0069] It is understood that the above three databases are all conventional technical means in this field. Those skilled in the art can choose different databases according to the actual needs of the content to be annotated. Therefore, the specific acquisition process and content of the databases will not be described here.

[0070] The reciprocal overlap ratio is used as the hit criterion: only when the overlap ratio between the target segment (i.e., the copy number status segment being compared) and the area included in the database is not less than 70%, it is determined that the database includes the copy number status segment; a data extraction operation is triggered to retrieve the clinical evidence record corresponding to the included area from the database.

[0071] For cases where the overlap ratio is less than 70% but the target segment is completely contained within the database region or the target segment completely contains the database region, they are recorded according to two categories of labels: "partial hit - subset" and "partial hit - superset", and the corresponding clinical evidence records are retrieved in the same way.

[0072] Based on the clinical evidence records retrieved from the database, multidimensional annotation information is added to each target segment, including pathogenicity score, five-level evidence classification (pathogenic, possibly pathogenic, unclear significance, possibly benign, benign), single dose inadequacy (HI) score and its corresponding region, triple dose sensitivity (TS) score and its corresponding region.

[0073] It also includes information such as chromosome naming, variation type, location zone, fragment length, number and names of coding genes contained in the target segment, and data annotation based on the target segment itself.

[0074] When extracting pathogenicity scores and five levels of evidence classification, the system queries in the following order according to preset priorities: the scores and classification results given by the first database are given priority; if the target segment is not included in the database, the results of the second database (structural variation multi-source annotation integration tool) are queried back; if the second database is still not included, the results of the third database (dose-sensitive region classification assessment tool) are queried back.

[0075] If all three databases provide inclusion results for the same segment, but there is a conflict in the five-level evidence classification, the classification level that is more inclined to be pathogenic should be accepted. The original outputs and conflict markers of the three databases should be retained in the annotation record for manual review by the interpreter in subsequent steps.

[0076] It should be noted that a conflict in the five levels of evidence classification refers to a difference of more than one level in the classification levels. For example, database A classifies evidence as pathogenic (Level 1), while database B classifies it as of unclear significance (Level 3). The difference between the two is two levels, and this is considered a conflict.

[0077] If none of the three databases contain the target segment, the pathogenicity score and evidence classification of the segment are marked as null, and the judgment is then made by the fallback rule based on length and coding region in step S6.

[0078] S3: Sort copy number status segments within the same chromosome in ascending order of their starting coordinates to obtain a segment list. Merge copy number status segments in the segment list that meet the preset conditions.

[0079] Since the segmentation algorithm in step S3 is based on a statistical model to identify local fluctuations, the copy number state segment set with annotation information output in step S4 may still contain fragmented segments that have been incorrectly cut due to noise interference.

[0080] Before merging, the true variant type is first determined for each copy number state segment sequence. The determination threshold is calculated based on the chromosome ploidy of the sample to be tested and preset determination parameters.

[0081] Specifically, the calculation formula is as follows: ; ; In the formula, This indicates the chromosome ploidy of the sample being tested. In this embodiment, the value is 2 for a standard diploid sample (that is, a normal person with two chromosomes). The threshold parameter is used to balance detection sensitivity and noise immunity, and in this embodiment, it is set to 0.05.

[0082] After calculation, based on the above values ​​(i.e., chromosome ploidy of 2 and a threshold parameter of 0.05), and substituting them into the formula, the lower limit for deletion determination is approximately -0.018, and the upper limit for duplication determination is approximately +0.018. That is, if the mean log ratio of a certain segment is below -0.018, it is determined as a deletion; above +0.018, it is determined as a duplication; and values ​​in between are determined as normal.

[0083] As the ploidy or threshold parameters are adjusted, the upper and lower boundaries of the judgment will be scaled accordingly, thereby adapting to detection scenarios with different species or different sensitivity requirements.

[0084] Subsequently, the chimerism of the copy number state segment (i.e., the proportion of cells in the sample carrying the copy number abnormality) is calculated according to different abnormality types.

[0085] The calculation method is as follows: Let the mean logarithmic ratio of the copy number state segment be... The ploidy of the sample to be tested is (In this embodiment, the value is 2), then the degree of chimerism corresponding to this segment is calculated according to the mutation type: ; ; In the above formula , The degree of chimerism corresponds to the missing and duplicate segments respectively, and the value range is limited to [0%, 100%]. Calculation results outside this range are truncated according to the boundary value. Indicates the ploidy of the sample to be tested; This is the mean of the logarithmic ratios of each bin window corresponding to the copy number status segment.

[0086] If clinical information confirms that this batch of samples contains maternal contamination (the proportion of maternal contamination is denoted as...), (Based on clinical information associated with step S1 or independent STR auxiliary test results), the chimerism is corrected according to the following formula: ; Indicates the corrected degree of chimerism. Similarly, the boundary [0%, 100%] is used as the input value for subsequent merging and filtering judgments; Indicates the uncorrected degree of nesting; The percentage of pollution originating from the mother plant is used. If no information on the percentage of pollution originating from the mother plant is available, the uncorrected percentage is used directly. As input values, mark "No parent source pollution correction performed" in the merged record.

[0087] After obtaining the abnormality type and corresponding chimerism of each copy number state segment, copy number state segments within the same chromosome that meet the preset conditions are merged based on the abnormality type and chimerism.

[0088] Combination Figure 2 The preset conditions include preconditions and postconditions. In the specific judgment process, the preconditions are judged first. If the preconditions are met, the merging strategy corresponding to the preconditions is used to merge the copy number state segments.

[0089] Prerequisites: The list of segments is obtained by sorting the starting coordinates of the copy number status segments within the same chromosome in ascending order.

[0090] In the first stage of chromosome-level merging, the total length of all abnormal copy number state segments under the two variant types of copy number deletion and copy number duplication is calculated.

[0091] If the sum of the total lengths of all copy number state segments under a certain variant type exceeds a preset proportion (e.g., 80%, with the remainder determined by staff experience), then all copy number state segments under that chromosome are merged into a single chromosome-wide merged segment.

[0092] After merging, the starting point of the merged segment is set to the smallest starting point in the original sequence, and the ending point is set to the largest ending point. The corresponding chimerism is calculated by weighting the physical lengths of each copy number state segment. The variant type that completes the whole chromosome-level merging will be marked as "whole chromosome-level variant" and will skip the second stage of processing described below.

[0093] The process of obtaining the weighted mean can be represented as: ; in, The degree of chimerism is obtained after weighted summation; The total number of copy number state segments participating in the merge; and The first The physical length and degree of interlocking of a copy number state segment.

[0094] Postconditions: For copy number state segments that have not triggered the first stage of processing, determine whether the postcondition is met. If the postcondition is met, then use the merging strategy corresponding to the postcondition to merge the copy number state segments.

[0095] Specifically, the merging strategy corresponding to the postcondition is the adjacent segment merging strategy. In this strategy, an iterative traversal method is used to traverse any two adjacent copy number state segments, and it is determined whether the two adjacent copy number state segments meet the merging condition. If the merging condition is met, the segments are merged until no copy number state segments are merged in a single round of traversal.

[0096] In each round of traversal, any adjacent pairs of segments that are immediately adjacent after sorting (denoted as the preceding segment A and the following segment B, where the end point of A is less than the starting point of B) are judged in turn.

[0097] Specifically, the postconditions include: It is understandable that when determining merging based on postconditions, there are two possible scenarios for the coordinate ranges corresponding to adjacent copy number state segments: continuous coordinates and discontinuous coordinates. This embodiment sets different conditions for each of these two scenarios. For ease of description, the postcondition for adjacent copy number state segments with continuous coordinates is defined as a continuous merging condition. The postcondition for adjacent copy number state segments with discontinuous coordinates is defined as a gap merging condition.

[0098] The conditions for consecutive mergers include: 1. Two adjacent copy number state segments have the same mutation type.

[0099] 2. The degree of chimerism of two adjacent copy number state segments is greater than the preset first threshold.

[0100] 3. The difference in chimerism between two adjacent copy number state segments is less than the second threshold.

[0101] Gap merging conditions include: 1. Two adjacent copy number state segments have the same mutation type.

[0102] 2. The degree of chimerism of two adjacent copy number state segments is greater than the preset first threshold.

[0103] 3. Two adjacent copy number state segments are located within the centromere p11.1 to q11.1 band of the same chromosome, or the physical distance between the start of the posterior segment B and the end of the posterior segment A is less than the physical distance threshold and the chimerism difference between the two adjacent copy number state segments is less than the second threshold.

[0104] In this embodiment, the physical distance threshold is 20kb, but it can be adjusted in other embodiments according to the actual situation.

[0105] The first and second thresholds are set based on experience. In this embodiment, the first threshold is set to 15% and the second threshold is set to 30%.

[0106] Understandably, when either the continuous merging condition or the gap merging condition is met, two adjacent copy number state segments are merged. During the merging operation, the generated new segment inherits the starting point of the preceding segment A, the ending point of the following segment B, and the same mutation type as the original segment. The degree of chimerism of the new segment is calculated by weighting the physical lengths of the original segment A and the following segment B, and the confidence level is taken as the minimum value between the confidence levels of the preceding segment A and the following segment B.

[0107] Furthermore, for the multidimensional annotation information before the merging operation, data aggregation and inheritance processing are performed simultaneously: the names and numbers of coding genes contained in the preceding segment A and the following segment B, as well as the overlapping haplo-inadequate (HI) and triple-dose sensitive (TS) regions, are mathematically combined to form the regional annotation features of the new copy number status segment; for pathogenicity scores and five levels of evidence classification, following the principle of strict clinical risk assessment, the evidence classification records in the original two segments that are more inclined to pathogenicity or have higher risk are automatically compared and accepted as the final pathogenicity annotation of the new copy number status segment.

[0108] If the above merging conditions are not met, preceding segment A and preceding segment B remain independent. Finally, the output is a list of segments integrated after the above two-stage processing.

[0109] For example: Example 1: Suppose that the sum of the physical lengths of the duplicated copy number states in chromosome 13 (physical length approximately 114 Mb) of the sample to be tested is 95 Mb, accounting for 83% of the total chromosome length, which meets the prerequisite. Therefore, all duplicated segments in this chromosome are merged into one chromosome-wide merged segment, with the starting point being the minimum value and the ending point being the maximum value.

[0110] Example 2: Preceding segment A is a deletion on chromosome 8, with coordinates ranging from 20.0 Mb to 20.4 Mb and a chimerism rate of 25%; postceding segment B is also a deletion on chromosome 8, with coordinates ranging from 20.41 Mb to 21.0 Mb and a chimerism rate of 22%. The gap between the two is 10 kb, and the difference in chimerism rate is 3%, which meets the gap merging condition. Therefore, they are merged into a single deletion segment with coordinates ranging from 20.0 Mb to 21.0 Mb.

[0111] Example 3: The preceding segment A is a deletion on chromosome 8, with coordinates ranging from 20,000,000bp to 20,400,000bp and a chimerism of 25%; the following segment B is also a deletion on chromosome 8, with coordinates ranging from 20,400,001bp to 21,000,000bp and a chimerism of 22%. Since their coordinates are continuous and the difference in chimerism is 3%, they meet the contiguous merging condition. Therefore, they are merged into a single deletion segment with coordinates ranging from 20,000,000bp to 21,000,000bp.

[0112] S4: Analyze the merged segments and output a report document.

[0113] First, filter out merged segments with a chimerism rate of less than 10%.

[0114] For the retained merged segments, perform a determination of missing paths and duplicate paths based on their type.

[0115] I. Path for determining missing anomalies: First, check whether the missing merged region contains a protein-coding / functional region. A protein-coding / functional region refers to a sequence region whose coordinate range physically overlaps (i.e., the number of overlapping bases is greater than 0) with the canonical coding sequence (CDS) region of any known transcript in a reference database (such as RefSeq or GENCODE database).

[0116] If the fragment does not contain a protein / functional region: further determine its length. If the fragment length is greater than 500kb, the output is "Manual Evaluation" (i.e., retained); if the fragment length is not greater than 500kb, it is determined as "Not Reported".

[0117] If the merged segment contains a protein / functional region, it is further determined whether the merged segment is located entirely within a region or gene with a very low haploinability score (HI=40, i.e., benign haploinability). If so, it is judged as "not reported"; otherwise, it is output as manual evaluation.

[0118] II. Decision path for duplicate exceptions: Similarly, the first step is to check whether the duplicated merged segment contains protein / functional regions.

[0119] If the fragment does not contain a protein / functional region: further determine its length. If the fragment length is greater than 1Mb, the output is "Manual Evaluation"; if the fragment length is not greater than 1Mb, it is determined as "Not Reported".

[0120] If it contains protein / functional regions, further examination is needed to identify regions or genes with a low three-fold dose sensitivity score (TS=40, i.e., benign three-fold dose sensitivity).

[0121] If no region or gene with TS=40 exists, the merged segment will be output as "Manual Evaluation".

[0122] If a region or gene with TS=40 exists, then the following exclusionary condition checks are performed sequentially: Condition A: The merged segment is consistent with the region / gene range of TS=40. If this condition is met, the condition is determined as "not reported"; otherwise, proceed to the determination of Condition B.

[0123] Condition B: The merged segment is contained within the region / gene of TS=40 (i.e., the coordinate range of the merged segment falls within the reference coordinate range of the region / gene of TS=40), and the breakpoint is not in the protein-coding region. If this condition is met, it is determined as "not reported"; otherwise, proceed to the determination of Condition C. Condition C: The merged region is larger than the region / gene with TS=40, but there is no protein encoded in the region it extends beyond. If this condition is met, it is judged as "not reported"; otherwise, it is output for manual evaluation.

[0124] It is understandable that not reporting the status indicates that the merged segment is at low clinical risk. By not outputting this status, redundant data is eliminated, reducing the burden of data processing and subsequent review.

[0125] Conversely, manually assessed variants that fail to meet the aforementioned safety filtering criteria pose a potential clinical pathogenic risk. These variants, along with the multidimensional annotation information obtained in the preceding processing, are then output to the final candidate variant set for clinical interpreters or medical experts to comprehensively assess the subject's actual clinical phenotype.

[0126] Finally, the filtered results are integrated with clinical information and rendered to generate a structured report document.

[0127] The system receives the merged segment set and clinical information files pre-organized and uploaded to a fixed path by interpreters. These clinical information files record elements such as sample number, submitting institution, whether it is a uniparental diploid, maternal contamination ratio, amplified fragment length polymorphism auxiliary test results, and report template type. This step first writes the quality control indicators, preliminary segmentation results, merged results, and filtered results into different worksheets within the same results file, highlighting merged segments within a pre-defined whitelist to form a single-sample result summary file. Subsequently, the filtered results of all samples within a batch are read and integrated with the corresponding clinical information files to form a batch result summary table. Finally, based on the report template type corresponding to the submitting institution for that batch of samples, the data binding engine is invoked to render and populate the information from the batch result summary table into the corresponding template, generating a structured batch report document.

[0128] It should be noted that the batch report generation process described in this step can also be run independently of the preceding automated delivery stage: the interpreter can manually add or delete merged sections in the batch result summary table and then re-trigger the report rendering to obtain the corresponding report document. This method is more flexible in scenarios where manual review and confirmation of individual merged sections are required.

[0129] This application also discloses a CNV-SEQ-based intelligent CNV analysis system for flow products, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a CNV-SEQ-based intelligent CNV analysis method for flow products according to this application.

[0130] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0131] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A CNV-SEQ-based intelligent analysis method for flow products, characterized in that, include: The alignment reads of the sample to be tested are obtained through a genome localization algorithm, and segmentation processing is performed based on the alignment reads to obtain copy number status segments. The copy number status segments within the same chromosome are sorted in ascending order of their starting coordinates to obtain a segment list. Copy number status segments in the segment list that meet preset conditions are merged to obtain merged segments. Analyze the merged sections and output a report document.

2. The CNV-SEQ-based intelligent analysis method for flow products according to claim 1, characterized in that, Merged segments are obtained by merging the copy count status segments in the segment list that meet preset conditions, including: Identify the anomaly types of copy number state segments and calculate the degree of chimerism of the copy number state segments for each anomaly; The preset conditions include preconditions and postconditions. In response to the fact that the copy number status segments in the segment list meet the preconditions, the copy number status segments are merged using the merging strategy corresponding to the preconditions. Otherwise, determine whether the copy number status segments in the segment list that meet the preset conditions meet the postconditions. If the postconditions are met, merge the copy number status segments using the merging strategy corresponding to the postconditions.

3. The CNV-SEQ-based intelligent analysis method for flow products according to claim 2, characterized in that, The prerequisite is: obtain the total length of the copy number state segments of the same mutation type in the segment list, and the total length is greater than a preset ratio of the total length of all abnormal copy number state segments; The merging strategy corresponding to the preconditions is: All copy number state segments under the same chromosome are merged into a single merged segment, with its starting point being the smallest starting point of all copy number state segments and its ending point being the largest ending point of the copy number state segments.

4. The CNV-SEQ-based intelligent analysis method for flow products according to claim 2, characterized in that, Postconditions include: The continuous merging condition and the gap merging condition are determined as follows: when the coordinates of two adjacent copy number state segments are continuous, it is determined whether the continuous merging condition is met; when the coordinates of two adjacent copy number state segments are not continuous, it is determined whether the gap merging condition is met. If either the continuous merging condition or the gap merging condition is met, the subsequent condition is determined to be met. The merging strategy corresponding to the postcondition is as follows: determine any adjacent copy number state segments, and merge the two adjacent copy number state segments in response to the satisfaction of the postcondition.

5. A CNV-SEQ-based intelligent analysis method for flow products, characterized in that, Based on the comparison read segment, segmentation processing is performed to obtain the copy number status segment, including: Based on the preset binning window, the number of alignment reads in the test sample within each binning window is counted to obtain the counting vector; The count vector is corrected based on the baseline matrix constructed from negative control samples; For any bin window, calculate the logarithmic ratio of the corrected count value of each bin window to the mean count value of the corresponding bin in the baseline matrix, and obtain the logarithmic ratio sequence. Perform cyclic binary segmentation on the logarithmic ratio sequence to obtain the copy number state segment.

6. The CNV-SEQ-based intelligent analysis method for flow products according to claim 5, characterized in that, Identifying the anomaly type of copy number status segment includes: for any copy number status segment, if the mean of the logarithmic ratios of the multiple binning windows corresponding to the copy number status segment is less than the preset missing threshold, the anomaly type of the copy number status segment is confirmed to be missing. If the mean of the logarithmic ratios of multiple bin windows corresponding to the copy number status segment is greater than the preset duplicate determination upper limit, the abnormal type of the copy number status segment is confirmed as duplicate. Otherwise, confirm that the copy number status segment is normal.

7. The CNV-SEQ-based intelligent analysis method for flow products according to claim 6, characterized in that, Calculate the degree of chimerism of each anomaly's copy number state segment, including: For copy number state segments with the anomaly type of missing, the formula for calculating their chimerism is: ； For a copy number state segment with the anomaly type of duplicate, the formula for calculating its degree of nesting is: ； In the formula, , The degree of chimerism corresponds to the missing and duplicate segments respectively, and the value range is limited to [0%, 100%]. Calculation results outside this range are truncated according to the boundary value. Indicates the ploidy of the chromosomes in the sample being tested; This is the mean of the logarithmic ratios of each bin window corresponding to the copy number status segment.

8. The CNV-SEQ-based intelligent analysis method for flow products according to claim 5, characterized in that, The log ratio sequence is subjected to cyclic binary segmentation, including: pairwise matching of bin windows in the sub-intervals on both sides of any candidate breakpoint to obtain matching pair windows, and weighted summation of the difference in log ratio between the two bin windows in the matching pair window to obtain the difference value; In each round of segmentation iteration, the candidate breakpoint with the largest difference value and the permutation test significance level meets the preset threshold is selected for actual segmentation until the preset maximum number of iterations is reached or the interval length does not meet the minimum number of bins.

9. The CNV-SEQ-based intelligent analysis method for flow products according to claim 7, characterized in that, Also includes: Multi-source database comparison is performed on the copy number state segment to obtain pathogenicity and gene function annotation information.

10. A CNV-SEQ-based intelligent analysis system for flow products, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a CNV-SEQ-based intelligent analysis method for flow products according to any one of claims 1-9.

Images