Methods, systems, devices, and media for simultaneous detection of copy number variations and genomic homozygosity regions

Abstract

The application discloses a method, system, device and medium for simultaneously detecting copy number variation and genome homozygous region, comprising the following steps: step S1, constructing a DNA probe set covering the whole genome, being highly uniformly distributed, having stable sequence characteristics and being rich in polymorphism information; step S2, filtering and processing original high-throughput sequencing data to convert into a standardized signal matrix for downstream analysis; step S3, applying independent signal processing algorithms to correct the original sequencing depth matrix and the original heterozygosity index matrix respectively through population geometric median normalization strategy; step S4, identifying CNV and ROH events respectively based on a statistical inference model, and determining UPD through logical integration; and step S5, system performance verification. The method and system provide an efficient, reliable and economical detection solution for the molecular etiology diagnosis of genetic diseases, developmental retardation, congenital malformation and tumor and the like.

Description

Technical Field

[0001] This invention relates to the field of bioinformatics, and more particularly to methods, systems, devices, and media for simultaneously detecting copy number variations and homozygous regions of the genome. Background Technology

[0002] The human genome is not absolutely stable and immutable; it contains numerous structural variations (SVs), among which copy number variations (CNVs) and homozygous regions (ROHs) are the two most common and clinically significant. Copy number variations refer to increases (duplications / amplifications) or decreases (deletions) in the copy number of DNA segments, typically longer than 1 kilobase pairs (kb), directly affecting the dosage effect of related genes. Homozygous regions are chromosomal regions with normal copy numbers (i.e., still retaining two alleles), where all allele sequences are identical, exhibiting a homozygous state. Specifically, when a ROH event occurs, if the copy number does not change (i.e., it remains diploid), it is essentially uniparental disomy (UPD), meaning that both pairs of homologous chromosomes in an individual originate from the same parent. These large genomic aberrations have been identified as key molecular bases for the occurrence, development, and prognosis of many common genetic diseases (such as autism spectrum disorder, intellectual disability, and epilepsy), congenital malformations, recurrent miscarriages, and various malignant tumors (such as leukemia and solid tumors).

[0003] For decades, karyotyping and fluorescence in situ hybridization (FISH) have been the primary methods for detecting large genomic aberrations, but their low throughput and limited resolution (typically >5-10 Mb) have made them insufficient for the needs of precision medicine. Subsequently, DNA microarray-based technologies, such as single nucleotide polymorphism arrays (SNP arrays) and comparative genomic hybridization arrays (aCGH), have become the standard methods for clinical detection of CNVs and ROHs due to their high throughput and relatively high resolution (up to 10-100 kb). However, these technologies have inherent limitations: First, their detection signals depend on the intensity of hybridization fluorescence, which is easily affected by experimental noise, GC content bias, and uneven probe design density, resulting in unstable data quality and requiring subjective intervention from experienced technicians to interpret the results; second, chip technology can only detect variations at preset probe sites and cannot detect novel variations outside the probe region, and its ability to accurately locate variation breakpoints is limited; finally, when detecting homozygous ROH (UPD) regions of the genome, although the chip platform can indirectly infer them through SNP genotyping information, it still requires family samples or matched normal control samples as a reference, which is difficult to obtain in many clinical scenarios.

[0004] With the rapid development of next-generation sequencing (NGS) technology and its exponentially decreasing cost, whole-genome sequencing (WGS) has gradually become the new gold standard for detecting genomic structural variations due to its unbiasedness, high resolution, and ability to provide sequence-level information. Currently, the main strategies for CNV detection using WGS include read depth-based methods, paired-end methods, split-read methods, and de novo assembly. However, to achieve sufficiently high sequencing depth for high sensitivity and specificity across the entire genome, the cost of WGS remains high, significantly limiting its widespread application in routine clinical diagnosis. As a compromise, whole-exome sequencing (WES) is widely used due to its lower cost, but its target regions only account for ~1-2% of the entire genome, and are mainly concentrated in coding regions. This non-uniform, non-genome-wide coverage pattern severely limits its ability to detect large CNVs and ROHs at the genome scale (especially non-coding regions), resulting in a huge detection blind spot.

[0005] To overcome the cost bottleneck of WGS and the coverage limitations of WES, targeted sequencing has been proposed as a cost-effective strategy. However, existing targeted sequencing methods mostly focus on specific pathogenic genes or known hotspot regions to detect small insertions / deletions (indels) or single nucleotide variants (SNVs), and their probe design principles are far removed from the needs of whole-genome structural variation detection. These methods face many challenges when analyzing CNVs / ROHs: the probe distribution on the genome is extremely uneven, with poor coverage of long genes and high-GC regions; the target regions (TRs) are not contiguous, posing significant difficulties for computational signal smoothing and region merging; most algorithms struggle to accurately distinguish between single-copy deletions and double-copy deletions, or triple-copy amplifications and higher fold amplifications; more importantly, they heavily rely on matched normal samples for standardization and background correction, which are often unavailable in many clinical scenarios such as prenatal diagnosis and unexplained developmental delays. Furthermore, there are very few reported targeted methods for detecting ROH (UPD), which usually require pedigree verification and lack the ability to make independent judgments. Summary of the Invention

[0006] There is an urgent need in the current technological field for a simultaneous whole-genome CNV / ROH / UPD detection scheme that can balance high detection performance, low cost, high throughput, and the need for matching control samples. This scheme requires a carefully designed, standardized target region set covering the entire genome, and a powerful, population-based bioinformatics system capable of automatically performing signal correction, variant identification, and integrated analysis. This invention is an innovative solution proposed based on this profound insight to address the aforementioned technical challenges.

[0007] To achieve the above-mentioned objectives, one aspect of the present invention provides a method for simultaneously detecting copy number variations and homozygous genomic regions, comprising the following steps:

[0008] Step S1: Construct a DNA probe set that covers the entire genome, is highly uniformly distributed, has stable sequence characteristics, and is rich in polymorphic information. The probe set maximizes the stability and representativeness of the signal and minimizes technical noise caused by GC content deviation, repetitive sequence interference, or uneven capture efficiency.

[0009] Step S2: The raw high-throughput sequencing data is filtered and converted into a standardized signal matrix for downstream analysis, including sequencing depth signals and heterozygosity index signals.

[0010] Step S3: Using a population geometric median normalization strategy, independent signal processing algorithms are applied to correct the original sequencing depth matrix and the original heterozygosity index matrix, respectively.

[0011] Step S4: Based on the statistical inference model, CNV and ROH events are identified respectively, and UPD is determined through logical integration;

[0012] Step S5: System performance verification.

[0013] Furthermore, step S1 includes the following steps:

[0014] Step S101: Select SNP sites with high heterozygosity from public databases to form a set. ;

[0015] Step S102, from set Extract the points that meet the comparison requirements to form a set. ;

[0016] Step S103, for the set After spatial distribution homogenization, the set is obtained. ;

[0017] Step S104, for the set Fill the gaps and determine the final probe position.

[0018] Furthermore, step S2 includes the following steps:

[0019] Step S201: The genome of the individual to be tested is broken down and repaired to construct the original paired-end sequencing library;

[0020] Step S202: Perform quality control and filtering on the raw sequencing data;

[0021] Step S203: Perform sequence alignment on the data processed in step S202 and sort it to generate a BAM file;

[0022] Step S204: Extract the raw sequencing depth and heterozygosity coefficient from the BAM file generated in step S203.

[0023] Furthermore, the sorted BAM files are traversed, and the total number of successfully aligned reads within each SeTR probe region is counted, denoted as . , where the subscript i represents the i-th probe region. Simultaneously, extract regions located within the probe that belong to the set. Allele count information for the SNP loci. For the j-th SNP locus, let the number of supporting reads for the non-reference allele (B-allele) be . Total coverage depth is The original B allele frequency is defined as: Thus, two original signal matrices are constructed: the original depth matrix: ,in =278,800;

[0024] Original heterozygosity matrix: ,in The number of high-quality SNP sites common to all samples.

[0025] Furthermore, step S3 includes the following steps:

[0026] Step S301: The sequencing depth signal is smoothed by moving average and then normalized by introducing the population geometric median.

[0027] Furthermore, to reduce the impact of fluctuations in local capture efficiency, the original depth of each sample was... Perform a moving average smoothing process. Define the smoothed depth value. for: , where window size =10, meaning the average value of the current probe and its 10 adjacent probes is taken. Further calculation of the global average depth ( The specific calculation is as follows: The depth coefficients were initially corrected. The specific calculation method is as follows: To eliminate systematic biases among different samples (such as library concentration and sequencing batch effects), a second normalization is performed using the geometric median of the population. Suppose K samples are involved in the analysis, then the geometric median Hᵢ of the i-th probe in the population is defined as: ,in This represents the depth coefficient of the k-th sample at the i-th probe. The final fully corrected depth ratio is obtained. The calculation method is as follows: Theoretically, this value should approach 1.0 in the region with a normal copy number (CN=2).

[0028] Step S302: Standardize the hybrid signal.

[0029] Furthermore, for each SNP locus j, its original heterozygosity index in the k-th sample is: Define its standardized heterozygosity coefficient. for: This transformation results in Q=0.5 for the completely heterozygous state (F=0.5) and Q=0 for the completely homozygous state (F=0 or 1). Similarly, the geometric median of this locus in the population is calculated. The calculation method is as follows: Final corrected heterozygosity coefficient Defined as: In the normal heterozygous state, In the ROH region, due to allele imbalance, .

[0030] Furthermore, step S4 includes the following steps:

[0031] Step S401: Applying a single-sample t-test model, using the population correction value as a reference, identify target regions where the depth ratio significantly deviates from 1.0, and infer the copy number status based on the direction and magnitude of the deviation, merging adjacent regions to form a CNV event; for each probe region i, its corrected depth ratio... In a population with normal copy numbers, the copy number follows a normal distribution. Now, for a given test sample, we test its... Does it deviate significantly from the group mean?

[0032] Construct a one-sample t-test statistic The calculation method is as follows:

[0033] in For this probe in the population The standard deviation of the value, in the denominator Used to compensate for population mean estimation errors. If If the corresponding p-value is less than 0.048, the probe is considered to have a significant copy number deviation.

[0034] Furthermore, at least 45 consecutive probes are divided into a potential CNV interval. At least three probes must have p-values ​​< 0.048, and the p-value of the fourth probe must not exceed 0.19 (i.e., four times the main threshold). The average value of the entire interval is then calculated. Calculate the final copy number: The `round` function represents rounding to the nearest integer. For example, if... =0.98, then CN = round(1.92) = 2; if =1.53, then CN = round(3.00) = 3.

[0035] Step S402: Based on the corrected homozygosity coefficient, apply the F-test model to detect whether its variance is significantly reduced relative to the population baseline, thus identifying ROH events; for any genomic region, if the heterozygosity coefficients of multiple SNP sites within it... A variance significantly lower than the population level suggests the occurrence of ROH. An F-test is used to compare the variance difference between the test sample and the population within this interval. Suppose a certain interval contains m SNP loci, and the variance of the test sample... The sample variance is V_test, and the population median is... If the sample variance is V_mul, then the F-statistic is defined as: All degrees of freedom If the p-value obtained from the F-test is less than 0.012, the null hypothesis is rejected, and the region is determined to have a ROH. The minimum span of a ROH event is three consecutive SNP sites. Expansion is allowed: as long as the p-value of the new probe is < 0.095, and the average p-value of the entire region is still below 0.095, they can be merged into a larger ROH region.

[0036] Step S403: Perform a logical intersection operation on the CNV and ROH results: If a region is simultaneously identified as a ROH event and its copy number is determined to be 2, then it is determined that the region has a single parent disomy event.

[0037] Furthermore, step S5 includes the following steps:

[0038] Step S501: Collect samples and construct a sample validation queue;

[0039] Step S502: Construct a comparative method and design experiments;

[0040] Step S503: Perform performance evaluation and analysis on the experimental results.

[0041] A second aspect of the present invention provides a system for simultaneously detecting copy number variations and homozygous genomic regions, comprising a selective target region set construction module, a sequencing data preprocessing and signal generation module, a population baseline correction and signal normalization module, a variant identification and integration analysis module, and a system validation module, wherein:

[0042] The selective target region set construction module is used to construct a set of DNA probes that cover the entire genome, are highly uniformly distributed, have stable sequence characteristics, and are rich in polymorphic information. The probe set maximizes the stability and representativeness of the signal and minimizes technical noise caused by GC content deviation, repetitive sequence interference, or uneven capture efficiency.

[0043] The sequencing data preprocessing and signal generation module performs quality control filtering on the raw NGS data (FASTQ format) obtained after target region enrichment (removing low-quality bases and adapter contamination), and aligns high-quality sequencing reads to the human reference genome, generating standard sequence alignment / mapping (SAM / BAM) files. Subsequently, the sequencing depth signal of each target region is calculated, and the allele frequencies of SNP sites located in the target region set are calculated to generate the raw sequencing depth matrix and the raw heterozygosity index matrix.

[0044] The population baseline correction and signal normalization module applies independent signal processing algorithms to correct the original sequencing depth matrix and the original heterozygosity index matrix, respectively. For the sequencing depth signal, a sliding window mean method is first used for smoothing to reduce local noise, and then normalization is performed using the geometric median of all samples in the target population at that target site to obtain the final corrected depth ratio. For the heterozygosity index, it is also corrected using the population geometric median to obtain the corrected heterozygosity coefficient.

[0045] The variant identification and integration analysis module uses a statistical inference model and a corrected depth ratio, applying a one-sample t-test model with the population correction value as a reference, to identify target regions where the depth ratio significantly deviates from 1.0. Based on the direction and magnitude of the deviation, it infers the copy number status and merges adjacent regions to form CNV events. Simultaneously, based on the corrected homozygosity coefficient, an F-test model is applied to detect whether its variance significantly decreases relative to the population baseline (i.e., tends towards 0) to identify ROH events. Finally, the CNV and ROH results are logically intersected: if a region is simultaneously identified as a ROH event and its copy number is determined to be 2, then the region is determined to have a single-parent disomy (UPD) event.

[0046] The system verification module is used to perform system performance verification.

[0047] A third aspect of the present invention provides an electronic device comprising a processor and a memory; the memory being used to store a program that, when executed by the processor, causes the processor to implement: a system and method for simultaneously detecting copy number variations and homozygous regions, for use in molecular diagnosis, prenatal screening, or preimplantation genetic testing of diseases, including hereditary developmental disorders, congenital malformations, recurrent miscarriages, intellectual disabilities, epilepsy, or tumors.

[0048] A fourth aspect of the present invention provides a computer-readable medium storing a computer program that, when executed by a processor, implements: a system and method for simultaneously detecting copy number variations and homozygous regions, for use in molecular diagnosis, prenatal screening, or preimplantation genetic testing of diseases, including hereditary developmental disorders, congenital malformations, recurrent miscarriages, intellectual disabilities, epilepsy, or tumors.

[0049] Compared with existing technologies, this system and method have the following advantages:

[0050] This invention innovatively integrates four core modules: (1) a selective target region set construction module; (2) a sequencing data preprocessing and signal generation module; (3) a population baseline correction and signal standardization module; and (4) a variant identification and integration analysis module. This overcomes the dependence of traditional CNV / ROH detection methods on matching normal control samples or family information, and for the first time achieves simultaneous detection of CNV, ROH, and UPD within the framework of targeted sequencing. This method significantly lowers the threshold for clinical application by replacing individual controls with population statistical modeling. Its designed selective target region set and innovative signal processing algorithm effectively overcome the signal noise problem caused by regional discontinuity and differences in capture efficiency in targeted sequencing, reducing the detection cost to less than 1 / 10 of whole-genome sequencing while ensuring high sensitivity and specificity. It possesses significant technological advancement, clinical applicability, and scalability. Attached Figure Description

[0051] Figure 1 A flowchart illustrating a method for simultaneously detecting copy number variations and homozygous regions of the genome.

[0052] Figure 2 ROC curves for validating the performance of cohorts representing homozygous regions of the real genome. Detailed Implementation

[0053] 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 some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] like Figure 1 The flowchart shown in Example 1 illustrates the method for simultaneously detecting copy number variations and homozygous genomic regions. By adjusting probe design parameters, sequencing depth, analysis module thresholds, and applicable populations, this method can be widely applied in clinical scenarios such as prenatal screening for various genetic diseases, neonatal disease diagnosis, genetic etiology analysis of infertility, risk assessment of recurrent miscarriage, and detection of genomic structural variations in tumor susceptibility. It is particularly suitable for complex genetic syndromes (such as Prader-Willi syndrome, Angelman syndrome, and Beckwith-Wiedemann syndrome) that require simultaneous detection of copy number abnormalities, homozygous genomic regions, and uniparental disomies (CNV / ROH / UPD), demonstrating high versatility and scalability.

[0055] This embodiment uses the simultaneous detection of copy number variations (CNV), homozygous regions (ROH), and unparental disomies (UPD) across the entire genome from 30 cell line samples with known karyotypes as an example to elaborate on the complete implementation steps of the "Three-in-One Structural Variation Detection System Based on Selective Target Region Set and Population Baseline Correction Model" (hereinafter referred to as CNROH-pipe) of this invention. All experiments were performed in a standard molecular biology laboratory and high-performance computing cluster environment to ensure the reproducibility and stability of data generation and analysis. Specifically, the following steps are included:

[0056] Step S1 involves constructing a DNA probe set that covers the entire genome, is highly uniformly distributed, has stable sequence characteristics, and is rich in polymorphic information, serving as the basis for subsequent targeted enrichment and sequencing. The design principle of this probe set aims to maximize signal stability and representativeness while minimizing technical noise caused by GC content bias, repetitive sequence interference, or uneven capture efficiency.

[0057] Step S101: Preliminary screening of candidate SNP sites.

[0058] First, information on all known single nucleotide polymorphisms (SNPs) on the human reference genome (GRCh37 / hg19 version) was obtained from the public databases released by the 1000 Genomes Project. SNPs with a population allele frequency (AF) between 0.12 and 0.88 were then selected and denoted as the set. The selection of this frequency range ensures that the selected loci have high heterozygosity in the population, which is beneficial for the subsequent identification of ROH events, while avoiding statistical bias that may be caused by rare alleles.

[0059] Step S102: Verification of sequence uniqueness and alignment reliability.

[0060] For each SNP site in set S1, 55 bases upstream and downstream of it were extracted to form a 110 bp local sequence fragment. These fragments were then aligned back to the complete GRCh37 reference genome using the BLASTN algorithm. A fragment was retained in the new set only if the best alignment result showed no mismatches and the mismatch rate of the second-best alignment result did not exceed 4.8%. This step effectively eliminates SNP sites located in repetitive sequence regions or with multiple highly homologous copies, ensuring that sequencing reads can be uniquely and accurately located.

[0061] Step S103: Spatial distribution homogenization processing.

[0062] Based on sets Sparsity processing is performed based on physical location. The ideal distance between adjacent sites is set to 10.6 kb. If multiple candidate SNPs exist within any consecutive 10 kb window, only the one closest to the window center is retained, and the rest are discarded. This process yields the set. Its members are more evenly distributed linearly on chromosomes, but there may still be large gaps.

[0063] Step S104: Fill the gap and determine the final probe position.

[0064] For sets In regions where the distance between adjacent SNPs exceeds 10.6 kb, additional target sites are inserted within these gaps to make the overall distribution more closely approximate one probe per 10.6 kb. These newly added sites do not necessarily have to be SNPs, but must meet the following conditions: (a) located in a non-coding region or intergenic region; (b) GC content within the range of 42%–58%; (c) not located in known simple repetitive sequences (such as microsatellites) or low-complexity regions; and (d) possessing good sequence alignment (mappability > 0.9). The final probe set is the "selective target region set," containing 278,800 independent probes with a total coverage length of approximately 41.8 Mb, representing 1.46% of the effective human genome (~2.86 Gb). The average length of each probe is 150 bp, and the median distance between adjacent probes is 10.6 kb.

[0065] This probe set can be synthesized using commercial liquid-phase hybridization capture technologies (such as Agilent SureSelect or RocheNimbleGen platforms) and used for subsequent targeted enrichment experiments.

[0066] Step S2 converts the raw high-throughput sequencing data into a standardized signal matrix that can be used for downstream analysis, including sequencing depth signals and heterozygosity index signals.

[0067] Step S201, Sample preparation and NGS library construction.

[0068] Genomic DNA (gDNA) from the individuals to be tested was randomly fragmented to an average fragment size of 200 bp using an ultrasonic disruptor. End repair, A-tailing, and adapter ligation were performed using the KAPA HyperPrep Kit to construct paired-end sequencing libraries. After PCR amplification, target regions were enriched using a customized selective target region set probe pool via liquid-phase hybridization. The enriched products were purified and paired-end sequencing (PE150 mode) was performed using the Illumina HiSeq X Ten platform to obtain raw FASTQ format data.

[0069] Step S202: Raw data quality control and filtering.

[0070] A comprehensive quality assessment of the raw sequencing data was performed using FastQC v0.12.0. Trimmomatic v0.40 was then used to perform adapter removal and trimming operations, with the specific parameter settings as follows:

[0071] ILLUMINACLIP:adapters.fa:2:30:10: Remove Illumina adapter sequence;

[0072] SLIDINGWINDOW:4:22: Reads with an average Phred quality below 22 within a sliding window (every 4 bp) are truncated; MINLEN:36: Reads ≥ 36 bp are retained. The filtered high-quality reads are used for the next alignment step.

[0073] Step S203: Sequence alignment and BAM file generation.

[0074] The filtered reads were aligned to the GRCh37 reference genome using BWA-MEM v0.7.17, with the following key parameters: -t 16: enable 16-thread parallel acceleration; -k 19: seed length at least 19 bp; -T 30: records with alignment scores below 30 are not output. The output SAM file was converted to compressed BAM format using SAMtools v1.17 and sorted by chromosome. Subsequently, Picard Tools was used to perform MarkDuplicates PCR to eliminate amplification bias.

[0075] Step S204: Extraction of raw sequencing depth and heterozygosity coefficient.

[0076] Traverse the sorted BAM files and count the total number of successfully aligned reads within each SeTR probe region, denoted as . , where the subscript i represents the i-th probe region. Simultaneously, extract regions located within the probe that belong to the set. Allele count information for the SNP loci. For the j-th SNP locus, let the number of supporting reads for the non-reference allele (B-allele) be . Total coverage depth is The original B allele frequency is defined as: Thus, two original signal matrices are constructed: the original depth matrix: ,in =278,800;

[0077] Original heterozygosity matrix: ,in The number of high-quality SNP sites common to all samples.

[0078] Step S3, by introducing the "population geometric median normalization" strategy, significantly improves the stability and cross-sample comparability of the signal, overcoming the limitation of traditional methods that rely on matched controls.

[0079] Step S301: Smoothing and normalization of sequencing depth signals.

[0080] To reduce the impact of fluctuations in local capture efficiency, the original depth of each sample was... Perform a moving average smoothing process. Define the smoothed depth value. for: , where window size =10, meaning the average value of the current probe and its 10 adjacent probes is taken. Further calculation of the global average depth ( The specific calculation is as follows: The depth coefficients were initially corrected. The specific calculation method is as follows: To eliminate systematic biases among different samples (such as library concentration and sequencing batch effects), a second normalization is performed using the geometric median of the population. Given K samples participating in the analysis, the geometric median of the i-th probe in the population is... Defined as: ,in This represents the depth coefficient of the k-th sample at the i-th probe. The final fully corrected depth ratio is obtained. The calculation method is as follows: Theoretically, this value should approach 1.0 in the region with a normal copy number (CN=2).

[0081] Step S302, standardization processing of hybrid signals.

[0082] For each SNP locus j, its original heterozygosity index in the k-th sample is: Define its standardized heterozygosity coefficient. for: This transformation results in Q=0.5 for the completely heterozygous state (F=0.5) and Q=0 for the completely homozygous state (F=0 or 1). Similarly, the geometric median of this locus in the population is calculated. The calculation method is as follows: Final corrected heterozygosity coefficient Defined as: In the normal heterozygous state, In the ROH region, due to allele imbalance, .

[0083] Step S4: Based on the statistical inference model, CNV and ROH events are identified respectively, and UPD is determined through logical integration.

[0084] Step S401, Copy number variation (CNV) detection.

[0085] For each probe region i, its correction depth ratio In a population with normal copy numbers, the copy number follows a normal distribution. Now, for a given test sample, we test its... Does it deviate significantly from the group mean?

[0086] Construct a one-sample t-test statistic The calculation method is as follows:

[0087] in For this probe in the population The standard deviation of the value, in the denominator Used to compensate for population mean estimation errors. If If the corresponding p-value is less than 0.048, the probe is considered to have a significant copy number deviation.

[0088] Furthermore, at least 45 consecutive probes are divided into a potential CNV interval. At least three probes must have p-values ​​< 0.048, and the p-value of the fourth probe must not exceed 0.19 (i.e., four times the main threshold). The average value of the entire interval is then calculated. Calculate the final copy number: The `round` function represents rounding to the nearest integer. For example, if... =0.98, then CN = round(1.92) = 2; if =1.53, then CN = round(3.00) = 3.

[0089] Step S402, detection of homozygous region (ROH).

[0090] For any given genomic region, if the heterozygosity coefficient of multiple SNP sites within it... A variance significantly lower than the population level suggests the occurrence of ROH. An F-test is used to compare the variance difference between the test sample and the population within this interval. Suppose a certain interval contains m SNP loci, and the variance of the test sample... The sample variance is V_test, and the population median is... If the sample variance is V_mul, then the F-statistic is defined as: All degrees of freedom If the p-value obtained from the F-test is less than 0.012, the null hypothesis is rejected, and the region is determined to have a ROH event. The minimum span of a ROH event is three consecutive SNP sites. Expansion is allowed: as long as the p-value of the new probe is < 0.095, and the average p-value of the entire region is still below 0.095, they can be merged into a larger ROH region.

[0091] Step S403, determination of single parent disomy (UPD).

[0092] The essence of UPD is a full-region ROH under the premise of normal copy number (CN=2). Therefore, this invention uses a logical intersection method to determine: if a certain genomic region simultaneously satisfies: (1) it is identified as a ROH event; (2) its average copy number CN = 2; then it is determined that the region has isodisomic UPD. This determination does not require pedigree information or parental samples and is completed independently based entirely on population statistical models.

[0093] Step S5: System performance verification.

[0094] To fully verify the technical advantages of this invention, "A detection method and system for simultaneously detecting copy number variations and homozygous genomic regions," particularly its innovation and advancement in uniparental disomy (UPD) detection, the applicant constructed a virtual validation cohort comprising 30 UPD syndrome patients. This cohort was designed strictly according to known human genomic imprinting regions, covering a variety of typical UPD-related diseases, aiming to systematically evaluate the performance differences between the method of this invention and a traditional UPD detection tool (named "UPDtool," whose principle relies on pedigree information and B allele frequency (BAF) analysis of matched normal control samples).

[0095] Step S501: Verify queue construction.

[0096] The 30-sample cohort constructed in this embodiment covers the four most common types of UPD syndrome, and the sample information statistics are detailed in Table 1. All samples were simulated to undergo targeted sequencing using the selective target region set (selective target region set) described in this invention, with an average sequencing depth of 75X. To simulate a real clinical scenario, no family pedigree samples from the parents or any matched normal control samples were provided for any samples; only the sequencing data of the individual being tested were provided.

[0097]

[0098] Table 1: Statistical Information of the UPD Syndrome Verification Cohort Samples

[0099] The construction of this cohort ensured that the genomic location, copy number status (all copy number neutral, CN=2), and clinical relevance of UPD events were consistent with those reported in the medical literature, providing a solid and reliable benchmark for performance evaluation.

[0100] Step S502, Comparison of methods and experimental design.

[0101] The method of this invention (CNROH-pipe) is strictly performed according to the process described in "Detailed Implementation". The core lies in using these 30 samples to construct an internal population baseline, identifying CNVs and ROHs through t-tests and F-tests respectively, and automatically inferring UPD events through logical integration. The entire process is completely independent of any external pedigree or matched control information.

[0102] Traditional method (UPDtool): This is a typical virtual tool that simulates existing commercial software (such as SNP-based analysis workflows). Its workflow requires input of sequencing data from the test sample and its parents (or at least one matched normal control). It identifies UPD by comparing the BAF deviation patterns of the test sample and the control sample. In this experiment, due to the unavailability of family pedigrees or matched controls, UPDtool was forced to run in "no control" mode, where the analysis was based solely on the BAF signal of the test sample itself, corrected using a pre-set static baseline from a public database. This mode has proven to have extremely low detection efficiency for copy number neutral events in practical applications.

[0103] Step S503, Performance evaluation results and analysis.

[0104] The experimental results clearly demonstrate the significant advantages of the method of this invention in UPD detection, as shown in the following comparison:

[0105] Sensitivity: CNROH-pipe successfully detected all 30 UPD events in 30 samples, achieving a sensitivity of 100%. All detected UPD regions (15q11-q13, 7q21-7q32, 11p15) perfectly matched their corresponding clinical syndromes. In contrast, due to a lack of necessary pedigree information, the "no control" mode of UPDtool performed extremely poorly, detecting UPD signals only in 7 paternal UPD15 (Angel's Syndrome) samples with abnormally strong BAF signals. It failed to detect maternal UPD15, UPD7, and UPD11p15 events, resulting in an overall sensitivity of only 23.3% (7 / 30). See Table 2 for details. Figure 2 The receiver operating characteristic (ROC) curves for the 30-case UPD validation cohort are shown. The results show that the area under the curve (AUC) of CNROH-pipe reaches 0.984 (95% confidence interval 0.977–0.984), indicating that it can achieve good discrimination performance at all classification thresholds.

[0106] Specificity and False Positive Control: CNROH-pipe detected 30 true UPD events without reporting any false positive UPD events in other regions of the genome. Through population baseline correction, it effectively reduced technical noise, ensuring the reliability of the ROH signal. While UPDtool also did not report false positives in undetected regions, this was mainly due to insufficient sensitivity rather than superior specificity. Of the 7 cases detected, one resulted in inaccurate boundary localization due to a significant difference between the baseline of the public database and the background of the sample population.

[0107] The ability to detect copy number neutral events: This is the most fundamental difference between the two methods. The method of this invention, through independent depth signal (for CNV) and hybridity signal (for ROH) analysis modules, and their integration, can clearly distinguish between two distinct situations: "copy number loss accompanied by ROH" and "normal copy number but ROH occurring (i.e., UPD). In contrast, UPDtool, lacking a control group, cannot effectively correct for individual background noise, and its BAF signal fluctuations are misjudged as technical errors, thus completely losing its ability to discriminate ROH. This is particularly evident in the detection rates of silver-Russell syndrome (UPD7) and Beaufort-Williams syndrome (UPD11p15) (CNROH-pipe: 14 / 14, UPDtool: 0 / 14).

[0108]

[0109] Table 2: Comparison of detection performance of the two methods in the UPD syndrome cohort

[0110] The performance verification results of this embodiment fully demonstrate that the CNROH-pipe method proposed in this invention, through its innovative "population baseline correction" and "CNV-ROH integrated analysis" strategies, successfully solves the problem of strong dependence on family and matched control samples in existing technologies for UPD detection. Even without providing any prior information, the method of this invention can still accurately identify various UPD syndromes with 100% sensitivity and high specificity, and its technical performance far exceeds that of traditional UPD detection tools. This advantage has significant practical implications for the common "proband-only" scenario in clinical diagnosis (i.e., only patient samples are available), providing a new, efficient, reliable, and cost-effective solution for the precise molecular diagnosis of hereditary diseases.

[0111] Example 2:

[0112] This invention provides a detection system for simultaneously detecting copy number variations and homozygous regions of the genome. The method includes the following five core modules:

[0113] The selective target region set construction module, based on preset criteria such as genome stability, moderate GC content, and polymorphism information content, uniformly selects tens of thousands of short DNA fragments as target probes from the human reference genome (e.g., hg19 / GRCh37). Each probe is approximately 120-160 bp in length, with the average physical distance between adjacent probes controlled at 9-12 kb. The probe set embeds high-frequency (e.g., population allele frequency in the range of 10%-90%) SNP sites. The target region set as a whole covers approximately 1.2%-1.6% of the entire genome.

[0114] The sequencing data preprocessing and signal generation module performs quality control filtering on the raw NGS data (FASTQ format) obtained after target region enrichment (removing low-quality bases and adapter contamination), and aligns high-quality sequencing reads to the human reference genome, generating standard sequence alignment / mapping (SAM / BAM) files. Subsequently, the sequencing depth signal of each target region is calculated, and the allele frequencies of SNP sites located in the target region set are calculated to generate the raw sequencing depth matrix and the raw heterozygosity index matrix.

[0115] The population baseline correction and signal normalization module applies independent signal processing algorithms to correct the original sequencing depth matrix and the original heterozygosity index matrix, respectively. For the sequencing depth signal, a sliding window mean method is first used for smoothing to reduce local noise, and then normalization is performed using the geometric median of all samples in the target population at that target site to obtain the final corrected depth ratio. For the heterozygosity index, it is also corrected using the population geometric median to obtain the corrected heterozygosity coefficient.

[0116] The variant identification and integration analysis module, based on the corrected depth ratio, applies a one-sample t-test model with the population correction value as a reference to identify target regions where the depth ratio significantly deviates from 1.0. It then infers the copy number status based on the direction and magnitude of the deviation, and merges adjacent regions to form CNV events. Simultaneously, based on the corrected homozygosity coefficient, an F-test model is applied to detect whether its variance significantly decreases relative to the population baseline (i.e., tends towards 0) to identify ROH events. Finally, the CNV and ROH results are logically intersected: if a region is simultaneously identified as a ROH event and its copy number is determined to be 2, then the region is determined to have a single-parent disomy (UPD) event.

[0117] The system verification module is used to perform system performance verification.

[0118] This invention constructs a highly integrated and cost-optimized three-in-one detection system for whole-genome structural variations (CNROH-pipe) for clinical genetic disease diagnosis and genomic medicine applications. Based on a carefully designed selective target region set (selective target region set), this system covers approximately 1.46% of key sites in the whole genome, reducing sequencing costs to less than one-tenth of whole-genome sequencing while ensuring signal representativeness. It innovatively introduces a "population baseline correction" statistical modeling strategy, completely eliminating the rigid dependence of traditional CNV / ROH detection on matched normal control samples, parental information, or pedigree data. For the first time within the targeted sequencing framework, it achieves simultaneous, independent, and high-confidence identification of copy number variations (CNV), homozygous regions (ROH), and uniparental disomies (UPD). Through multi-level signal processing algorithms such as sliding window smoothing, geometric median normalization, and combined single-sample t-test and F-test modeling, it significantly suppresses technical noise introduced by probe capture bias, GC content fluctuations, and PCR amplification bias, and can stably detect greater than 500 copies / times even with an average sequencing depth as low as 8×. This invention identifies pathogenic structural variations in the genome (kb) and further integrates the logical intersection rules of copy number state and homozygous state to accurately determine UPD, providing a novel technical approach for non-invasive prenatal screening and early neonatal diagnosis of imprinted diseases such as Prader-Willi, Angelman, and Beckwith-Wiedemann, as well as recessive genetic diseases. Furthermore, this invention is highly modular and customizable—by subsetting selective target regions and probe sets (e.g., compressed to 5 Mb), it can be flexibly adapted to different budgets and application scenarios. It can be used for large-scale population cohort screening as well as embedded in specialized clinical testing processes such as reproductive health, tumor susceptibility assessment, and recurrent miscarriage etiology analysis. In summary, this invention not only breaks through the cognitive boundary that targeted sequencing cannot detect whole-genome structural variations in terms of technical principles, but also sets a benchmark for next-generation high-throughput genetic disease genomic diagnostic tools in terms of clinical practicality, economy, and scalability. It is expected to replace traditional microarray technology and partially supplement the application gap of whole-genome sequencing, providing strong support for the genetic disease prevention and control system in the era of precision medicine.

[0119] Example 3:

[0120] This invention provides an electronic device for executing the aforementioned detection method. This device is specifically designed to implement the aforementioned simultaneous detection workflow for whole-genome CNV, ROH, and UPD based on a selective target region set. The electronic device includes at least one processor, a memory, an input / output interface, and a bus. The memory stores program instructions and data, specifically including: an operating system, a sequencing data management module, a bioinformatics analysis software package (i.e., the "CNROH-pipe" system), and a database. The "CNROH-pipe" system is the core of this invention, containing a complete codebase and implementing all the method steps described in the first aspect. Specifically, it covers: (i) a data import interface for receiving and parsing raw sequencing data (FASTQ or BAM format); (ii) a low-level computing engine for performing sequencing data quality control, alignment, and signal extraction; (iii) a signal processing core implementing key correction algorithms such as sliding window smoothing and geometric median normalization; (iv) a statistical analysis module integrating t-test and F-test models; and (v) a logical decision-making unit for CNV, ROH region merging, and UPD event inference. When the processor runs the "CNROH-pipe" system program stored in memory, it calls upon various hardware resources via the bus, receives sequencing data from the input / output interfaces, and automatically executes the complete analysis process from raw data to final variant reports, following the procedure described in the first aspect. This device can process single or batch samples, thus flexibly adapting to different research or clinical needs.

[0121] Example 4:

[0122] This invention provides a computer-readable medium storing a computer program. When executed by a processor, the computer program enables the processor to implement all the steps of the detection method for simultaneously detecting copy number variations and homozygous genomic regions as described in the first aspect. The computer program is essentially a collection of executable code or scripts for the "CNROH-pipe" system, with its functional modules corresponding one-to-one with the four core modules described in the first aspect. Specifically, the program includes: metadata code for defining a set of selective target regions (selective target region set), guiding the system to identify valid analysis regions; a set of instructions for performing sequencing data preprocessing and signal generation; a function library implementing population baseline correction and signal normalization algorithms, particularly optimization code for calculating the sliding window mean and geometric median; and a master control script for implementing variant identification and integration analysis, which coordinates the operation of subroutines such as t-test, F-test, region merging, and UPD determination. The computer-readable medium can be any form of non-volatile storage device, such as a hard disk drive, solid-state drive (SSD), USB flash drive, optical disc (CD / DVD), or cloud storage. By deploying this program, any general-purpose computing platform that meets the configuration requirements can be transformed into a dedicated CNV / ROH / UPD detection and analysis workstation, greatly enhancing the portability and application value of the method of this invention.

[0123] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for simultaneously detecting copy number variations and homozygous regions of the genome, characterized in that, Includes the following steps: Step S1: Construct a DNA probe set that covers the entire genome, is highly uniformly distributed, has stable sequence characteristics, and is rich in polymorphic information. The probe set maximizes the stability and representativeness of the signal and minimizes technical noise caused by GC content deviation, repetitive sequence interference, or uneven capture efficiency. Step S2 involves filtering the raw high-throughput sequencing data and converting it into a standardized signal matrix for downstream analysis, including sequencing depth signals and heterozygosity index signals. Step S3: Using a population geometric median normalization strategy, independent signal processing algorithms are applied to correct the original sequencing depth matrix and the original heterozygosity index matrix, respectively. Step S4: Based on the statistical inference model, CNV and ROH events are identified respectively, and UPD is determined through logical integration; Step S5: System performance verification.

2. The method for simultaneously detecting copy number variations and homozygous genomic regions according to claim 1, characterized in that, Step S1 includes the following steps: Step S101: Select SNP sites with high heterozygosity from public databases to form a set. ; Step S102, from set Extract the points that meet the comparison requirements to form a set. ; Step S103, for the set After spatial distribution homogenization, the set is obtained. ; Step S104, for the set Fill the gaps and determine the final probe position.

3. The method for simultaneously detecting copy number variations and homozygous regions of the genome according to claim 1, characterized in that, Step S2 includes the following steps: Step S201: The genome of the individual to be tested is broken down and repaired to construct the original paired-end sequencing library; Step S202: Perform quality control and filtering on the raw sequencing data; Step S203: Perform sequence alignment on the data processed in step S202 and sort it to generate a BAM file; Step S204: Extract the raw sequencing depth and heterozygosity coefficient from the BAM file generated in step S203.

4. The method for simultaneously detecting copy number variations and homozygous genomic regions according to claim 1, characterized in that, Step S3 includes the following steps: Step S301: The sequencing depth signal is smoothed by moving average and then normalized by introducing the population geometric median. Step S302: Standardize the hybrid signal.

5. The method for simultaneously detecting copy number variations and homozygous genomic regions according to claim 4, characterized in that, The step S301, which introduces the geometric median of the population for quadratic normalization, is as follows: If K samples are involved in the analysis, then the geometric median of the i-th probe in the population is... Defined as: ,in The depth coefficient of the k-th sample at the i-th probe is then the fully corrected depth ratio. The calculation method is as follows: Theoretically, this value should approach 1.0 in the normal copy number region.

6. The method for simultaneously detecting copy number variations and homozygous genomic regions according to claim 1, characterized in that, Step S4 includes the following steps: Step S401: Apply the single-sample t-test model, with the population correction value as a reference, to identify target regions where the depth ratio deviates significantly from 1.0, and infer the copy number status based on the direction and magnitude of the deviation, and merge adjacent regions to form CNV events; Step S402: Based on the corrected homozygosity coefficient, apply the F-test model to detect whether its variance is significantly reduced relative to the population baseline, and identify the ROH event. Step S403: Perform a logical intersection operation on the CNV and ROH results: If a region is simultaneously identified as a ROH event and its copy number is determined to be 2, then it is determined that the region has a single-parent disomy event.

7. The method for simultaneously detecting copy number variations and homozygous regions of the genome according to claim 1, characterized in that, Step S5 includes the following steps: Step S501: Collect samples and construct a sample validation queue; Step S502: Construct a comparative method and design experiments; Step S503: Perform performance evaluation and analysis on the experimental results.

8. A system for simultaneously detecting copy number variations and homozygous genomic regions, characterized in that, It includes a selective target region set construction module, a sequencing data preprocessing and signal generation module, a population baseline correction and signal normalization module, a variant identification and ensemble analysis module, and a system validation module, among which: The selective target region set construction module is used to construct a set of DNA probes that cover the entire genome, are highly uniformly distributed, have stable sequence characteristics, and are rich in polymorphic information. The probe set maximizes the stability and representativeness of the signal and minimizes technical noise caused by GC content deviation, repetitive sequence interference, or uneven capture efficiency. The sequencing data preprocessing and signal generation module is used to filter and process the raw high-throughput sequencing data and transform it into a standardized signal matrix for downstream analysis, including sequencing depth signals and heterozygosity index signals. The population baseline correction and signal normalization module applies independent signal processing algorithms to correct the original sequencing depth matrix and the original heterozygosity index matrix through a population geometric median normalization strategy. The variant identification and integration analysis module is used to identify CNV and ROH events based on a statistical inference model, and to determine UPD through logical integration. The system verification module is used to perform system performance verification.

9. An electronic device, characterized in that, It includes a processor and a memory; the memory is used to store a program that, when executed by the processor, causes the processor to perform: the system and method for simultaneously detecting copy number variations and homozygous regions as described in any one of claims 1 to 8, for use in molecular diagnosis, prenatal screening or preimplantation genetic testing of diseases, including hereditary developmental disorders, congenital malformations, recurrent miscarriages, intellectual disabilities, epilepsy or tumors.

10. A computer-readable medium, characterized in that, The computer-readable medium stores a computer program that, when executed by a processor, implements: a system and method for simultaneously detecting copy number variations and homozygous regions as described in any one of claims 1 to 8, for use in molecular diagnosis, prenatal screening, or preimplantation genetic testing of diseases, including hereditary developmental disorders, congenital malformations, recurrent miscarriages, intellectual disabilities, epilepsy, or tumors.

Images