A CNV detection method, system and reliability verification method based on small panel

Abstract

The present invention belongs to the technical field of CNV detection, and discloses a CNV detection method, system and reliability verification method based on a small panel. The detection method comprises: obtaining an original data matrix, wherein one column in the original data matrix is ​​the depth data of the sample to be tested, and the remaining columns are the depth data of the control sample; calculating the ratio between each depth data in each row of the original data matrix and the geometric mean of the row to obtain an intermediate data matrix; calculating the ratio between each depth data in each column of the intermediate data matrix and the median of the column to obtain a correction data matrix; solving the ratio between the sample to be tested and the control sample in each amplification region based on the correction data matrix to obtain the CNV type of the sample to be tested. Thereby effectively improving the accuracy of the detection result. The reliability verification method is based on mathematical logic, which is simpler and more feasible than the existing method based on a large number of simulation calculations, and the reliability of the result is better.

Description

Technical Field

[0001] The present invention relates to the technical field of CNV detection, and in particular to a CNV detection method, system and reliability verification method based on a small panel. Background Art

[0002] Copy Number Variation (CNV) is a gene mutation with a length ranging from 1kb to 3Mb, caused by genome rearrangement. It is widely used in the detection of genetic diseases and auxiliary diagnosis of some common diseases.

[0003] Targeted sequencing technology is a genetic testing method that enriches specific genomic regions related to certain genetic diseases for sequencing. It has the advantages of low output of single sample sequencing data and fast analysis speed. At the same time, because it can perform deep sequencing on the target genomic region, it also has the advantages of high detection sensitivity and accuracy. Small panel refers to a sequencing sample that contains only dozens of genes. Therefore, compared with CNV detection based on the whole genome, CNV detection based on small panels and combined with targeted sequencing technology not only has stable test results and low cost, but also can meet the testing needs of most people.

[0004] The existing CNV detection principle is to compare the coverage changes between the test sample and the control sample to obtain the inconsistent interval between the two, and then detect the potential copy number variation area. Based on the ratio between the test sample and the control sample in each amplification region, CNV is divided into heterozygous deletion, homozygous deletion and duplication. For heterozygous deletion, the ratio is 0.5, indicating that the copy number of the test sample is reduced compared with the control sample. For homozygous deletion, the ratio is 0, indicating that the copy number of the test sample is lost compared with the control sample; for duplication, the ratio is 1.5, indicating that the copy number is repeated compared with the control sample.

[0005] In the process of testing based on the above CNV detection principle, data correction is an important factor in determining the accuracy of the test results. However, when the targeted sequencing area based on a small panel is small, and the area where CNV occurs is large (that is, the area where CNV occurs occupies too large a proportion of the targeted sequencing area), the results detected by the conventional CNV data correction method will be abnormal. For example, when a larger CNV occurs in the middle of a sample, since the conventional CNV data correction method performs correction based on the average depth of the sample, the depth of the larger CNV area will lower the average depth. This may cause false positives on the fragments on both sides where CNV does not actually occur; it may also cause missed detection of the middle fragment where CNV actually occurs. Summary of the invention

[0006] The purpose of the present invention is to provide a CNV detection method, system and reliability verification method based on a small panel, so as to improve the technical problems of errors and missed detections that are prone to occur in the existing CNV detection based on a small panel.

[0007] To achieve the above object, the present invention proposes the following technical solutions:

[0008] A CNV detection method based on a small panel, comprising:

[0009] Obtaining an original data matrix, wherein one column of the original data matrix is ​​depth data of the sample to be tested, and the remaining columns are depth data of the control sample;

[0010] Obtaining a geometric mean of each depth data in each row of the original data matrix, and calculating a first ratio between each depth data in each row of the original data matrix and the geometric mean of the row, and forming an intermediate data matrix according to the positions of the depth data corresponding to the first ratio;

[0011] Obtaining the median or mean of each depth data in each column of the intermediate data matrix, and calculating a second ratio between each depth data in each column of the intermediate data matrix and the median or mean of the column, and forming a correction data matrix according to the position of the depth data corresponding to the second ratio;

[0012] The ratio between the sample to be tested and the control sample in each amplified region is solved based on the calibration data matrix to obtain the CNV type of the sample to be tested.

[0013] Furthermore, the first column of the original data matrix is ​​the depth data of the sample to be tested.

[0014] Furthermore, the calculation formula of the geometric mean of each row in the original data matrix is:

[0015]

[0016] in:

[0017] i=1, 2...n, j=1, 2...m; n is the total number of rows of the original data matrix, and m is the total number of columns of the original data matrix;

[0018] a ij is the element in the i-th row and j-th column in the original data matrix.

[0019] Further, obtaining the CNV type of the sample to be tested includes:

[0020] Calculate the average or median of the depth data of each control sample in each row of the calibration data matrix to obtain a control average or a control median;

[0021] Solving the ratio between the depth data of the sample to be tested in each row of the calibration data matrix and the control mean or control median of the row;

[0022] The CNV type of the sample to be tested is determined according to the ratio.

[0023] A CNV detection system based on a small panel, comprising:

[0024] An acquisition module is used to acquire an original data matrix, wherein one column of the original data matrix is ​​depth data of the sample to be tested, and the remaining columns are depth data of the control sample;

[0025] A first calculation module is used to obtain a geometric mean of each depth data in each row of the original data matrix, and calculate a first ratio between each depth data in each row of the original data matrix and the geometric mean of the row, and form an intermediate data matrix according to the first ratio and the position of the depth data corresponding thereto;

[0026] a second calculation module, used for obtaining the median or mean of each depth data in each column of the intermediate data matrix, and calculating a second ratio between each depth data in each column of the intermediate data matrix and the median or mean of the column, and forming a correction data matrix according to the position of the depth data corresponding to the second ratio;

[0027] The CNV detection module is used to solve the ratio between the sample to be tested and the control sample in each amplification region based on the correction data matrix to obtain the CNV type of the sample to be tested.

[0028] A method for verifying the reliability of CNV detection based on a small panel, used to verify the reliability of the detection method, comprising:

[0029] Assume an m×(n+1) dimensional original data matrix:

[0030] Among them, the first column is the sample to be tested, and the remaining columns are control samples;

[0031] For the sample to be tested, it is assumed that CNV occurs on the kth to k+lth genes; where τ (i) is the actual coefficient of variation of the ith gene that has mutated; the ideal coefficient of variation of the gene is recorded as τ, and the first constraint is established: |τ (i) -τ|<δ;

[0032] Wherein, δ is the difference threshold between the actual coefficient of variation of the i-th gene and the ideal coefficient of variation;

[0033] For the control samples, is the element in the j+1th column and the ith row; is the actual ratio of the element in the control sample to the corresponding element in the sample to be tested; the ideal ratio of the control sample in the j+1th column to the sample to be tested is recorded as b j , and create the second constraint:

[0034] Among them, ε j is the difference threshold between the actual ratio in the j+1th column and the i-th row and the ideal ratio in the j+1th column;

[0035] Get the geometric mean of each row element in the original data matrix:

[0036] g i =a i med i ;

[0037] For regions without CNV, that is, i<k or i>k+l: calculate the geometric mean of the actual ratios described in row i Calculate the geometric mean of the ideal ratios in row i And establish the third constraint: |med i -med|≤θ;

[0038] Wherein, θ is the difference threshold between the geometric mean of the actual ratios described in the i-th row and the geometric mean of the corresponding ideal ratios; θ=max{θ1,θ2}; Wherein, Among them, Ο() is an order function, ε=max{ε1,…,ε j ,…,ε n};

[0039] For the CNV region, that is, k≤i≤k+l; calculate the geometric mean of the actual ratios described in the i-th row Calculate the geometric mean of the ideal ratios in row i And establish the fourth constraint: |med i -med|≤Θ;

[0040] Where Θ is the difference threshold between the geometric mean of the actual ratios in the i-th row and the geometric mean of the corresponding ideal ratios; Θ = max{Θ1, Θ2}; Where,

[0041] Calculate a first ratio between each element in each row of the original data matrix M and the geometric mean value of the row, and form an intermediate data matrix M1 according to the position of the corresponding depth data using the first ratio;

[0042] Get the median of each element in each column of the intermediate data matrix M1:

[0043] For the control sample, the median μ of the j+1th column is calculated based on the third constraint and the fourth constraint. j , we can get:

[0044] in, β = max{θ, Θ};

[0045] For the sample to be tested, the median μ0 is calculated based on the third constraint and the fourth constraint. If it does not exist, This indicates that the reliability of the method described is low; otherwise, continue to calculate the second ratio between each depth data in each column of the intermediate data matrix and the median or average of the column, and form a correction data matrix M2 according to the position of the corresponding depth data using the second ratio;

[0046] The third ratio between the first column and the j+1th column in the correction matrix M2 is calculated to determine the CNV type:

[0047] For regions without CNV, that is, i<k or i>k+l: if there is Then there exists R1≈1; wherein R1 is the third ratio in the CNV-free region;

[0048] For CNV regions, that is, k≤i≤k+l: if there is Then there exists R2≈τ; wherein R2 is the third ratio in the CNV region;

[0049] This means that the verified CNV region and type are consistent with the assumed situation, that is, the reliability of the method described is high; otherwise, it means that the reliability of the method described is poor.

[0050] Furthermore, the process of establishing the third constraint includes:

[0051] Establish the first inequality:

[0052] Establish the second inequality:

[0053] The third constraint is established based on the second constraint, the first inequality, and the second inequality.

[0054] Furthermore, the process of establishing the fourth constraint includes:

[0055] Establish the third inequality:

[0056]

[0057] Establish the fourth inequality:

[0058]

[0059] The fourth constraint is established based on the second constraint, the third inequality, and the fourth inequality.

[0060] Furthermore, for the intermediate data matrix M1:

[0061] For the region without CNV, i<k or i>k+1, based on the third constraint, we can get:

[0062]

[0063] For CNV regions, k≤i≤k+1, based on the fourth constraint, we can obtain:

[0064]

[0065] Beneficial effects:

[0066] It can be seen from the above technical solutions that the technical solution of the present invention provides a CNV detection method based on a small panel to improve the detection anomalies caused by data correction in the existing CNV detection method.

[0067] The detection method includes: obtaining an original data matrix, wherein one column of the original data matrix is ​​depth data of the sample to be tested, and the remaining columns are depth data of the control sample; calculating the ratio between each depth data in each row of the original data matrix and the geometric mean of the row to obtain an intermediate data matrix; calculating the ratio between each depth data in each column of the intermediate data matrix and the median of the column to obtain a correction data matrix; and solving the ratio between the sample to be tested and the control sample in each amplification region based on the correction data matrix to obtain the CNV type of the sample to be tested.

[0068] In the existing detection process, when data correction is performed, the average or median of the depth data of each column in the original data matrix is ​​directly solved, and then the original data matrix is ​​processed by column based on the average or median to obtain the correction data matrix. Since each column in the original data matrix belongs to the sequencing depth of different gene regions in the same sample, that is, the sample to be tested or the control sample; and processing it by column at this time weakens the difference in the sequencing depth of genes in different regions in the original sample to be tested or the control sample. And CNV detection is based on the difference between the sample to be tested and the control sample in each amplification region, so the existing data correction is bound to lead to result errors. In the present technical solution, the depth data is first processed by row to obtain an intermediate data matrix; and the elements in each row are the sequencing depth data of the same position in the sample to be tested or the control sample, so the difference between different amplification regions is guaranteed. Secondly, the intermediate data matrix is ​​processed based on the median of each column in the intermediate data matrix, and the median further guarantees the difference between each amplification region compared with the method based on the average processing in the existing correction. Therefore, when CNV detection is performed based on the correction data matrix in this technical solution, the accuracy of the detection results is effectively improved and the false detection situation is improved. At the same time, compared with the average correction method, this technical solution does not have the situation where the ratio does not meet the threshold and is considered a normal copy; therefore, missed detection is also effectively avoided.

[0069] The present technical solution also provides a CNV detection reliability verification method based on a small panel. The existing detection reliability verification process is mostly achieved through a large number of detection processes, which is not only cumbersome and complicated, but also time-consuming. In particular, when performing detection verification, the type of verification sample has a greater impact on whether the verification itself is reliable, and artificial sample design or selection often cannot comprehensively consider the changeable and complex CNV. As a result, the reliability of the verification result is questionable. The present technical solution provides a reliability verification method based on numerical calculations, which gets rid of the dependence on repeated detection and verifies the reliability of the CNV detection based on the rationality of numerical logic. Not only is the verification process simple and convenient, but the verification results are more accurate and reliable. It is effectively proved that the test results obtained by the above-mentioned CNV detection are always more consistent with the actual situation.

[0070] It should be appreciated that all combinations of the foregoing concepts, as well as additional concepts described in greater detail below, may be considered to be part of the inventive subject matter of the present disclosure, provided such concepts are not mutually inconsistent.

[0071] The foregoing and other aspects, embodiments and features of the present invention can be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the present invention, such as the features and / or beneficial effects of the exemplary embodiments, will be apparent from the following description or learned from the practice of the specific embodiments according to the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] The drawings are not intended to be drawn to scale. In the drawings, each identical or nearly identical component shown in various figures may be represented by the same reference numeral. For clarity, not every component is labeled in every figure. Embodiments of various aspects of the present invention will now be described by way of example and with reference to the accompanying drawings, in which:

[0073] Figure 1 is a flow chart of the CNV detection method based on a small panel described in this embodiment;

[0074] Figure 2 for Figure 1 Flow chart for obtaining CNV types;

[0075] Figure 3 This is a CNV detection result diagram when the average correction method in the prior art is used;

[0076] Figure 4 This is a graph showing the CNV detection results using the detection method described in this embodiment. DETAILED DESCRIPTION

[0077] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Unless otherwise defined, the technical terms or scientific terms used herein should be the common meanings understood by people with ordinary skills in the field to which the present invention belongs.

[0078] The words "first", "second" and similar words used in the patent application specification and claims of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, unless the context clearly indicates otherwise, the singular form of "a", "an" or "the" and other similar words do not indicate a quantitative limitation, but indicate the existence of at least one. "Include" or "comprise" and other similar words mean that the elements or objects appearing before "include" or "comprise" include the features, wholes, steps, operations, elements and / or components listed after "include" or "comprise", and do not exclude the existence or addition of one or more other features, wholes, steps, operations, elements, components and / or their collections. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0079] In the existing CNV detection process, the test results are often abnormal due to the influence of the correction method. The reliability test process based on a large number of verification samples is not only cumbersome and complicated, but also cannot always effectively verify the complex and changeable CNV.

[0080] Data verification in CNV detection in the prior art is performed through the following steps:

[0081] Step S001: Obtaining the original data matrix

[0082] Step S003, calculating the geometric mean of the original data matrix by column, and obtaining:

[0083]

[0084] Step S005: Calculate the correction coefficient to obtain:

[0085] Step S007: The corrected matrix is:

[0086]

[0087] It can be seen from the above steps 001 to 007 that in the existing detection process, when performing data correction, the average or median of the depth data of each column in the original data matrix is ​​directly solved, and then the original data matrix is ​​processed column by column based on the average or median to obtain the correction data matrix. Since each column in the original data matrix belongs to the sequencing depth of different gene regions in the same sample, that is, the test sample or the control sample; processing it by column at this time weakens the difference in the sequencing depth of genes in different regions in the original test sample or the control sample. CNV detection is based on the difference between the test sample and the control sample in each amplification area, so the existing data correction is bound to lead to result errors.

[0088] Example 1

[0089] In order to improve the defects of the above existing correction methods in actual use, the CNV detection method described in this embodiment is as follows Figure 1 As shown, the specific steps are as follows:

[0090] Step S102, obtaining an original data matrix, wherein one column of the original data matrix is ​​depth data of the sample to be tested, and the remaining columns are depth data of the control sample.

[0091] As a specific implementation, in order to facilitate calculation, the original data matrix is ​​set so that the first column is the depth data of the sample to be tested, and the remaining columns are the depth data of the control samples.

[0092] The original data matrix in this embodiment is:

[0093] As a specific implementation, the acquisition of the original data matrix includes:

[0094] Step S100: constructing a genome alignment and correction file of the sample.

[0095] In specific implementation, this step specifically includes: first, basic quality control of sample sequencing data is performed, and data quality control includes qualified data volume, average sequencing depth reaching requirements, and data quality Q20>90%, Q30>85%.

[0096] Q20 and Q30: Each base in the sequencing data has a corresponding quality value. If the quality value is Q20, the probability of misidentification is 1%, that is, the error rate is 1%, or the accuracy is 99%; if the quality value is Q30, the probability of misidentification is 0.1%, that is, the error rate is 0.1%, or the accuracy is 99.9%;

[0097] Then use the bwa module of sentieon software (NGS gene data analysis acceleration software) to compare with the reference genome to obtain the comparison data, and then use the UMI processing module in sentieon software to perform subsequent processing on the compared bam file to obtain the final bam file, that is, the comparison data.

[0098] Step S101, construct a matrix file, and calculate the depth of each amplification area according to the designed amplification area.

[0099] In specific implementation, this step includes: according to the designed amplification region coordinate information, using the R package Rsamtools to calculate the number of sequences covering the amplification region, constructing the number of covering sequences for each amplification region of the sample, and summarizing them into a matrix file.

[0100] The sample to be tested and the historical negative sample (control sample) are calculated using the above calculation method to calculate the control sample matrix file. The sequencing sequence number file of the test sample and the file of the control sample are merged according to the same amplification region position to obtain a matrix file. Each column in the matrix file represents a sample, and each row represents the sequencing sequence number file of an amplification region.

[0101] Step S104, obtaining the geometric mean of each depth data in each row of the original data, and calculating a first ratio between each depth data in each row of the original data matrix and the geometric mean of the row, and forming an intermediate data matrix according to the position of the corresponding depth data using the first ratio.

[0102] In this embodiment, the calculation formula for the geometric mean of each row in the original data matrix is:

[0103]

[0104] in:

[0105] i=1, 2...n, j=1, 2...m; n is the total number of rows of the original data matrix, and m is the total number of columns of the original data matrix;

[0106] a ij is the element in the i-th row and j-th column in the original data matrix.

[0107] Then the intermediate data matrix is ​​obtained as:

[0108] Step S106, obtaining the median or mean of each depth data in each column of the intermediate data matrix, and calculating a second ratio between each depth data in each column of the intermediate data matrix and the median or mean of the column, and forming a correction data matrix according to the position of the corresponding depth data using the second ratio.

[0109] In this embodiment, the median is used for calculation; the median of each column is recorded as: med 01 ,med 02 ,…,med 0m .

[0110] Then the correction data matrix is ​​obtained:

[0111]

[0112] Step S108: Calculate the ratio between the sample to be tested and the control sample in each amplification region based on the calibration data matrix to obtain the CNV type of the sample to be tested.

[0113] As a specific implementation method, Figure 2 As shown, the step S108 includes:

[0114] Step S108.2, calculating the mean or median of the depth data of each control sample in each row of the calibration data matrix to obtain a control mean or a control median.

[0115] In this embodiment, the control mean is specifically selected to be obtained.

[0116] Step S108.4, calculating the ratio between the depth data of the sample to be tested in each row of the calibration data matrix and the control mean or control median of the row.

[0117] Step S108.6: Determine the CNV type of the sample to be tested according to the ratio.

[0118] As can be seen from the above steps, this embodiment first processes the depth data by row to obtain an intermediate data matrix; and each row is the sequencing depth data of the same position in the sample to be tested or the control sample, thereby ensuring the differences between different amplification regions. Secondly, the intermediate data matrix is ​​processed based on the median of each column in the intermediate data matrix, and the median further ensures the differences between the amplification regions compared with the method based on the average processing in the existing correction. Therefore, when performing CNV detection based on the correction data matrix in the technical solution, the accuracy of the detection result is effectively improved. At the same time, compared with the average correction method, the technical solution does not have a situation where the ratio does not meet the threshold and is considered to be a normal copy; therefore, it also effectively avoids missed detection.

[0119] In the specific implementation, for three samples to be tested, the CNVs actually occurred in the middle region, and the CNV fragments were 10 exon, 30 exon, and 40 exon, respectively, using the existing average correction method and the detection method described in this embodiment to perform CNV detection. Figure 3 As shown in the figure, for the existing average correction method, as the CNV area increases, the ratio of the left and right sides and the actual CNV area increases, which leads to false positives in the detection results of the normal areas on both sides as the actual CNV area increases, while the detection results of the actual CNV area in the middle are missed because the ratio does not meet the threshold. Figure 4 As shown, for the detection method described in this embodiment, as the CNV region increases, both the actual CNV region and the normal region have better detection results, which further indicates that this embodiment has better detection accuracy.

[0120] The above method process can be run in a processor, or can also be stored in a memory (or computer-readable medium), which includes permanent and non-permanent, removable and non-removable media. Information storage can be achieved by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0121] These computer programs can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps of the functions specified in one or more blocks can be implemented by different modules corresponding to different steps.

[0122] Example 2

[0123] This embodiment also discloses a CNV detection system based on a small panel, and the system is built based on the detection method, and includes: an acquisition module, a first calculation module, a second calculation module and a CNV detection module.

[0124] The acquisition module is used to acquire an original data matrix, in which one column is depth data of the sample to be tested, and the remaining columns are depth data of the control sample.

[0125] The first calculation module is used to obtain the geometric mean of each depth data in each row of the original data, and calculate the first ratio between each depth data in each row of the original data matrix and the geometric mean of the row, and form an intermediate data matrix according to the position of the corresponding depth data using the first ratio.

[0126] The second calculation module is used to obtain the median or mean of each depth data in each column of the intermediate data matrix, and calculate the second ratio between each depth data in each column of the intermediate data matrix and the median or mean of the column, and form a correction data matrix according to the position of the corresponding depth data using the second ratio.

[0127] The CNV detection module is used to solve the ratio between the sample to be tested and the control sample in each amplification region based on the calibration data matrix to obtain the CNV type of the sample to be tested.

[0128] It can be seen that when CNV detection is performed based on the detection system, it also has the advantage of high accuracy of detection results; thereby greatly improving the false detection existing in the existing detection system and avoiding the phenomenon of missed detection.

[0129] Example 3

[0130] The existing test reliability verification process is mostly achieved through a large number of testing processes, which is not only cumbersome and complicated, but also time-consuming. In particular, when conducting test verification, the type of verification sample affects the reliability of the verification itself, and artificial sample design or selection often cannot take into account the variable and complex CNVs. As a result, the reliability of the verification results is questionable.

[0131] The reliability verification method of the present application is based on mathematical logic to verify the CNV detection method disclosed in the above embodiment 1. It includes:

[0132] First, set an m×(n+1)-dimensional original data matrix:

[0133] Among them, the first column is the sample to be tested, and the remaining columns are control samples;

[0134] For the sample to be tested, it is assumed that CNV occurs on the kth to k+lth genes; where τ (i) is the actual coefficient of variation of the ith gene that has mutated; the ideal coefficient of variation of the gene is recorded as τ, and the first constraint is established: |τ (i) -τ|<δ;

[0135] Wherein, δ is the difference threshold between the actual coefficient of variation of the ith gene and the ideal coefficient of variation.

[0136] For the control samples, is the element in the j+1th column and the ith row; is the actual ratio of the element in the control sample to the corresponding element in the sample to be tested; the ideal ratio of the control sample in the j+1th column to the sample to be tested is recorded as b j , and create the second constraint:

[0137] Among them, ε j is the difference threshold between the actual ratio in the j+1th column and the i-th row and the ideal ratio in the j+1th column.

[0138] Secondly, obtain the geometric mean of each row element in the original data matrix:

[0139] g i =a i med i ;

[0140] For regions without CNV, i < k or i > k + l; It is the geometric mean of the actual ratios described in the i-th row; the geometric mean of the ideal ratios described in the i-th row is And based on the second constraint, establish the third constraint: |med i -med|≤θ;

[0141] Wherein, θ is the difference threshold between the geometric mean of the actual ratios described in the i-th row and the geometric mean of the corresponding ideal ratios; θ=max{θ1,θ2}; Wherein, Among them, Ο() is an order function, ε=max{ε1,…,ε j ,…,ε n}.

[0142] In this embodiment, the process of establishing the third constraint includes:

[0143] First, establish the first inequality:

[0144]

[0145] Next, establish the second inequality:

[0146]

[0147] Finally, the third constraint is established based on the second constraint, the first inequality and the second inequality.

[0148] For the CNV region, that is, k≤i≤k+l; It is the geometric mean of the actual ratios described in the i-th row; the geometric mean of the ideal ratios described in the i-th row is And based on the second constraint, the fourth constraint is established: |med i -med|≤Θ;

[0149] Where Θ is the difference threshold between the geometric mean of the actual ratios in the i-th row and the geometric mean of the corresponding ideal ratios; Θ = max{Θ1, Θ2}; Where,

[0150] In this embodiment, the process of establishing the fourth constraint includes:

[0151] First, establish the third inequality:

[0152]

[0153] Next, establish the fourth inequality:

[0154]

[0155] Finally, the fourth constraint is established based on the second constraint, the third inequality and the fourth inequality.

[0156] A first ratio between each element in each row of the original data matrix M and the geometric mean value of the row is calculated, and an intermediate data matrix M1 is formed according to the first ratio and the position of the corresponding depth data.

[0157] In this embodiment, for the intermediate data matrix M1:

[0158] For the region without CNV, i<k or i>k+1, based on the third constraint, we can get:

[0159]

[0160] For CNV regions, k≤i≤k+1, based on the fourth constraint, we can obtain:

[0161]

[0162] Furthermore, the median of each element in each column of the intermediate data matrix M1 is obtained:

[0163] For the control sample, the median of the j+1th column is recorded as μ j , based on the third constraint and the fourth constraint, we can get:

[0164] in, β = max{θ, Θ}.

[0165] For the sample to be tested, the median is recorded as μ0. Based on the third and fourth constraints, if there is no This indicates that the reliability of the method described is low; otherwise, continue to calculate the second ratio between each depth data in each column of the intermediate data matrix and the median or mean of the column, and use the second ratio to form a correction data matrix M2 according to the position of the corresponding depth data.

[0166] Finally, the third ratio between the first column and the j+1th column in the correction matrix M2 is calculated to determine the CNV type:

[0167] For the region without CNV, that is, i<k or i>k+l; based on the above process, if there is Wherein, R1 is the third ratio in the CNV-free region; and based on the μ0 determined above, R1≈1 exists.

[0168] At the same time, for the CNV region, k≤i≤k+l; based on the above process, if there is also Wherein R2 is the third ratio in the CNV region; and based on the first constraint and the above-determined μ0, R2≈τ exists;

[0169] This indicates that the CNV region and type verified are consistent with the assumed situation, that is, the reliability of the described method is high; otherwise, it indicates that the reliability of the described method is poor.

[0170] For the detection method described in Example 1, R1≈1 and R2≈τ finally hold, which verifies that the detection method has good reliability.

[0171] As can be seen from the above steps, this embodiment discloses a reliability verification method based on numerical calculation. It gets rid of the dependence on repeated detection and verifies the reliability of the CNV detection based on the rationality of numerical logic. Not only is the verification process simple and convenient, but the verification results are more accurate and reliable. It is effectively proved that the results obtained by the CNV detection are always more consistent with the actual results.

[0172] The above method process can be run in a processor, or can also be stored in a memory (or computer-readable medium), which includes permanent and non-permanent, removable and non-removable media. Information storage can be achieved by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0173] These computer programs can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps of the functions specified in one or more blocks can be implemented by different modules corresponding to different steps.

[0174] Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. A person with ordinary knowledge in the technical field to which the present invention belongs may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be determined by the definition of the claims.

Claims

1. A CNV detection method based on a small panel, characterized in that: include: Obtaining an original data matrix, wherein one column of the original data matrix is ​​depth data of the sample to be tested, and the remaining columns are depth data of the control sample; Obtaining a geometric mean of each depth data in each row of the original data matrix, and calculating a first ratio between each depth data in each row of the original data matrix and the geometric mean of the row, and forming an intermediate data matrix according to the positions of the depth data corresponding to the first ratio; Obtaining the median or mean of each depth data in each column of the intermediate data matrix, and calculating a second ratio between each depth data in each column of the intermediate data matrix and the median or mean of the column, and forming a correction data matrix according to the position of the depth data corresponding to the second ratio; The ratio between the sample to be tested and the control sample in each amplified region is solved based on the calibration data matrix to obtain the CNV type of the sample to be tested.

2. The CNV detection method based on a small panel according to claim 1, characterized in that: The first column of the original data matrix is ​​the depth data of the sample to be tested.

3. The CNV detection method based on a small panel according to claim 1, characterized in that: The calculation formula of the geometric mean of each row in the original data matrix is: in: i=1, 2...n; j=1, 2...m; n is the total number of rows of the original data matrix, and m is the total number of columns of the original data matrix; a ij is the element in the i-th row and j-th column in the original data matrix.

4. The CNV detection method based on a small panel according to claim 1, characterized in that: The obtaining of the CNV type of the sample to be tested includes: Calculate the average or median of the depth data of each control sample in each row of the calibration data matrix to obtain a control average or a control median; Solving the ratio between the depth data of the sample to be tested in each row of the calibration data matrix and the control mean or control median of the row; The CNV type of the sample to be tested is determined according to the ratio.

5. A CNV detection system based on a small panel, characterized in that: include: An acquisition module is used to acquire an original data matrix, wherein one column of the original data matrix is ​​depth data of the sample to be tested, and the remaining columns are depth data of the control sample; A first calculation module is used to obtain a geometric mean of each depth data in each row of the original data matrix, and calculate a first ratio between each depth data in each row of the original data matrix and the geometric mean of the row, and form an intermediate data matrix according to the first ratio and the position of the depth data corresponding thereto; a second calculation module, used for obtaining the median or mean of each depth data in each column of the intermediate data matrix, and calculating a second ratio between each depth data in each column of the intermediate data matrix and the median or mean of the column, and forming a correction data matrix according to the position of the depth data corresponding to the second ratio; The CNV detection module is used to solve the ratio between the sample to be tested and the control sample in each amplification region based on the correction data matrix to obtain the CNV type of the sample to be tested.

6. A method for verifying the reliability of CNV detection based on a small panel, characterized in that: Used to verify the reliability of the detection method described in any one of claims 2 to 4, comprising: Assume an m×(n+1) dimensional original data matrix: Among them, the first column is the sample to be tested, and the remaining columns are control samples; For the sample to be tested, it is assumed that CNV occurs on the kth to k+lth genes; where τ (i) is the actual coefficient of variation of the ith gene that has mutated; the ideal coefficient of variation of the gene is recorded as τ, and the first constraint is established: |τ (i) -τ|<δ; Wherein, δ is the difference threshold between the actual coefficient of variation of the i-th gene and the ideal coefficient of variation; For the control samples, is the element in the j+1th column and the ith row; is the actual ratio of the element in the control sample to the corresponding element in the sample to be tested; the ideal ratio of the control sample in the j+1th column to the sample to be tested is recorded as b j , and create the second constraint: Among them, ε j is the difference threshold between the actual ratio in the j+1th column and the i-th row and the ideal ratio in the j+1th column; Get the geometric mean of each row element in the original data matrix: g i =a i med i ; For regions without CNV, that is, i<k or i>k+l: calculate the geometric mean of the actual ratios described in row i Calculate the geometric mean of the ideal ratios in row i And establish the third constraint: |med i -med|≤θ; Wherein, θ is the difference threshold between the geometric mean of the actual ratios described in the i-th row and the geometric mean of the corresponding ideal ratios; θ=max{θ1,θ2}; Wherein, Among them, Ο() is an order function, ε=max{ε1,…,ε j ,…,ε n }; For CNV regions, that is, k≤i≤k+l: calculate the geometric mean of the actual ratios described in the i-th row Calculate the geometric mean of the ideal ratios in row i And establish the fourth constraint: |med i -med|≤Θ; Where Θ is the difference threshold between the geometric mean of the actual ratios in the i-th row and the geometric mean of the corresponding ideal ratios; Θ = max{Θ1, Θ2}; Where, Calculate a first ratio between each element in each row of the original data matrix M and the geometric mean value of the row, and form an intermediate data matrix M1 according to the position of the corresponding depth data using the first ratio; Get the median of each element in each column of the intermediate data matrix M1: For the control sample, the median μ of the j+1th column is calculated based on the third constraint and the fourth constraint. j , we can get: Among them, β=max{θ,θ}; For the sample to be tested, the median μ0 is calculated based on the third constraint and the fourth constraint. If it does not exist, This indicates that the method described in claims 2-4 has low reliability; otherwise, continue to calculate the second ratio between each depth data in each column of the intermediate data matrix and the median or mean of the column, and use the second ratio to form a correction data matrix M2 according to the position of the corresponding depth data; The third ratio between the first column and the j+1th column in the correction matrix M2 is calculated to determine the CNV type: For regions without CNV, that is, i<k or i>k+l: if there is Then there exists R1≈1; wherein R1 is the third ratio in the CNV-free region; For CNV regions, that is, k≤i≤k+l: if there is Then there exists R2≈τ; wherein R2 is the third ratio in the CNV region; This means that the verified CNV region and type are consistent with the assumed situation, that is, the method described in claims 2-4 has high reliability; otherwise, it means that the method described in claims 2-4 has poor reliability.

7. The method for verifying the reliability of CNV detection based on a small panel according to claim 6, characterized in that: The process of establishing the third constraint includes: Establish the first inequality: Establish the second inequality: The third constraint is established based on the second constraint, the first inequality, and the second inequality.

8. The method for verifying the reliability of CNV detection based on a small panel according to claim 7, characterized in that: The process of establishing the fourth constraint includes: Establish the third inequality: Establish the fourth inequality: The fourth constraint is established based on the second constraint, the third inequality, and the fourth inequality.

9. The method for verifying the reliability of CNV detection based on a small panel according to claim 6, characterized in that: For the intermediate data matrix M1: For the region without CNV, i<k or i>k+1, based on the third constraint, we can get: For CNV regions, k≤i≤k+1, based on the fourth constraint, we can obtain:

Images