System, readable medium and application thereof for determining sample type based on DNA methylation level

Abstract

The present invention discloses a system for determining sample types based on DNA methylation levels, a readable medium and its application. The system includes a data analysis module, a feature extraction module and a scoring and typing module, and the specific contents are shown in the text of the present invention. The present invention can analyze the methylation information of known sample types based on the input information and judge the sample type to which the sample to be tested belongs; and has high sensitivity (up to 100%), reliable results, and can effectively solve the problem of low detection rate of existing products (such as CT). The tumor methylation variation information that needs to be invasively obtained by obtaining tissue samples in the prior art can be obtained non-invasively through the present invention.

Description

Technical Field

[0001] The present invention relates to the fields of bioinformatics, in vitro diagnosis and detection, and in particular to a system for determining sample type based on DNA methylation level, a readable medium and applications thereof. Background Art

[0002] DNA methylation is an important way for organisms to regulate the physiological activity of the genome. By methylating the promoter region of a gene, the expression capacity of the gene can be regulated, so that the same DNA can express different proteins in different cells without changing the DNA structure, thereby enabling the cells to perform different functions. At present, a large number of studies have reported the phenomenon of specific methylation of cancer-related genes observed in tumor tissues. In theory, by statistically summarizing such phenomena, the risk of an individual suffering from a certain cancer can be judged / predicted. However, in actual operation, since it is impossible to obtain tissue samples for analysis in ordinary screening, cancer screening requires the use of peripheral fluid (blood, saliva, urine, sweat, etc.) samples that can be obtained non-invasively. Although there is clear evidence that tumor cells release DNA fragments with cancer-specific markers into the peripheral fluid during apoptosis and can be detected, how to accurately detect its weak target signal in the complex and changing environment of the peripheral fluid is still a huge challenge in bioengineering.

[0003] The patent (application number PCT / CN2017 / 085150) applied by the applicant on May 19, 2017, discloses a method for identifying samples based on the distribution of reads (reads are DNA fragments interrupted during the sequencing process, i.e., sequencing reads) generated during the sequencing process of the second-generation sequencing technology on the reference genome. In essence, it is to identify samples based on the variation information of the DNA structure. However, for some diseases, DNA structural variation is not inevitable, so the technology of this application is not suitable for detecting signals of such diseases. Relatively speaking, changes in gene methylation are more common, with significant differences between healthy tissues, and are theoretically more suitable for tissue tracing of diseases.

[0004] There are methylation sites based on base C in the genome of organisms. These sites are often located in the promoter region of genes, affecting cell biochemistry and even the occurrence and development of tumors. Measuring the methylation rate of methylation sites is the basis for analyzing methylation information. However, on the one hand, due to the complex and changeable environment of peripheral fluid and the scarcity of free tumor DNA (especially for early tumors), it is very difficult to accurately obtain the methylation rate of each site, which brings great uncertainty to the analysis of tumor signals at the site level.

[0005] On the other hand, most detection methods based on methylation variation tend to treat the promoter region of a gene as a whole and analyze the overall indicators, resulting in excessive loss of details and decreased sensitivity. For example, the review "DNA-Methylation-Based Detection of Urological Cancer in Urine: Overview of Biomarkers and Considerations on Biomarker Design, Source of DNA, and Detection Technologies" published by Louise KL et al. lists a series of studies involving cancer detection through urine methylation data as of 2019. Among the DNA methylation biomarkers of primary bladder cancer, pyrophosphate sequencing was used. When and only when at least six of SOX1, TJP2, MYOD, HOXA9_1, HOXA9_2, VAMP8, CASP8, SPP1, IFNG, CAPG, HLADPA1 and RIPK3 were positive, the sensitivity and specificity of the detection were both 100%.

[0006] Therefore, the art lacks a highly sensitive detection system that can determine sample type based on methylation information. Summary of the invention

[0007] In response to the technical problem in the prior art of lacking a detection system capable of analyzing sample types based on methylation information and high-sensitivity DNA methylation levels, the present invention proposes a system for determining sample types based on DNA methylation levels, a readable medium and its application, which combines tumor biology knowledge and mathematical statistical analysis methods to analyze the methylation information of samples of the same type at the site level and the window level in turn, assisting in the judgment of the type to which the sample belongs, and obtaining a high-sensitivity detection system capable of analyzing sample types based on methylation information.

[0008] The inventor tried to analyze the DNA methylation level based on the patent applied for (application number PCT / CN2017 / 085150), and found that the difficulty lies in that the distribution pattern of reads (sequencing reads) relied on by the aforementioned patent belongs to regional structural variation, that is, reads abnormally increase or decrease in a certain area. At this time, relevant information can be obtained by dividing windows on the genome and counting the number of reads at the window level, but methylation variation belongs to site variation, and information at the window level cannot be directly obtained. On the one hand, in order to be able to analyze methylation variation information at the window level, the inventor analyzed the divided windows based on the methylation rate of the site, constructed the window methylation rate, screened out tumor-related features, and inferred the type of sample based on the degree of fit between the unknown type of sample and the said feature; on the other hand, the inventor used the S-fold cross-validation method to select the discriminant model, which further improved the sensitivity and accuracy.

[0009] The present invention is achieved through the following technical solutions:

[0010] The first aspect of the present invention provides a system for determining sample types based on DNA methylation levels, comprising: a data analysis module, a feature extraction module and a scoring and typing module, wherein:

[0011] The data analysis module is used to divide the reference genome into windows, calculate the window methylation rate of each sample based on the sequencing reads of the methylation sequencing library of the sample to be tested and each sample belonging to at least two known sample types, and number and sort the windows according to the obtained window methylation rates, thereby obtaining the window sequence of each sample;

[0012] The feature extraction module is used to screen the same-order subsequences of each sample and filter the common high-frequency same-order subsequences between the known sample types according to the window sequence of each sample of the known sample type, so as to obtain the feature sequence of each known sample type;

[0013] The scoring and typing module is used to assign weights to the feature sequences obtained in the feature extraction module, weight the window sequence of the sample to be tested corresponding to the feature sequences of each known sample type, obtain the score of the sample to be tested relative to the known sample type, and thereby determine the sample type to which the sample to be tested belongs.

[0014] The window division includes: dividing the reference genome into windows, wherein each window is a region containing a fixed number of methylation sites.

[0015] The reference genome may be conventional in the art, preferably hg39 or hg38.

[0016] In a preferred embodiment of the present invention, the dividing window divides the reference genome into windows in sequence based on the promoter region.

[0017] The fixed number of methylation sites may be 5-500, preferably 10-350, and more preferably 30-150.

[0018] In a preferred embodiment of the present invention, the number of the fixed number of methylation sites is 50.

[0019] The step of numbering and sorting the windows according to the methylation rates of the windows refers to: numbering the windows, and sorting the numbered windows in ascending or descending order according to the methylation rates of the windows.

[0020] The calculation formula of the methylation rate of the window is: the number of sequencing reads in which the base of any methylation site in the window is methylated / the total number of sequencing reads aligned to any methylation site in the window×100%.

[0021] In the feature extraction module, the high-frequency same-order subsequence refers to a same-order subsequence that appears in samples of the same sample type at a ratio higher than a set threshold.

[0022] The ratio threshold is 30-70%, preferably 50%.

[0023] The same order subsequence refers to a sequence consisting of numbers of windows with the same order in two or more window sequences. The same order subsequence is a sequence fragment obtained by connecting the sequence information of each window. If a window sequence contains all windows of a same order subsequence, and the numbering order of these windows is consistent with that in its same order subsequence, then the same order subsequence is said to appear in the window sequence, or the same order subsequence appears in the window sequence.

[0024] The filtering refers to removing high-frequency homologous subsequences that appear in at least two different sample types.

[0025] In the scoring and typing module, the weight assignment includes:

[0026] (1) Calculate the p-value of Fisher's exact test of the feature sequence based on the number of samples P1 that appear in the samples of the sample type to which the feature sequence belongs and the number of samples P2 that do not appear, and the number of samples P3 that appear in the samples of another sample type other than the sample type to which the feature sequence belongs and the number of samples P4 that do not appear;

[0027] (2) The p-value obtained in (1) is converted by a conversion function and used as the weight of the corresponding feature sequence.

[0028] The conversion function may be conventional in the art, including -1*log, -1*ln or the reciprocal.

[0029] A weight value is calculated for each feature sequence. The weight value is usually for at least two sample types to reflect the ability of the feature sequence to distinguish samples of the two sample types.

[0030] In a preferred embodiment of the present invention, said (1) further includes: discarding feature sequences with p-values ​​greater than a test threshold; preferably, the test threshold includes 0.05 or 0.01.

[0031] In the scoring and typing module, determining the sample type to which the sample to be tested belongs includes:

[0032] 1) For samples of different known sample types, preferably 10-fold cross validation is used to calculate the score set of samples corresponding to the known sample types;

[0033] 2) Calculate the 95% confidence intervals for different known sample types based on the score set in 1);

[0034] When the scores of the sample to be tested fall into the 95% confidence intervals of different known sample types at the same time, the type of the sample to be tested is rejected, otherwise 3) is performed;

[0035] 3) Compare the distance between the score of the sample to be tested and the mean or median of the score set in 1). When the distance between the score of the sample to be tested and the mean or median of the score set of the first known sample type is smaller than the distance between the score of the sample to be tested and the mean or median of the score set of the second known sample type, the sample to be tested is judged to belong to the first known sample type; otherwise, it belongs to the second known sample type.

[0036] In a preferred embodiment of the present invention, the system may further include a methylation sequencing module, which may perform methylation sequencing on samples of the test type to obtain sequencing data; the sequencing data may include a number of sequencing reads.

[0037] The methylation sequencing module may be a conventional second-generation sequencing platform or a single-molecule sequencing platform in the art, such as BGI-seq500.

[0038] DNA methylation sequencing and library construction are performed using conventional methylation sequencing techniques in the art, such as bisulfite sequencing technology or TET-assisted pyridine borane sequencing (TAPS) technology.

[0039] The known sample type includes a healthy type or a diseased type; the diseased type is preferably a tumor type.

[0040] The diseased type samples and healthy type samples are samples that have been obtained from peripheral fluid, tissues, etc.

[0041] The second aspect of the present invention provides a computer-readable medium, wherein the computer-readable medium stores a computer program, and when the computer program is executed by a processor, it can realize the functions of the system for determining the sample type based on the DNA methylation level as described in the first aspect of the present invention.

[0042] The third aspect of the present invention provides an apparatus for determining a sample type based on a DNA methylation level, comprising the computer-readable medium according to the second aspect of the present invention; and,

[0043] A processor is used to execute the computer program to realize the function of determining the sample type based on the DNA methylation level.

[0044] The fourth aspect of the present invention provides a use of the system as described in the first aspect of the present invention in preparing the computer-readable medium as described in the second aspect of the present invention and the device as described in the third aspect of the present invention.

[0045] On the basis of being in accordance with the common sense in the art, the above-mentioned preferred conditions can be arbitrarily combined to obtain the preferred embodiments of the present invention.

[0046] The reagents and raw materials used in the present invention are commercially available.

[0047] The positive and progressive effects of the present invention are:

[0048] The present invention provides a system for determining sample types based on DNA methylation levels and a storage medium that can implement the system and output results when executed. The system for determining sample types based on DNA methylation levels can analyze the methylation information of known sample types based on input information and judge the sample type to which the sample to be tested belongs; the inventor creatively proposes the concept of "window sequence" to solve the problem that the methylation rates of different samples cannot be directly compared; and the present invention has high sensitivity (up to 100%), high specificity (97%), and reliable results, which can effectively solve the problem of low detection rate of existing products (such as CT). The tumor methylation variation information that needs to be obtained invasively from tissue samples in the prior art can be obtained non-invasively through the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the process of the present invention.

[0050] Figure 2 This is a schematic diagram of the workflow of the data analysis module of the present invention;

[0051] In the figure: (a) is the site methylation rate calculation, (b) is the window methylation rate calculation, and (c) is the window sorting.

[0052] Figure 3This is a schematic diagram of the workflow of the feature extraction module and the scoring and typing module of the present invention;

[0053] In the figure: (a) is the feature extraction module, (b) is the scoring and typing module;

[0054] In the weight part of (b), the dotted line is the window sequence whose weight is lower than the set threshold, and the solid line is the window sequence whose weight is higher than the set threshold.

[0055] Figure 4 Schematic diagram of the method for determining the type of the sample to be tested in Example 3.

[0056] Figure 5 This is a schematic diagram of the effect of predicting the type of samples to be tested in Example 3. DETAILED DESCRIPTION

[0057] The present invention is further described below by way of examples, but the present invention is not limited to the scope of the examples. The experimental methods in the following examples without specifying specific conditions are carried out according to conventional methods and conditions, or selected according to the product specifications.

[0058] Example 1

[0059] like Figures 1 to 3 As shown, this embodiment includes: a methylation sequencing module, a data analysis module, a feature extraction module and a scoring and typing module, wherein: peripheral fluid samples are collected from suspected bladder cancer patients who have not yet received treatment and confirmed bladder cancer patients (tumor type) and healthy people (healthy type), and DNA methylation sequencing is performed using the BGI-seq500 sequencing platform to construct methylation sequencing libraries.

[0060] The methylation sequencing module performs DNA methylation sequencing on the constructed methylation sequencing library to obtain sequencing data; the sequencing data includes a plurality of sequencing reads.

[0061] The data analysis module is used to divide the reference genome into windows, calculate the window methylation rate of each sample based on the sequencing reads of the methylation sequencing library of the sample to be tested, the healthy type sample and the tumor type sample, and number and sort the windows according to the obtained window methylation rate, thereby obtaining the window sequence of each sample;

[0062] The feature extraction module is used to screen the homologous subsequences of each sample and filter the common high-frequency homologous subsequences between the healthy type and the tumor type according to the window sequences of each sample of the known sample type obtained, such as the healthy type sample and the tumor type sample, to obtain the feature sequences of the healthy type and the tumor type;

[0063] The scoring and typing module is used to assign weights to the feature sequences obtained in the feature extraction module, weight the feature sequences of the window sequence of the sample to be tested corresponding to the healthy type and the tumor type, obtain the score of the sample to be tested relative to the healthy type and the tumor type, and thereby determine the sample type to which the sample to be tested belongs.

[0064] As mentioned above, it is difficult to accurately measure the methylation rate of a single site through peripheral fluid samples, and most tumors do not depend on the status of a single site. Therefore, the inventors creatively proposed the concept of window methylation rate based on the site methylation rate. Figure 2 As shown in (a), the principle of calculating the site methylation rate is: assuming that there is a methylation site at a certain position on the genome, and some sequencing reads generated during the methylation sequencing process are aligned to the position (the range of fragment alignment on the genome covers the site), then the calculation formula for the methylation rate of the site is: the number of sequencing reads whose bases at the corresponding position of the site are methylated / the total number of sequencing reads aligned to the site. The so-called window refers to an area (DNA fragment) artificially divided on a reference genome (such as hg39 or hg38). Obviously, the reference genome can be divided into a series of windows, which may be of equal length or unequal length, adjacent or non-adjacent, or even overlapping. The principle of analyzing windows is: window division is performed for the reference genome region, and each window contains a fixed number of methylation sites. Since the methylation sites are not evenly distributed on the genome, this means that different windows are of unequal length in this scheme. As Figure 2 As shown in (b), each window contains 5 methylation sites. Since the division across promoter regions may result in the last one or more methylation sites in some promoter regions not being divided into windows due to insufficient number, thereby increasing the detection noise and causing the accuracy of the detection result to decrease, it is preferred to divide the window for each promoter region.

[0065] The methylation rate of the defined window is calculated as: the number of sequencing reads with methylated bases at any methylation site in the window / the total number of sequencing reads aligned to any methylation site in the window × 100%. Note that not all sequencing reads aligned to the window will be calculated, but they must cover at least one methylation site. The window methylation rate refers to the situation of multiple methylation sites at the same time, and in theory will be less susceptible to interference from environmental fluctuations.

[0066] The window sequence is abstracted based on the window methylation rate, and the methylation rate of the window is calculated based on the number of sequencing reads, so as to avoid the situation in which the total number of sequencing reads between different samples is different due to factors such as tumor purity and sequencing depth in actual operation, thereby making the methylation rates between different samples incomparable. To this end, the inventor creatively proposed the concept of "window sequence" to solve the problem that the methylation rates between different samples cannot be directly compared.

[0067] The window sequence refers to sorting a series of numbered windows according to the order of window methylation rate from high to low (or from low to high), and the sequence composed of the numbers of these sorted windows is called the window sequence. Figure 2 As shown in (c), the methylation levels of the five windows numbered 1 to 5 are in the order of 5>2>4>1>3, and the sequence obtained according to the high-to-low rule is: 5,2,4,1,3. A sample will get a window sequence, and the windows involved in the window sequence will span multiple promoter regions (the windows are numbered at the genome level to ensure that the window numbers of each promoter region are different). Since different samples are numbered at the same genome level, the window sequences of the samples are theoretically not easily affected by the fluctuations in the total number of sequencing reads, making the sequences of different samples comparable.

[0068] The window sequence of a sample involves the promoter regions of all genes, but in reality a tumor (or other diseases) often involves only a few key genes. That is, in order to extract the most critical partial sequence (i.e., characteristic sequence) from a complete window sequence, this embodiment introduces the concepts of homologous subsequences and high-frequency homologous subsequences.

[0069] The same-order subsequence refers to a sequence consisting of windows with the same order in two or more window sequences, that is, a set of windows with the same order of numbers, which is a sequence fragment obtained by connecting the order information of each window.

[0070] If there is a subsequence a in window sequence A, the numbers involved in subsequence a exist in the other two window sequences B and C, and the order of these numbers in window sequence A, window sequence B and window sequence C are the same, that is, subsequence a is called a subsequence of window sequence A, window sequence B and window sequence C with the same order, and subsequence a is said to appear in window sequence A, window sequence B and window sequence C.

[0071] like Figure 3As shown in (a), suppose there are three window sequences "5,2,4,1,3", "2,5,4,1,3" and "5,4,1,2,3". It can be observed that there is a longest subsequence "5,4,1,3", and the order of the four numbers 5,4,1,3 involved is consistent with the order in the three window sequences. In this case, the subsequence can be called the same-order subsequence of the three window sequences.

[0072] Those skilled in the art should understand that one sample will obtain one window sequence, and after comparing the window sequences of multiple samples of the same type, many subsequences of the same sequence can be extracted from the window sequences of multiple samples because the samples can be freely combined and the subsequences of the same sequence are not required to appear in all samples.

[0073] For multiple samples of the same type, it is necessary to find the same-order subsequence of the window sequence corresponding to these samples as the feature of the type. At this time, the subsequence can be called a high-frequency same-order subsequence of the type (because this subsequence appears frequently in samples of this type). In actual operation, it is not required that the subsequence appear in all samples of the same type. It only needs to exceed a certain ratio threshold to be considered as a high-frequency same-order subsequence of the type. That is, if the frequency of the same-order subsequence of multiple samples of the same type appearing in all samples exceeds the set ratio threshold, the same-order subsequence is considered to be a high-frequency same-order subsequence of the type.

[0074] In this embodiment, the ratio threshold is 50%. Generally speaking, the lower the ratio threshold, the more conducive it is to discovering subtypes in samples of this type (high-frequency subsequences unique to subtypes often do not appear in all samples of the parent type), and to eliminating the interference of abnormal samples (high-frequency subsequences are not required to appear in abnormal samples), but a lower ratio threshold will lead to an exponential increase in the amount of calculation.

[0075] Multiple window sequences of a type can be used to extract a series of subsequences of the same order, but these subsequences of the same order are not necessarily unique to the type, and it is necessary to further extract the subsequences unique to the type, that is, the feature sequence. Figure 3As shown in (b), suppose there are two types of samples, A and B. A series of homologous subsequences are extracted from each of the two types of samples to form their own homologous subsequence sets. The common elements in the two sets, that is, the high-frequency subsequences that appear in both types of samples, are removed and filtered. For the remaining homologous subsequences (the remaining homologous subsequences may appear with low frequency in sample A and high frequency in sample B, or appear with high frequency in sample A and low frequency in sample B, or appear with low frequency in both samples A and B, or may be a mixture of the above three situations, etc.), weights are assigned to the sequences according to the difference in the number of times each subsequence appears in different types of samples. A subsequence extracted from sample A will be assigned a higher weight if it appears more frequently in sample A and less frequently in sample B, and vice versa, thereby obtaining a characteristic sequence of this type.

[0076] This embodiment calculates the weight by the p-value of Fisher's exact test. The specific process is as follows:

[0077] (1) Set four parameters of the characteristic sequence of type A, which are the number of samples that appear in the samples of type A (P1) and the number of samples that do not appear (P2), and the number of samples that appear in the samples of type B (P3) and the number of samples that do not appear (P4);

[0078] (2) The p value in (1) is converted into the weight of the corresponding feature sequence by taking the reciprocal.

[0079] The reciprocal may also be other conversion functions such as f(x)=-1*log10(x), f(x)=-1*ln(x), f(x)=1 / x, etc., where log10 refers to a log function with base 10, ln refers to a log function with base e, and similarly, the base of the log function here may also be other numerical values ​​that make the log function mathematically meaningful (such as greater than 0), x represents the p value of the high-frequency same-order subsequence in (1), and f(x) represents the weight of the feature sequence.

[0080] The subsequences of the same order whose weights are greater than the set threshold are regarded as the characteristic sequences of the type to which they belong; the occurrence of characteristic sequences of different types of samples in the sample of the type to be tested is examined, and the sum of the weights of the characteristic sequences that appear is used as the scoring value of the type.

[0081] The threshold is set to p<0.01, and this parameter is affected by the number of currently available subsequences.

[0082] When it is confirmed that a certain type of feature sequence does not appear frequently in other types of samples, the feature sequence is a complementary sequence for other types.

[0083] Those skilled in the art should understand that a certain type may refer to a healthy type sample or a tumor type sample, and tumor type samples may be further divided into many subtypes according to the type of disease, such as lung cancer samples, colon cancer samples, gastric cancer samples, bladder cancer samples, pancreatic cancer samples, liver cancer samples, etc. In this embodiment, healthy type samples and bladder cancer samples are used.

[0084] Similar to the complementary sequence, there is a complementary type for a certain type or multiple types. The symptom characteristics of the complementary type sample are opposite to the symptom characteristics of the certain type or multiple types of samples. For example, the healthy type sample is the complementary type of the diseased sample, and the diseased sample is the complementary type of the healthy sample. Specifically, this embodiment takes the healthy type sample and the bladder cancer sample as an example, assuming that the high-frequency homologous subsequence set A is obtained based on the sample mining of healthy people, and the high-frequency homologous subsequence set B is obtained based on the bladder cancer patients. The high-frequency homologous subsequences in set A are filtered using the samples of bladder cancer patients and the weight assignment is calculated to obtain the characteristic sequence, so the characteristic sequence of set A becomes the complementary sequence of the bladder cancer patient samples; vice versa, the characteristic sequence of set B becomes the complementary sequence of the healthy person samples.

[0085] It is generally believed that the sample to be tested is more likely to belong to the corresponding type with a larger score. In some cases where high judgment accuracy is required and no judgment is allowed, if the score difference is less than a certain judgment threshold, it is considered impossible to judge. In theory, the larger the judgment threshold, the more accurate the judgment can be, but it will also cause more samples to be unable to judge. Those skilled in the art will understand that for a certain sample to be tested, it can be directly compared with a certain type and its complementary type, and it can also be compared with multiple types and their complementary types. For example, it can be compared with bladder cancer samples and healthy samples, and it can also be compared with bladder cancer samples, liver cancer samples and healthy samples.

[0086] In this embodiment, the scoring and typing module determines the type of sample to be tested by:

[0087] 1) A 10-fold cross validation was performed on tumor type samples and healthy type samples to obtain the score sets of tumor type samples and healthy type samples respectively;

[0088] 2) Calculate the 95% confidence interval of tumor type and healthy type according to the score set in 1);

[0089] When the score of the sample of the type to be tested falls into both the 95% confidence interval of the tumor type and the 95% confidence interval of the healthy type, the type of the sample to be tested is rejected, otherwise 3) is performed;

[0090] 3) Compare the distance between the score of the sample of the test type and the mean or median of the two score sets in 1). When the distance between the score of the sample of the test type and the mean or median of the score set of the tumor type is smaller than the distance between the score of the sample of the test type and the mean or median of the score set of the healthy type, the sample of the test type is judged to be a tumor type; otherwise, it is a healthy type.

[0091] Those skilled in the art will appreciate that all or part of the functions of the various methods in the above-mentioned embodiments can be implemented by hardware or by computer programs. When all or part of the functions of the various methods in the above-mentioned embodiments are implemented by computer programs, the program can be stored in a computer-readable storage medium, and the storage medium can include: read-only memory, random access memory, disk, optical disk, hard disk, etc., and the program is executed by a computer to implement the above-mentioned functions. For example, the program is stored in the memory of the device, and the program in the memory is executed to implement all or part of the above-mentioned functions. In addition, when all or part of the functions in the above-mentioned embodiments are implemented by computer programs, the program can also be stored in a storage medium such as a server, another computer, disk, optical disk, flash disk or mobile hard disk, and can be downloaded or copied and saved in the memory of the local device, or the system of the local device is updated. When the program in the memory is executed by the processor, all or part of the functions in the above-mentioned embodiments can be implemented.

[0092] Example 3

[0093] The samples in this example were obtained from urine of donors (30 bladder cancer patients and 10 healthy subjects). Methylation sequencing was performed using conventional methylation library construction, BGQ-seq500 sequencing platform, and sequencing depth of 30×. The number of methylation sites included in each window was set to 50, the ratio threshold of high-frequency homologous subsequences was 50%, the weight was the inverse of the p-value of Fisher's exact test (converted using the log function and multiplied by -1), and 10-fold cross validation was performed for a total of 10 rounds of testing.

[0094] In the present embodiment, 50 methylation sites are one window, which is the optimal parameter for this batch of data. Generally speaking, the more methylation sites a window contains, the more stable the statistical characteristics are, because it refers to more sites, but the more details are lost, because the same window only considers the overall index (window methylation rate), and ignores the internal differences; on the contrary, the fewer the number of methylation sites a window contains, the more details of methylation rate changes can be captured, but it is easy to cause the statistical characteristics to be unstable, and then cause false positives to increase. The improvement of sequencing depth helps to suppress / slow down the problem of false positives increased due to the number of methylation sites contained in the window.

[0095] The model trained in 10 rounds of testing is used to score each sample in the test set, and the validation set is scored with the same standard. The scoring process is as follows: a high-frequency homologous subsequence set A is obtained based on samples from healthy people, and a high-frequency homologous subsequence set B is obtained based on samples from bladder cancer patients. Since these two sets use each other's samples to filter and calculate the weights of homologous subsequences, they become each other's complementary sequences. For a window sequence of a sample of a type to be judged, its score is set to 0. The sum of the weights of the feature sequences matching the window sequence in set A and set B is counted separately (weighted), and the sum of the weights is the score of the test sample.

[0096] The matching means that if the common elements of two sequences are consistent in order, the two sequences are considered to be matched.

[0097] like Figure 4 As shown, in this embodiment, the type determination of the sample to be tested includes:

[0098] 1) A 10-fold cross validation was performed on tumor type samples and healthy type samples to obtain the score sets of tumor type samples and healthy type samples respectively;

[0099] 2) Calculate the 95% confidence interval of tumor type and healthy type according to the score set in 1);

[0100] When the score of the sample of the type to be tested falls into both the 95% confidence interval of the tumor type and the 95% confidence interval of the healthy type, the type of the sample to be tested is rejected, otherwise 3) is performed;

[0101] 3) Compare the distance between the score of the sample of the test type and the mean or median of the two score sets in 1). When the distance between the score of the sample of the test type and the mean or median of the score set of the tumor type is smaller than the distance between the score of the sample of the test type and the mean or median of the score set of the healthy type, the sample of the test type is judged to be a tumor type; otherwise, it is a healthy type.

[0102] The score shows the probability that the sample is from a particular type of donor, and the results are as follows Figure 5 shown.

[0103] The results of this example show that the method proposed by the present invention can effectively distinguish urine samples of healthy people and bladder cancer patients, with a sensitivity of 100% and a specificity of 97% when 0 is used as the dividing threshold. It can be applied to the screening and treatment detection of bladder cancer patients.

Claims

1. A system for determining sample type based on DNA methylation level, characterized in that: It includes: Data analysis module, feature extraction module and scoring and typing module, including: The data analysis module is used to divide the reference genome into windows, calculate the window methylation rate of each sample based on the sequencing reads of the methylation sequencing library of the sample to be tested and each sample belonging to at least two known sample types, and number and sort the windows according to the obtained window methylation rates, thereby obtaining the window sequence of each sample; The window division is as follows: the reference genome is divided into windows, wherein each window is a region containing a methylation site; the methylation rate of the window is calculated as follows: the number of sequencing reads in which the base of any methylation site in the window is methylated / the number of all sequencing reads aligned to any methylation site in the window×100%; The feature extraction module is used to screen the same-order subsequences of each sample and filter the common high-frequency same-order subsequences between the known sample types according to the window sequence of each sample of the known sample type, so as to obtain the feature sequence of each known sample type; The same-order subsequence refers to a sequence consisting of numbers of windows with the same order in two or more window sequences; The high-frequency homonymous subsequence refers to a homonymous subsequence that appears in samples of the same sample type at a rate higher than a set ratio threshold; the ratio threshold is 30-70%; the scoring and typing module is used to assign weights to the feature sequences obtained in the feature extraction module, weight the window sequence of the sample to be tested against the feature sequences of each known sample type, obtain the score of the sample to be tested relative to the known sample type, and thereby determine the sample type to which the sample to be tested belongs; The weight assignment is: assigning weights to the sequences according to the difference in the number of times each subsequence appears in samples of different types; In the scoring and typing module, determining the sample type to which the sample to be tested belongs includes: 1) Sample verification calculations for different known sample types are performed to obtain a score set of samples corresponding to the known sample types; 2) Calculate the 95% confidence intervals for different known sample types based on the score set in 1); When the scores of the sample to be tested fall into the 95% confidence intervals of different known sample types at the same time, the type of the sample to be tested is rejected, otherwise proceed to 3); 3) Compare the distance between the score of the sample to be tested and the mean or median of the score set in 1). When the distance between the score of the sample to be tested and the mean or median of the score set of the first known sample type is smaller than the distance between the score of the sample to be tested and the mean or median of the score set of the second known sample type, the sample to be tested is judged to belong to the first known sample type; otherwise, it belongs to the second known sample type.

2. The system according to claim 1, characterized in that In the data analysis module, Each window is a region containing a fixed number of methylation sites; The step of numbering and sorting the windows according to the obtained methylation rates of the windows refers to: numbering the windows, and sorting the numbered windows in ascending or descending order according to the methylation rates of the windows.

3. The system according to claim 1 or 2, characterized in that: In the data analysis module, the window division is as follows: window division is performed on the reference genome hg39 or hg38.

4. The system according to claim 1 or 2, characterized in that: In the data analysis module, the reference genome is divided into windows in sequence based on the promoter region.

5. The system according to claim 1 or 2, characterized in that: In the data analysis module, each window is a region containing 5 to 500 methylation sites.

6. The system according to claim 5, characterized in that Each window is a region containing 10 to 350 methylation sites.

7. The system according to claim 6, characterized in that Each window is a region containing 30 to 150 methylation sites.

8. The system according to claim 7, characterized in that Each window contains 50 methylation sites.

9. The system according to claim 2, characterized in that The numbers are sequential numbers.

10. The system according to claim 1 or 2, characterized in that: The filtering refers to removing high-frequency homologous subsequences that appear in at least two different sample types.

11. The system according to claim 1 or 2, characterized in that: In the feature extraction module, the ratio threshold is 50%.

12. The system of claim 1, wherein: In the scoring and typing module, the weight assignment includes: (1) Calculate the p-value of Fisher's exact test of the feature sequence based on the number of samples P1 that appear in the samples of the sample type to which the feature sequence belongs and the number of samples P2 that do not appear, as well as the number of samples P3 that appear in the samples of another sample type other than the sample type to which the feature sequence belongs and the number of samples P4 that do not appear; (2) The p-value obtained in (1) is converted by a conversion function and used as the weight of the corresponding feature sequence; the conversion function includes -1*log, -1*ln or the reciprocal.

13. The system of claim 12, wherein: The above (1) also includes: discarding feature sequences whose p-values ​​are greater than the test threshold.

14. The system of claim 13, wherein: The test threshold includes 0.05 or 0.

01.

15. The system of claim 1, wherein: In the scoring and typing module, the verification calculation is performed using a 10-fold cross-validation calculation; The confidence intervals are 95% confidence intervals.

16. The system of claim 1, wherein: The system also includes a methylation sequencing module, which can perform methylation sequencing on the sample to obtain sequencing data; the sequencing data includes a plurality of sequencing reads.

17. The system of claim 16, wherein: The methylation sequencing module is a second-generation sequencing platform or a single-molecule sequencing platform.

18. The system of claim 17, wherein: The methylation sequencing model is BGI-seq500.

19. The system according to any one of claims 16 to 18, characterized in that: The known sample types include complementary types.

20. The system of claim 19, wherein: The complementary types are a healthy type and a diseased type.

21. The system of claim 20, wherein: The disease type is a tumor type.

22. A computer readable medium, characterized in that The computer-readable medium stores a computer program, and when the computer program is executed by a processor, it can realize the functions of the system for determining the sample type based on the DNA methylation level as described in any one of claims 1 to 21.

23. A device for determining sample type based on DNA methylation level, characterized in that: include: The computer readable medium of claim 22; A processor is used to execute the computer program to realize the function of determining the sample type based on the DNA methylation level.

24. Use of the system according to any one of claims 1 to 21 in preparing the computer-readable medium according to claim 22.

25. Use of the system according to any one of claims 1 to 21 in the preparation of the device according to claim 23.

Images