Method for screening methylation detection markers
By screening and utilizing methylation markers and combining machine learning algorithms, the problems of low content and heterogeneity of tumor free DNA in cfDNA are solved, and efficient and low-cost early screening detection of cancer are achieved, improving the accuracy of detection and Sensitivity.
Patent Information
- Application Number
- CN202310666087.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-06
- Publication Date
- 2024-12-06
AI Technical Summary
In the prior art, early cancer screening detection based on cfDNA has extremely low tumor free DNA content and heterogeneity problems, resulting in high detection cost and high invasiveness. Traditional methylation quantification methods are insensitive to trace ctDNA and are difficult to distinguish. Random noise.
By screening methylation markers with differentiation, using methylation sequencing data to determine candidate areas, selecting significant differences as markers, and combining machine learning algorithms such as random forests to build a binary classification discriminant model to achieve healthy populations Distinguishing from the patient population.
With small sample size, screening and model training of methylated markers is achieved, achieving specificity of more than 80% and sensitivity of more than 90%, reducing detection costs and improving the accuracy of early cancer screening .
Smart Images

Figure SMS_1 
Figure SMS_2 
Figure SMS_8
Abstract
Description
Technical Field
[0001] The present application relates to the field of bioinformatics, and in particular to a method for screening markers based on methylation detection based on a bioinformatics method, and uses of the markers. Background Art
[0002] There are corresponding detection methods for various types of cancer in clinical practice, but those methods that can be used as the gold standard for detection are often costly and invasive, which greatly affects people's willingness to test. This is also an important reason for the low prevalence of early cancer screening. In recent years, some successful applications of plasma free DNA (cfDNA) in the fields of non-invasive prenatal diagnosis and disease detection have shown its great prospects as a new generation of molecular markers. Among them, abnormal cfDNA methylation patterns are highly promising molecular markers for disease diagnosis.
[0003] cfDNA enters the circulation when cells die as part of normal cell renewal or pathology. In healthy people, most of the free DNA molecules in plasma come from blood cells. When an organ or tissue has a pathology with increased cell death, more free cell DNA from the affected tissue or organ will be present in the circulation. Due to the obvious tissue differences in DNA methylation, the increase of cfDNA from a specific tissue or organ will change its normal methylation pattern. In addition, changes in the methylation pattern of a specific tissue or organ caused by changes in physiological state will also affect the methylation pattern of plasma cfDNA when it enters the blood circulation. Therefore, changes in the proportion of cfDNA from a specific tissue, or changes in the methylation pattern of cfDNA from a characteristic tissue, can lead to abnormalities in the final plasma free DNA methylation pattern. Therefore, identifying abnormal methylation patterns from a patient's plasma free DNA can effectively diagnose their diseased state and locate their diseased tissues or organs.
[0004] However, the extremely low tumor-free DNA (ctDNA) content in plasma in the early stages of cancer, as well as the heterogeneity within and between tumors, poses a huge challenge to cfDNA-based early cancer screening. Due to the extremely low tumor-free DNA (ctDNA) content in plasma in the early stages of cancer, traditional detection methods such as high-depth sequencing or small panels can only capture very small amounts of ctDNA fragments, resulting in extremely high false negatives. In addition, the traditional methylation quantification method, the beta value, is defined as the proportion of methylated reads at a CpG site. It is insensitive to changes in methylation levels caused by trace amounts of ctDNA and is difficult to distinguish from random noise. Summary of the invention
[0005] The present application provides a method for screening methylation markers, using the screening method to screen the obtained methylation markers, and using the methylation markers to distinguish between healthy people and sick people. In some embodiments, the methylation markers of healthy people and sick people have a degree of distinction in cfDNA. In some embodiments, the sick people are esophageal cancer patients. In some embodiments, the methylation markers have a higher sensitivity for distinguishing between healthy people and sick people than other markers.
[0006] Specifically, this application relates to:
[0007] 1. A method for screening markers for methylation detection, comprising:
[0008] (1) The step of determining the candidate area comprises:
[0009] Acquire first target data based on tissue, wherein the first target data is methylation sequencing data, which includes methylation level information of each CpG site in the tissue cell type;
[0010] Determine a candidate region based on the first target data, wherein the candidate region includes more than x CpG sites and the methylation levels of all CpG sites in the candidate region in the tissue cell type are comparable;
[0011] (2) The step of selecting a marker from the candidate region comprises:
[0012] Obtaining second target data of each candidate region in each sample of the first population and the second population, wherein the second target data is data obtained after statistical calculation based on the methylation sequencing data of each candidate region in each sample,
[0013] The candidate regions of the second target data that differ to a certain extent between the first population and the second population are used as markers based on methylation detection.
[0014] The first population and the second population are two groups of people with different characteristics that do not overlap.
[0015] Among them, it is necessary to make the methylation levels of all CpG sites in the candidate region equivalent in the tissue cell type, because adjacent CpG sites often form a continuous methylation functional region with a consistent methylation pattern. Therefore, it is required that the methylation pattern of CpG sites in the candidate region is consistent with biological laws and has biological significance.
[0016] 2. The method according to item 1, wherein the value of x is 3 to 6, preferably 4. When the value range of x is less than 3, there will be too few effective CpG sites, and the estimation error of the methylation degree of the read will be large; when the value range of x is greater than 6, it is easy to cause the candidate region to be too large, increase the cost of panel design, and it is unnecessary to verify.
[0017] 3. The method according to item 1, in the step of selecting a marker from the candidate region,
[0018] For each candidate region, obtain z methylation sequencing reads with different sequences from each sample, and each read contains data of at least y CpG sites, and obtain the methylation level of each read;
[0019] For z reads of each candidate region of each sample, the ratio of the number of reads in which the methylation level is within a specific numerical range to z is calculated, which is the second target data of the candidate region of the sample.
[0020] Based on the difference degree of the second target data, a person skilled in the art can easily distinguish the first target population from the second target population. 4. According to the method of item 3, wherein y=x. This is used to define the lower limit of the number of CpG sites in each read. When the number y is too small, there will be too few effective CpG sites, and the estimation error of the methylation degree of the read will be large.
[0021] 5. The method according to item 3, wherein the methylation level of the read is the ratio of the number of methylated CpG sites in the read to the total number of CpG sites in the read.
[0022] 6. The method according to item 3, wherein the z reads cover 50% to 100% of the CpG sites in the candidate region, preferably 75% of the CpG sites. When the proportion of CpG sites covered by the reads is too small, there are too few valid CpG sites, and the estimation error of the methylation degree of the reads is large; when the proportion of CpG sites covered by the reads is too large, the conditions are too strict, which easily leads to too few valid reads that meet the conditions;
[0023] 7. The method according to item 3, wherein z is at least 30 to 50. The value range of z is a suitable value range optimized by the inventors, because when the z value is too small, the number of valid reads is too small, and the estimation error of the methylation degree of the candidate region is large; and when the z value is too large, the condition is too strict, which easily leads to too few valid intervals that meet the condition;
[0024] 8. The method according to item 3, wherein the specific numerical range is any sub-range from 0 to 1, preferably: 0 to 25%, >25% and <75% or 75% to 1. In the embodiment, the specific numerical range is any one of the three equal ranges from 0 to 1.
[0025] 9. A method according to any one of items 1 to 8, wherein:
[0026] The first population is a population suffering from the target disease, and the second population is a population not suffering from the target disease.
[0027] 10. The method according to item 9, wherein the difference between the second target data of the candidate area in the first population and the second population reaches a certain degree means:
[0028] The sensitivity of distinguishing the first group of people from the second group of people based on the second target data of the candidate area reaches a certain level.
[0029] 11. The method according to item 10, wherein the sensitivity of the candidate area refers to:
[0030] When the median of the second target data of the second population is smaller than the median of the second target data of the first population, the ratio of the samples whose second target data of the candidate area in the first population is greater than or equal to the discrimination threshold to the total samples of the first population, wherein the discrimination threshold is greater than or equal to the 50% quantile of the second target data of the candidate area in the second population, preferably the 55% quantile, 60% quantile, 65% quantile, 70% quantile, 75% quantile, 80% quantile, 85% quantile, 90% quantile or 95% quantile; or
[0031] When the median of the second target data of the second population is greater than the median of the second target data of the first population, the ratio of samples of the candidate area in the first population whose second target data is less than or equal to the discrimination threshold to the total samples of the first population, wherein the discrimination threshold is less than or equal to the 50% quantile of the second target data of the candidate area in the second population, preferably the 45% quantile, 40% quantile, 35% quantile, 30% quantile, 25% quantile, 20% quantile, 15% quantile, 10% quantile or 5% quantile.
[0032] 12. The method according to item 11, wherein the sensitivity of the candidate area reaches a certain level means:
[0033] The candidate areas are sorted as a whole from high to low according to the sensitivity value, and the sensitivity of the candidate area is located in the first w candidate intervals in the sorting.
[0034] 13. The method according to item 12, wherein the value of w ranges from 5 to 50, preferably 20.
[0035] The discrimination threshold determines the accuracy of using the candidate area as a marker to distinguish the first and second populations. Because the purpose of the marker is to accurately distinguish the first and second populations, the sensitivity is meaningful only when the accuracy reaches a certain level. In the prior art, the accuracy that can be achieved by methylation markers is usually 60%, and the present application also determines the value range of parameter w through a large number of experiments on the premise of ensuring an accuracy of more than 80%. When w is less than 5, there are too few markers for subsequent modeling, and the model performance performs poorly in the test set with an AUC of 0.71. Continuing to increase the number of markers can bring about an improvement in model performance; when w is greater than 50, there are too many markers for subsequent modeling, the model is overfitted, and the model performance decreases. When w is about 20, such as 15, 16, 17, 18, 19, 20, 21, 22, 23, 24 or 25, the model performance is near the inflection point of performance improvement.
[0036] 14. The method according to item 1, wherein the methylation levels of the CpG sites in the candidate region are different in at least two tissue cell types, which can ensure that the candidate region has the most basic biological differences and can provide available biological information for subsequent marker screening and modeling.
[0037] 15. The method according to claim 14, wherein the at least two tissue cell types are selected from:
[0038] Adipocytes, bladder epithelial cells, blood B cells, blood granulocytes, blood monocytes, blood NK cells, blood T cells, osteoblasts, breast basal epithelial cells, breast ductal epithelial cells, colon epithelial cells, colon fibroblasts, skin fibroblasts, endometrial epithelial cells, epidermal keratinocytes, bone marrow erythroid progenitor cells, esophageal epithelial cells, fallopian tube epithelial cells, gallbladder epithelial cells, gastric mucosal epithelial cells, cardiomyocytes, cardiac fibroblasts, kidney epithelial cells, hepatocytes, alveolar epithelial cells, lung bronchial epithelial cells, neuronal cells, oligodendrocytes, ovarian epithelial cells, pancreatic acinar cells, pancreatic alpha cells, pancreatic beta cells, pancreatic delta cells, pancreatic ductal cells, prostate epithelial cells, skeletal muscle cells, small intestinal epithelial cells and thyroid epithelial cells.
[0039] 16. According to the method of item 1, the sample is a blood sample, and the blood sample is selected from: whole blood, plasma or serum.
[0040] When the marker is used to distinguish between the first and second populations through a blood sample, since the blood sample may contain cfDNA from various tissue cells of the body, the marker of the present application preferably has the ability to distinguish the above tissue cell types through methylation sequencing data. The tissue cell types are not limited to the above tissue cells, as long as the methylation degree of each CpG site in the screened marker has a degree of distinction in at least two tissue cells. The degree of distinction may be different depending on the statistical method.
[0041] 17. The method according to item 9, wherein the target disease is selected from one or more of cancer, liver disease, nonalcoholic fatty hepatitis, metabolic disease, autoimmune disease, cardiovascular disease, hypertension, systemic hypertension, pulmonary hypertension, vascular disease, atherosclerosis, diabetic vascular disease, vascular calcification, arrhythmia, heart failure, hypertrophy, fibrosis, cardiomyocyte proliferation and apoptosis, inflammation, diabetic cardiomyopathy, angiogenesis, neurological disorders, viral diseases, drug addiction, conditions caused by drug addiction, increased adipose tissue, insulin resistance, and osteoporosis;
[0042] Preferably,
[0043] The cancer is selected from any one of liver cancer, lung cancer, pancreatic cancer, breast cancer, cervical cancer, endometrial cancer, colorectal cancer, nasopharyngeal cancer, ovarian cancer, prostate cancer, skin cancer, bladder cancer, stomach cancer, prostate cancer, colon cancer, bone cancer, brain cancer, rectal cancer, esophageal cancer, tongue cancer, kidney cancer, uterine body cancer, testicular cancer, urinary cancer, melanoma, astrocytic carcinoma, meningioma, Hodgkin's lymphoma, non-Hodgkin's lymphoma, acute lymphocytic leukemia, chronic lymphocytic leukemia, acute myeloid leukemia, chronic myeloid leukemia, adult T-cell leukemia lymphoma, hepatocellular carcinoma, bronchial cancer, multiple myeloma, basal cell tumor, seminoma, chondrosarcoma, myosarcoma, and fibrosarcoma.
[0044] 18. The method according to item 9, wherein the target disease is esophageal cancer.
[0045] Through the method of screening markers for methylation detection of the present application, the parameters involved in the above items can be adjusted according to different diseases to achieve the purpose of screening the markers. In particular, in the embodiments of the present application, for esophageal cancer, the method of screening markers for methylation detection of the present application can achieve a sensitivity of 0.923 when the markers screened by the method have a specificity of 0.833.
[0046] 19. The method according to item 1, wherein the ratio of the number of samples of the first population to the number of samples of the second population is 2:1 to 1:2, preferably 1:1. In an ideal modeling practice, samples of different groups are required to be as uniform as possible, that is, the ratio of the two groups of samples is required to be close to 1:1, which is difficult to achieve in practical applications, but at least it is necessary to ensure that the number of samples in any group does not exceed twice that of the other group.
[0047] 20. The method according to item 1, wherein the number of samples of the first population and the second population is not less than 20 to 50, preferably not less than 30. The larger the sample size, the more samples can be used for modeling and model evaluation, the more stable the model obtained, and the more reliable the result of model evaluation, but the corresponding cost will be greatly increased. The sample size range of 20 to 50 can achieve the specificity and higher sensitivity required for clinical practice at a lower cost.
[0048] 21. A method for constructing a model for distinguishing a first population from a second population, comprising:
[0049] Obtaining methylation sequencing statistics of markers from samples of the first population and the second population;
[0050] Using the statistical data as input to construct a binary classification discriminant model;
[0051] Wherein, the marker is:
[0052] A marker obtained by the method of any one of items 1 to 20, or
[0053] Two or more or all of the markers selected from the following table:
[0054]
[0055]
[0056] As shown in the examples of the present application, using the markers obtained by the method of any one of items 1-20 or one or more, for example 20, of the markers in the above table to construct a model to distinguish between a first population and a second population can achieve a sensitivity of 0.923 while satisfying a specificity of 0.833.
[0057] 22. The method according to item 21, wherein the method for constructing a binary classification discriminant model is selected from:
[0058] Random forests, logistic regression, and support vector machines.
[0059] 23. The method according to item 22, wherein the step of obtaining methylation sequencing statistics of the marker comprises:
[0060] For the marker, z methylation sequencing reads with different sequences are obtained from each sample, and each read contains data of at least y CpG sites, and the methylation level of each read is obtained;
[0061] For z reads of each marker of each sample, the ratio of the number of reads in which the methylation level is within a specific numerical range to z is calculated, which is the methylation sequencing statistics of the marker of the sample.
[0062] 24. The method according to item 23, wherein the methylation level of the read is the ratio of the number of methylated CpG sites in the read to the total number of CpG sites in the read.
[0063] 25. The method according to item 24, wherein the z reads cover 50% to 100% of the CpG sites of the marker, preferably 75% of the CpG sites.
[0064] 26. A method according to item 23, wherein z is at least 30, preferably ≥50.
[0065] 27. The method according to item 23, wherein the specific numerical range is from 0 to 1, preferably: 0 to 25%, >25% and <75% or 75% to 1.
[0066] 28. A method for distinguishing a first population from a second population using a model constructed according to any one of the methods of items 21 to 27.
[0067] 29. The method according to item 21 or 28, wherein the first population is a population suffering from the target disease, and the second population is a population not suffering from the target disease, wherein the target disease is selected from one or more of cancer, liver disease, non-alcoholic fatty liver disease, metabolic disease, autoimmune disease, cardiovascular disease, hypertension, systemic hypertension, pulmonary hypertension, vascular disease, atherosclerosis, diabetic vascular disease, vascular calcification, arrhythmia, heart failure, hypertrophy, fibrosis, cardiomyocyte proliferation and apoptosis, inflammation, diabetic cardiomyopathy, angiogenesis, neurological disorders, viral diseases, drug addiction, conditions caused by drug addiction, increased adipose tissue, insulin resistance, and osteoporosis;
[0068] Preferably,
[0069] The cancer is selected from any one of liver cancer, lung cancer, pancreatic cancer, breast cancer, cervical cancer, endometrial cancer, colorectal cancer, nasopharyngeal cancer, ovarian cancer, prostate cancer, skin cancer, bladder cancer, stomach cancer, prostate cancer, colon cancer, bone cancer, brain cancer, rectal cancer, esophageal cancer, tongue cancer, kidney cancer, uterine body cancer, testicular cancer, urinary cancer, melanoma, astrocytic carcinoma, meningioma, Hodgkin's lymphoma, non-Hodgkin's lymphoma, acute lymphocytic leukemia, chronic lymphocytic leukemia, acute myeloid leukemia, chronic myeloid leukemia, adult T-cell leukemia lymphoma, hepatocellular carcinoma, bronchial cancer, multiple myeloma, basal cell tumor, seminoma, chondrosarcoma, myosarcoma, and fibrosarcoma.
[0070] 30. A probe composition comprising probes targeting one or more markers selected from the following table:
[0071] Marker number chromosome Starting position End position Target genes 1 chr1 224804716 224805088 CNIH3 2 chr2 38302118 38302190 CYP1B1 3 chr3 27766032 27766367 EOMES 4 chr4 113432549 113432657 NEUROG2 5 chr4 134073584 134073906 PCDH10 6 chr5 3594738 3595978 IRX1 7 chr5 16179012 16179268 MARCHF11 8 chr6 100055731 100056756 PRDM13 9 chr6 105388800 105388961 LIN28B-AS1 10 chr10 22623460 22623607 SPAG6 11 chr11 20181456 20182036 DBX1 12 chr12 113917865 113918236 LHX5 13 chr13 28500781 28501299 PDX1 14 chr13 53775490 53775596 OLFM4 15 chr14 29237294 29237532 FOXG1 16 chr14 37132008 37132598 PAX9 17 chr14 60977414 60977962 SIX6 18 chr18 13868635 13869240 MC5R 19 chr18 25755415 25756043 CDH2 20 chr19 54481832 54482102 MIR935
[0072] 31. A methylation sequencing composition comprising primers targeting one or more markers selected from the following table:
[0073] Marker number chromosome Starting position End position Target genes 1 chr1 224804716 224805088 CNIH3 2 chr2 38302118 38302190 CYP1B1 3 chr3 27766032 27766367 EOMES 4 chr4 113432549 113432657 NEUROG2 5 chr4 134073584 134073906 PCDH10 6 chr5 3594738 3595978 IRX1 7 chr5 16179012 16179268 MARCHF11 8 chr6 100055731 100056756 PRDM13 9 chr6 105388800 105388961 LIN28B-AS1 10 chr10 22623460 22623607 SPAG6 11 chr11 20181456 20182036 DBX1 12 chr12 113917865 113918236 LHX5 13 chr13 28500781 28501299 PDX1 14 chr13 53775490 53775596 OLFM4 15 chr14 29237294 29237532 FOXG1 16 chr14 37132008 37132598 PAX9 17 chr14 60977414 60977962 SIX6 18 chr18 13868635 13869240 MC5R 19 chr18 25755415 25756043 CDH2 20 chr19 54481832 54482102 MIR935
[0074] In summary, the implementation scheme of the present application can achieve the screening of methylation markers with a small sample size (20 to 50) through an innovative processing method for methylation sequencing data and the optimization of each parameter of the method one by one. And the methylation markers screened by the method can be used for the training of a binary classification discriminant model, so that the trained binary classification discriminant model has an accuracy of more than 90% in distinguishing between sick and non-sick people under the premise of not less than 80% specificity. In addition, the methylation markers screened by the method can be used to distinguish DNA from different tissue sources, and therefore can also be used for the detection of blood cfDNA samples, before obtaining tissue samples by surgery or puncture, it costs less to achieve early screening of cancer.
[0075] The preferred embodiments of the present application are described in detail above, but the present application is not limited thereto. Within the technical concept of the present application, the technical solution of the present application can be subjected to a variety of simple modifications, including the combination of various technical features in any other suitable manner, and these simple modifications and combinations should also be regarded as the contents disclosed in the present application, and all belong to the protection scope of the present application. The aspects and embodiments of the present application described herein include aspects and embodiments of "comprising", "consisting of" and "essentially consisting of...". BRIEF DESCRIPTION OF THE DRAWINGS
[0076] Figure 1The URA values of the 20 feature intervals of the training set samples were used as input, and a binary classification discriminant model was constructed using the random forest algorithm. The ROC curves were obtained by using the binary classification discriminant model to predict the training set and test set samples respectively. Specific implementation plan
[0077] The present application is described in detail below. Although specific embodiments of the present application are shown, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.
[0078] Unless otherwise indicated, the implementation of this application will adopt conventional molecular biology (including recombinant technology), microbiology, cell biology, biochemistry and genetics techniques, which are all within the scope of conventional technical means in this area. Such techniques are described in detail in the literature such as Molecular Cloning: A Laboratory Manual, Second Edition (Sambrook et al., 1989); Oligonucleotide Synthesis (MJ Gait, 1984 edition); Animal Cell Culture (RI Freshney, 1987 edition); Methods in Enzymology series (American Academic Press Co., Ltd.); Current Protocols in Molecular Biology (FM Ausubel et al., 1987 edition, and regular updates); PCR: The Polymerase Chain Reaction (Mullis et al., 1994 edition). The primers, probes, blockers and kits used in this application can be prepared using standard techniques known in the art.
[0079] Unless otherwise defined herein, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
[0080] definition
[0081] For the purpose of interpreting this specification, the following definitions will apply, and where appropriate, terms used in the singular will also include the plural, and vice versa. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by a person of ordinary skill in the art to which the technology belongs. All technical and patent disclosures cited herein are incorporated herein by reference in their entirety.
[0082] The application's "primer" means an oligonucleotide that can be used as a starting point for synthesis when placed under conditions that induce the synthesis of a primer extension product complementary to a nucleic acid chain, i.e., in the presence of nucleotides and an inducer such as a DNA or RNA polymerase and at a suitable temperature and pH, whether it is naturally occurring or synthetically produced in a purified restriction digest. Primers can be single-stranded or double-stranded, and must be long enough to induce the synthesis of a desired extension product in the presence of an inducer. The exact length of a primer depends on a variety of factors, including temperature, primer source, and the method used. For example, for diagnostic and prognostic applications, an oligonucleotide primer typically contains at least or more than about 9, 10, or 15, or 20, or 25 or more nucleotides, but it can contain fewer nucleotides or more nucleotides. Factors involved in determining the appropriate length of a primer are well known to those skilled in the art.
[0083] The "probe" of the present application, when referring to a nucleic acid sequence, is used in its ordinary sense to refer to a selected nucleic acid sequence that can hybridize with a target sequence under specified conditions and can be used to detect the presence of the target sequence. It will be understood by those skilled in the art that, in some cases, a probe can also be used as a primer, and a primer can be used as a probe.
[0084] "Homology", "identity" and "similarity" of the present application refer to the sequence similarity between two nucleic acid molecules. "Homology", "identity" or "similarity" can be determined by comparing the positions in each sequence, and the sequences can be aligned for the purpose of comparison. When equivalent positions in the compared sequences are occupied by the same base, the molecules are identical at that position; when equivalent sites are occupied by the same or similar amino acid (for example, similar in spatial properties or charged properties) residues, the molecules can be said to be homologous (similar) at that position. The expression of homology / similarity or identity percentage refers to the function of the number of identical or similar amino acids at the positions shared by the compared sequences. "Unrelated" or "non-homologous" sequences share less than 40% identity, preferably less than 25% identity with the sequences of the present application. When comparing two sequences, the absence of residues (amino acids or nucleic acids) or the presence of extra residues also reduces identity and homology / similarity. In a specific embodiment, two or more sequences or subsequences are considered to be substantially or significantly homologous, similar or identical if their sequences are about 60%, or about 65%, 70%, 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99% or more identical over a specified region, as determined using the BLAST or BLAST 2.0 sequence comparison algorithm with the default parameters described below or by manual alignment and visual inspection provided online by, for example, the National Center for Biotechnology Information (NCBI). This definition also relates to or can be used to test the complement of a sequence. Thus, to the extent permitted by the context of the present invention, for example, if the nucleotide sequence can be predicted to occur naturally in a DNA duplex, or can occur naturally in the form of one or both of the complementary strands, then the nucleotide sequence that is complementary to a specified target sequence or a variant thereof is itself considered to be "similar" to the target sequence, and when referring to "similar" nucleic acid sequences, single-stranded sequences, their complementary sequences, double-stranded strand complexes, sequences that can encode the same or similar polypeptide products, and any allowed variants of any of the above. Circumstances in which similarity must be limited to the analysis of a single nucleic acid strand sequence can include, for example, the detection and quantification of the expression of a specific RNA sequence or coding sequence in a cell. This definition also includes sequences with deletions and / or additions, as well as sequences with substitutions.In embodiments, the identity or similarity can be over a region that is at least about 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 10, 21, 22, 23, 24, 25, or more nucleotides in length, or over a region that is more than about 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, or more than about 100 nucleotides in length.
[0085] As used herein, the term "about" refers to the usual error range of each value that is readily known to those skilled in the art. Reference to an "about" value or parameter herein includes (and describes) embodiments for the value or parameter itself. As used herein, when the term "about" precedes a numerical value, it means within a range of 10% above or below the numerical value. For example, "about 100" covers 90 and 110.
[0086] As used herein, the singular forms "a," "an," and "the" include plural forms unless otherwise indicated.
[0087] In the present application, unless otherwise specified, "methylation sequencing" includes any means for obtaining methylation information of nucleic acid molecule sequences. The present application does not limit the method of methylation sequencing, and those skilled in the art can make appropriate choices according to actual conditions. For example, a variety of different methods can be used to detect nucleic acid molecules. Nucleic acid detection methods include, for example, PCR and nucleic acid hybridization (for example, Southern blot, Northern blot or in situ hybridization). Specifically, oligonucleotides (for example, oligonucleotide primers) capable of amplifying target nucleic acids can be used for PCR reactions. PCR methods generally include the following steps: obtaining a sample, isolating nucleic acids (for example, DNA, RNA or both) from the sample, and contacting the nucleic acid with one or more oligonucleotide primers, wherein the primers specifically hybridize with the template nucleic acid under conditions that enable template nucleic acid amplification to occur. In the presence of template nucleic acid, an amplification product is produced. The conditions for nucleic acid amplification and amplification product detection are known to those skilled in the art. A variety of improvements to basic PCR techniques have been developed, including but not limited to anchored PCR, RACE PCR, RT-PCR and ligase chain reaction (LCR). The primer pairs in the amplification reaction must anneal to opposite strands of the template nucleic acid and should maintain an appropriate distance from each other so that the polymerase can efficiently polymerize across the region and so that the amplified product can be easily detected, for example, using electrophoresis. For example, computer programs such as OLIGO (Molecular Biology Insights Inc., Cascade, Colo.) can be used to design oligonucleotide primers to help design primers with similar melting temperatures. Typically, the oligonucleotide primer is 9-30 or 40 or 50 nucleotides in length (e.g., 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49 or 50 nucleotides in length), but the oligonucleotide primer can be longer or shorter, as long as appropriate amplification conditions are used. In some embodiments, the method of methylation sequencing comprises Sanger sequencing, shotgun sequencing, second generation sequencing, or single nanopore sequencing.
[0088] "DNA methylation" of the present application refers to the addition of a methyl group to the 5-position of cytosine (C), which is usually (but not necessarily) in the case of a CpG (cytosine followed by guanine) dinucleotide. "Increased methylation" or "significant methylation" as used herein refers to the presence of at least one methylated cytosine nucleotide in a DNA sequence, wherein the corresponding C in a normal control sample (e.g., a DNA sample extracted from a non-cancerous cell or tissue sample or a DNA sample treated for methylation of DNA residues) is unmethylated, and in certain embodiments, at least 2, 3, 4, 5, 6, 7, 8, 9, 10 or more Cs may be methylated, wherein the Cs at these positions in the control DNA sample are unmethylated.
[0089] Unless otherwise specified, in this application, the "specificity" of a candidate region or marker refers to the ratio of negative test results in samples from patients without a specific clinical disease; the "sensitivity" of a candidate region or marker refers to the ratio of positive test results in samples from patients with a clear clinical disease.
[0090] As used herein, the term "cross validation" is a practical method for statistically grouping data into subsets and then using them for training and learning. The basic idea of cross validation is to randomly split the original data set according to the grouping situation, with one part used as a training set for training the learning model and determining parameters, and the other part used as a test set. The test set is used to test the predictive performance of the trained model or classifier, and to calculate and evaluate the performance indicators of the model or classifier.
[0091] The term "random forest" essentially belongs to a major branch of machine learning - Ensemble Learning, which is a method of integrating many decision trees into a forest and using it to predict the final result. Random forest, as the name implies, is to build a forest in a random way. The forest consists of many decision trees, and there is no correlation between each decision tree in the random forest. After obtaining the forest, when a new input sample enters, each decision tree in the forest is allowed to make a judgment to see which category the sample should belong to (for classification algorithms), and then see which category is selected the most, and predict the sample as that category.
[0092] Methods for screening markers
[0093] In one aspect, the present application provides a method for screening markers for methylation detection, comprising:
[0094] (1) The step of determining the candidate area comprises:
[0095] Acquire first target data based on tissue, wherein the first target data is methylation sequencing data, which includes methylation level information of each CpG site in the tissue cell type;
[0096] Determine a candidate region based on the first target data, wherein the candidate region refers to a region containing more than x CpG sites and the methylation levels of all CpG sites in the candidate region in the tissue cell type are comparable;
[0097] (2) The step of selecting a marker from the candidate region comprises:
[0098] Obtaining second target data of each candidate region in each sample of the first population and the second population, wherein the second target data is data obtained after statistical calculation based on the methylation sequencing data of each candidate region in each sample,
[0099] The candidate regions of the second target data that differ to a certain extent between the first population and the second population are used as markers based on methylation detection.
[0100] The first population and the second population are two groups of people with different characteristics that do not overlap.
[0101] "Methylation detection" of the present application refers to any determination of the methylation state of one or more CpG dinucleotide sequences in a DNA sequence.Methylation detection can use any method that can be used to detect the change of CpG dinucleotide methylation in DNA, including, for example, methylation-sensitive restriction endonuclease (MSRE) assay, methylation-specific or methylation-sensitive PCR (MS-PCR) assay, methylation-sensitive single nucleotide primer extension (Ms-SnuPE), high-resolution melting (HRM) analysis, bisulfite sequencing, pyrophosphate sequencing, methylation-specific single-stranded conformation analysis (MS-SSCA), combined bisulfite restriction analysis (COBRA), methylation-specific denaturing gradient gel electrophoresis (MS-DGGE), methylation-specific melting curve analysis (MS-MCA), methylation-specific denaturing high-performance liquid chromatography (MS-DHPLC), methylation-specific microarray (MSO).These determinations can be PCR analysis, quantitative analysis or southern imprint analysis using fluorescent labels.
[0102] "Detection" of the present application refers to any process of observing markers or marker changes (e.g., changes in marker methylation state or expression levels of nucleic acid or protein sequences) in a biological sample, regardless of whether the marker or marker change is actually detected. In other words, the act of detecting a marker or marker change in a sample is "detection", even if the marker is determined to be absent or below a sensitivity level. Detection can be quantitative, semi-quantitative or non-quantitative observation, and can be based on comparison with one or more control samples.
[0103] The "candidate region" and the "marker for methylation detection" screened out from the candidate region both represent a DNA fragment with specific DNA sequence characteristics. The characteristics at least include that the DNA sequence contains "more than x CpG sites". The term "CpG" is the abbreviation of "cytosine (C)-phosphate-guanine (G)" dinucleotide, on which the base sequence appears as cytosine C followed by guanine G, and the two are connected by a phosphodiester bond, where C is located at the 5' end and G is located at the 3' end. Under natural conditions, the cytosine in the CpG site can be methylated to 5'-methylcytosine. The occurrence of physiological processes such as diseases will change the methylation state of the CpG site, causing the cytosine to be methylated or demethylated. "x" is an integer greater than 1. In some embodiments, the value of x is 3, 4, 5 or 6. For example, when the value of x is 4, the number of CpG dinucleotides contained in the candidate region and the marker for methylation detection finally determined by screening is not less than 4. The number of CpG refers to the number of CpG dinucleotides in the candidate region or the marker for methylation detection when the bases in the DNA sequence are not changed after treatment. For example, the number of CpG dinucleotides in the candidate region or the marker for methylation detection in the original state without bisulfite treatment to convert C therein into uracil (U) or further into thymine (T).
[0104] For samples from a single individual, the methylation level of a CpG site refers to the proportion of reads that methylate the site in the reads that cover the CpG site of the sample. The methylation level of the CpG site in a sample set refers to the mean of the methylation levels of all single individual samples in the CpG site in this sample set. Therefore, whether it is the methylation level of a CpG site in a single sample or the methylation level of a CpG site in a sample set, it is a numerical value between 0 and 1 (including the endpoints). When describing "the methylation level of the CpG site in the tissue cell type", it theoretically refers to the methylation level of the CpG site in the sample set of the tissue cells of all individuals of a species (such as all mankind). For a specific species, the methylation level of its specific CpG site should be determined. However, when the method is actually used, since it is impossible to collect all the biological samples of the species, a sample set containing as many samples as possible (such as the WGBS sequencing data set specifically used in the embodiment of the present application) is usually selected to evaluate the methylation level of each CpG site. The sample set containing as many samples as possible can be appropriately selected by those skilled in the art as needed. For example, a sample set or sample data set from a healthy population is selected. When describing "the methylation levels of all CpG sites in the tissue cell type are equivalent" of a certain DNA region (such as a candidate region), it refers to all samples from the same tissue cell type (such as training set samples, samples of the tissue cell type in the WGBS sequencing data set, etc.), and the methylation levels of each CpG site in the DNA region are located in the same numerical interval. The same numerical interval is any sub-interval in the interval of 0 to 1 (including the endpoints). In some embodiments, the sub-interval is any sub-range in 0 to 1 (including the endpoints), including but not limited to any sub-range containing or not containing the endpoints obtained after the numerical range of 0 to 1 is divided into 2 equal parts, 3 equal parts, 4 equal parts, 5 equal parts, 6 equal parts, 7 equal parts, 8 equal parts, 9 equal parts, 10 equal parts and more equal parts. In some embodiments, the "specific numerical range" is the range of any union of two or more adjacent sub-ranges containing or excluding the endpoints obtained by dividing the numerical range from 0 to 1 (including the endpoints) into 2, 3, 4, 5, 6, 7, 8, 9, 10 and more equal parts. For example, in some embodiments, "the methylation levels of all CpG sites in the tissue cell type are equivalent" means that the methylation levels of each CpG site in the candidate region are 0 to 1 / 3, 1 / 3 to 2 / 3 or 2 / 3 to 1.
[0105] In some embodiments, in the method for screening markers for methylation detection, the first target data for each candidate region of each tissue is a screened read set obtained after deduplication and / or removal of unqualified methylation sequencing reads (reads or reads) based on the original data of methylation sequencing, wherein the number of screened reads for each candidate region of each tissue is z. It should be understood that z is a positive integer variable, and the z value for the same candidate region of different tissues, the z value for different candidate regions of different tissues, or the z value for different candidate regions of the same tissue can be the same or different. In this application, "deduplication" means that for multiple reads with exactly the same sequence, only one of them is retained, and the other reads are removed. Therefore, after deduplication, the sequences of any two reads for the same candidate region of the same tissue are different from each other. The sequences of reads are different from each other, which means that the read sequence lengths are different, and / or they do not have 100% sequence identity through sequence alignment. In some embodiments, the unqualified read refers to the read containing less than y CpG sites. In other words, the number of CpG sites in the screened read is greater than or equal to y. In some embodiments, x=y, that is, the minimum number of CpG sites allowed in the screened read is the same as the minimum number of CpG sites contained in the candidate region. In some embodiments, the x may also be greater than y or less than y, and those skilled in the art may adjust the y value according to the final required accuracy. In some embodiments, y is 3, 4, 5 or 6. It should be noted that in the present application, whenever the number of CpG sites of any category in the read is mentioned, it refers to the number of CpG sites of the category actually contained in the DNA segment corresponding to the read before the bisulfite treatment. Therefore, some DNA segments contain a number of CpG sites greater than y (including the case of being equal to y), because they have unmethylated CpG sites. Therefore, in the methylation sequencing process, due to the bisulfite treatment, the C in the unmethylated CpG site is converted to T, so that the number of CpG in the read reflecting the DNA sequence of the segment may be less than y, but the read is not removed during the data processing process. In some embodiments, the unqualified read refers to a read that covers too small a range of the candidate region, such as a read that covers less than 50% of the candidate region. In other words, the coverage of the candidate region by the screened reads is selected from any value between 50% and 100%. In some embodiments, the screened reads cover at least 60%, 62%, 65%, 67%, 75%, 77%, 80%, 82%, 85%, 87%, 90% or more of the sequence of the candidate region. In some embodiments, wherein z is at least any integer between 30 and 50.In some embodiments, z is at least 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50 or more.
[0106] In some embodiments, the first target data is represented by a readMeth result, wherein the readMeth includes the methylation (methylated or unmethylated) information of the CpG site in the read, such as the candidate region number (or interval index) covered by the read, the read number (or read index), and the coordinates (base number) of the CpG site.
[0107] The second target data is based on the statistical data of the first target data. In some embodiments, the method for obtaining the statistical data includes analyzing and counting the z reads, calculating the ratio of the number of reads whose methylation levels are within a specific numerical range in the z reads of each candidate region of each tissue to z, and the ratio is the second target data. In the present application, the methylation level of a certain read refers to the degree of methylation of the DNA sequence segment corresponding to the read. In some embodiments, the methylation level of a certain read is the ratio of the number of methylated CpG sites in the read to the number of all CpG sites in the read. Therefore, in some embodiments, the methylation level is represented by any fraction from 0 to 1 (including 0 and 1). In some embodiments, the "specific numerical range" is any sub-range from 0 to 1 (including sub-ranges with endpoints containing 0 or 1), including but not limited to the sub-ranges obtained by dividing the numerical range from 0 to 1 into 2 equal parts, 3 equal parts, 4 equal parts, 5 equal parts, 6 equal parts, 7 equal parts, 8 equal parts, 9 equal parts, 10 equal parts and more equal parts, including or not including the endpoints. In some embodiments, the "specific numerical range" is a union of two or more adjacent sub-ranges obtained by dividing the numerical range of 0 to 1 into 2 equal parts, 3 equal parts, 4 equal parts, 5 equal parts, 6 equal parts, 7 equal parts, 8 equal parts, 9 equal parts, 10 equal parts and more equal parts. In some embodiments, the "specific numerical range" is, for example, 0 to 25% (including the endpoint), 25% to 75% (excluding the endpoint), or 75% to 1 (including the endpoint), etc. In some embodiments, the specific numerical range refers to a sub-range or a union of sub-ranges containing an endpoint value of 0 or 1. In some embodiments, the specific numerical range refers to a sub-range or a union of sub-ranges containing an endpoint value of 0, then the read whose methylation level value is within the specific numerical range is called Uread, and the second target data is called URA. In some embodiments, the specific numerical range refers to a sub-range or a union of sub-ranges containing an endpoint value of 1, then the read whose methylation level value is within the specific numerical range is called M read, and the second target data is called MRA. In some embodiments, the specific numerical range refers to a subrange or a union of subranges that does not include endpoint values 0 and 1, and the read whose methylation level value is within the specific numerical range is called X read, and the second target data is called XRA.
[0108] In the present application, the first population and the second population are populations with specific differences in physiological characteristics. In some embodiments, the first population is a population suffering from the target disease, and the second population is a population not suffering from the target disease. In some embodiments, the target disease is selected from cancer, liver disease, non-alcoholic fatty hepatitis, metabolic diseases, autoimmune diseases, cardiovascular diseases, hypertension, systemic hypertension, pulmonary hypertension, vascular diseases, atherosclerosis, diabetic vascular disease, vascular calcification, arrhythmia, heart failure, hypertrophy, fibrosis, cardiomyocyte proliferation and apoptosis, inflammation, diabetic cardiomyopathy, angiogenesis, neurological disorders, viral diseases, drug addiction, conditions caused by drug addiction, increased adipose tissue, insulin resistance, osteoporosis, one or more of them. In some embodiments, the target disease is cancer, and the cancer is selected from any one of liver cancer, lung cancer, pancreatic cancer, breast cancer, cervical cancer, endometrial cancer, colorectal cancer, nasopharyngeal cancer, ovarian cancer, prostate cancer, skin cancer, bladder cancer, stomach cancer, prostate cancer, colon cancer, bone cancer, brain cancer, rectal cancer, esophageal cancer, tongue cancer, kidney cancer, uterine body cancer, testicular cancer, urinary cancer, melanoma, astrocytic carcinoma, meningioma, Hodgkin's lymphoma, non-Hodgkin's lymphoma, acute lymphatic leukemia, chronic lymphatic leukemia, acute myeloid leukemia, chronic myeloid leukemia, adult T-cell leukemia lymphoma, hepatocellular carcinoma, bronchial cancer, multiple myeloma, basal cell tumor, seminoma, chondrosarcoma, myosarcoma, and fibrosarcoma. In some embodiments, the target disease is esophageal cancer. In some embodiments, the cancer is primary or secondary.
[0109] In some embodiments, the number of samples of the first population is greater than the number of samples of the second population. In some embodiments, the number of samples of the first population is less than the number of samples of the second population. In some embodiments, the number of samples of the first population is equal to the number of samples of the second population. Those skilled in the art can adjust the ratio of the number of samples of the first population and the number of samples of the second population according to the final screening of the candidate area, and can re-supplement or increase or decrease the samples of the first population or the second population. In some embodiments, the ratio of the number of samples of the first population to the second population is any value from 2:1 to 1:2. As used herein, the "number of samples" of the first population and the second population refers to the number of samples or data of the samples of the same candidate area of the same tissue. For example, when the sample of the first population is a sample of n healthy human esophageal epithelial cells, or when the data of the first population is sample data of n healthy human esophageal epithelial cells (e.g., methylation sequencing data, first target data, or second target data), the number of samples of the first population is n. For example, when the sample of the second population is an embassy epithelial cell sample of m esophageal cancer patients, or when the data of the second population is sample data of n healthy human esophageal epithelial cells (e.g., methylation sequencing data, first target data, or second target data), the number of samples of the second population is m. In some embodiments, the number of samples of the first population and the second population is not less than any integer from 20 to 50. In some embodiments, the number of samples of the first population and the second population is not less than 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, or 50. Those skilled in the art can comprehensively select the number of samples of the first population and the second population according to the difficulty of data acquisition and the test results of the screened markers.
[0110] As mentioned above, the marker for methylation detection is a candidate region where the second target data differs to a certain extent between the first population and the second population, and the difference to a certain extent means:
[0111] The sensitivity of distinguishing the first population from the second population based on the second target data of the candidate area reaches a certain level. In the present application, the sensitivity of the candidate area refers to: when the median of the second target data of the second population is less than the median of the second target data of the first population, the ratio of the samples whose second target data of the candidate area in the first population is greater than or equal to the discrimination threshold to the total samples of the first population, wherein the discrimination threshold is greater than or equal to the 50% quantile of the second target data of the candidate area in the second population, such as the 55% quantile, 60% quantile, 65% quantile, 70% quantile, 75% quantile, 80% quantile, 85% quantile, 90% quantile or when the median of the second target data of the second population is greater than the median of the second target data of the first population, the ratio of the samples whose second target data of the candidate area in the first population is less than or equal to the discrimination threshold to the total samples of the first population, wherein the discrimination threshold is less than or equal to the 50% quantile of the second target data of the candidate area in the second population, such as the 45% quantile, 40% quantile, 35% quantile, 30% quantile, 25% quantile, 20% quantile, 15% quantile, 10% quantile or 5% quantile. The median of the second target data of the first population or the second population refers to the median of the second target data of the same candidate area obtained from different samples in the population. The discrimination threshold is determined according to the specificity required to distinguish the second population. According to the requirements of specificity, for example, if the specificity of the marker for distinguishing the second population is required to reach 90%, for a certain candidate area, when the median of the second target data of the second population is smaller than the median of the second target data of the first population, the discrimination threshold is set to the 90% quantile of the second target data of the candidate area in the second population; when the median of the second target data of the second population is larger than the median of the second target data of the first population, the discrimination threshold is set to the 10% quantile of the second target data of the candidate area in the second population.
[0112] In some embodiments, the sensitivity of the candidate area reaches a certain level, which means that the sensitivity of the candidate area is in the first w candidate intervals in the ranking, and the candidate areas are ranked according to the sensitivity value from high to low. In some embodiments, the value range of w is 1 to 50, such as 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, or 49.
[0113] Since some samples may include DNA from at least two tissue cell sources, for such samples, in some embodiments, it is also necessary that the methylation level of any CpG site in the candidate region is different in at least two tissue cell types. In some embodiments, the sample is a blood sample. In some embodiments, the blood sample is selected from: whole blood, plasma or serum. In some embodiments, the sample is plasma free DNA (cfDNA). In some embodiments, the at least two tissue cell types are selected from the group consisting of: adipocytes, bladder epithelial cells, blood B cells, blood granulocytes, blood monocytes, blood NK cells, blood T cells, osteoblasts, breast basal epithelial cells, breast ductal epithelial cells, colon epithelial cells, colon fibroblasts, skin fibroblasts, endometrial epithelial cells, epidermal keratinocytes, bone marrow erythroid progenitor cells, esophageal epithelial cells, fallopian tube epithelial cells, gallbladder epithelial cells, gastric mucosal epithelial cells, cardiomyocytes, cardiac fibroblasts, kidney epithelial cells, hepatocytes, alveolar epithelial cells, lung bronchial epithelial cells, neurons, oligodendrocytes, ovarian epithelial cells, pancreatic acinar cells, pancreatic α cells, pancreatic β cells, pancreatic δ cells, pancreatic duct cells, prostate epithelial cells, skeletal muscle cells, small intestinal epithelial cells, and thyroid epithelial cells.
[0114] Specifically, in some embodiments, the method for screening markers for methylation detection is a method for screening tumor markers for cfDNA methylation detection. In some embodiments, the tumor is esophageal cancer. In some embodiments, the method comprises the following steps:
[0115] (1) Identification of candidate intervals
[0116] From the acquired public data, candidate intervals are identified. The original methylation sequencing data of the first population is obtained from the public data, and the data includes WGBS sequencing data of purified cell types from various major tissues and organs, and the homogeneous methylation intervals (homogenously methylated blocks, hereinafter referred to as HMBs) are identified using an internal self-built process, so that the methylation level between each CpG site in each HMB interval in each cell type is uniform, that is, it is either uniformly high methylation or uniformly low methylation. The "uniform" means that the methylation levels of all CpG sites in the HMB interval in the tissue cell type are between 0 and 1 / 3, 1 / 3 to 2 / 3, or 2 / 3 to 1.
[0117] (2) Design a panel (probe combination) for specific capture and high-depth methylation sequencing of candidate intervals
[0118] Based on the HMB intervals identified in the previous step, the intervals whose methylation levels have a certain degree of difference in each cell type are screened as candidate intervals, the CpG sites therein are identified, and probes are designed and synthesized to obtain a panel for specifically capturing and enriching cfDNA molecules in the candidate intervals. Among them, the intervals with a certain degree of difference in each cell type refer to the methylation levels of each CpG site in the interval being different in at least two tissue cell types.
[0119] (3) Extract ReadMeth
[0120] Blood samples were obtained from the test subjects, cfDNA samples were extracted from the test samples, and then panel capture and methylation library sequencing were performed to obtain sequence reads from the cfDNA samples.
[0121] Through a series of basic downstream analyses, the alignment results and methylation site information of reads are obtained. First, the sequencing quality is checked using fastp quality control software, and low-quality sequencing bases or reads are removed. Then, the quality-controlled data are aligned to the reference genome hg19 using bismark alignment software, and sequencing duplications are removed to obtain the deduplicated alignment results (bam file), which contains detailed alignment results of reads and methylation information of each site on the reads.
[0122] Analyze the detailed alignment results of the above reads and the methylation information of each site on the reads to obtain the reads falling within the candidate interval and the methylation status information of the CpG sites therein, which is called readMeth. Use the internal self-built analysis process to complete the task of extracting readMeth information.
[0123] (4) Identification of UXM reads
[0124] For each candidate interval, only reads containing at least 4 CpG sites were retained, and the methylation level u of each read was calculated as the ratio of the number of methylated CpG sites in the total number of CpG sites, that is, Where C is the number of methylated CpG sites and T is the number of unmethylated CpG sites. Then the read is classified as U read (mostly unmethylated read); if Then the read is classified as M read (mostly methylated read); otherwise The read is classified as X read (mixed methylated read).
[0125] (5) Calculation of URA
[0126] For each candidate interval, the U reads abundance (URA) is calculated, that is, the proportion of U reads to all reads in the interval, that is, In order to improve the accuracy of URA quantification, the minimum number of valid reads is set to 50, that is, U+X+M≥50. This method is also applicable to calculating the abundance of X or M reads. For the convenience of description below, only the case of URA is discussed in this embodiment. XRA and MRA can also be used in the subsequent steps of this embodiment.
[0127] (6) Marker screening and modeling evaluation
[0128] Healthy human plasma samples and cancer plasma samples are divided into training sets and test sets according to a certain ratio. The number of healthy human samples and cancer samples in the training set is required to be close (for example, the ratio of the number of healthy human samples to cancer samples in the training set is close to 1:1), that is, to ensure the balance of samples in the training set.
[0129] In the training set, two groups of candidate intervals with discriminability in URA features were screened. The screening method is that for each candidate interval, the cancer sample is set as the positive group and the healthy person sample is set as the negative group. Under the premise of setting the specificity of the negative group at 0.9, the sensitivity of the positive group is calculated. The calculation method is that if the median of the negative group is smaller than that of the positive group, the 90% quantile of the negative group is calculated as the threshold, and the value of the positive group is compared with the threshold. The proportion of the number of positive groups with values not lower than the threshold in all positive groups is the sensitivity value; if the median of the negative group is larger than that of the positive group, the 10% quantile of the negative group is calculated as the threshold, and the value of the positive group is compared with the threshold. The proportion of the number of positive groups with values not greater than the threshold in all positive groups is the sensitivity value. The 20 candidate intervals with the largest sensitivity values are screened as the feature intervals for post-selection modeling.
[0130] Modeling evaluation. The URA values of the 20 feature intervals of the training set samples were used as input, and a binary classification model was constructed using the random forest algorithm. The test set samples were predicted and the model prediction performance was evaluated.
[0131] Further, in some embodiments, the method for screening markers also includes performing one or more (e.g., 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 or more) repeated experiments on the aforementioned embodiments, and selecting the intersection of the markers screened in each experiment as the final selected marker for methylation detection. In some embodiments, the repeated experiments use different sample sets or sample data sets to determine the methylation level of the candidate region CpG site. In some embodiments, the repeated experiments use different test sets (first population and second population) samples to determine the marker from the candidate region. In some embodiments, the repeated experiments use different sample sets or sample data sets to determine the methylation level of the candidate region CpG site, and the repeated experiments use different test sets (first population and second population) samples to determine the marker from the candidate region.
[0132] Modeling approach
[0133] A second aspect of the present application also provides a method for constructing a model for distinguishing a first population from a second population, comprising:
[0134] Obtaining methylation sequencing statistics of markers from samples of the first population and the second population;
[0135] Using the statistical data as input to construct a binary classification discriminant model;
[0136] Wherein, the marker is:
[0137] A marker determined according to the aforementioned method for screening markers, or
[0138] Any two, more or all of the gene loci or their corresponding nucleic acid sequence fragments selected from the following table:
[0139]
[0140]
[0141] In some embodiments, the more types are 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18 or 19. Any binary classification discriminant model can be used in the modeling method of this aspect. In some embodiments, the method for constructing the binary classification discriminant model is random forest method, logistic regression, support vector machine, or any combination thereof.
[0142] It should be understood that the methylation sequencing statistical data may be any statistical data that can distinguish samples of the first population and the second population based on the methylation sequencing data. The present application does not limit the statistical method of the data, and those skilled in the art may make appropriate selections according to actual conditions.
[0143] In some embodiments, the methylation sequencing statistical data is the second target data collected from the marker, and the step of obtaining the methylation sequencing statistical data of the marker is the same as the step of obtaining the second target data of the candidate region in the aforementioned method for screening markers, including:
[0144] For the marker, z methylation sequencing reads with different sequences are obtained from each sample of the first population and the second population, and each read contains data of at least y CpG sites, and the methylation level of each read is obtained;
[0145] For each marker of each sample, z reads, the ratio of the number of reads in which the methylation level is within a specific numerical range to z is calculated, which is the statistical data. In some embodiments, the coverage range of the marker by the reads is selected from any value between 50% and 100%. In some embodiments, the screened reads cover at least 60%, 62%, 65%, 67, 75%, 77%, 80%, 82%, 85%, 87%, 90% or more of the sequence of the marker. In some embodiments, wherein z is at least any integer between 30 and 50. In some embodiments, z is at least 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50 or more.
[0146] In some embodiments, the method for obtaining the statistical data includes analyzing and counting the z reads, calculating the ratio of the number of reads whose methylation levels are within a specific numerical range in the z reads for each marker of each tissue to z, and the ratio is the second target data. In the present application, the methylation level of a certain read refers to the degree of methylation of the DNA sequence segment corresponding to the read. In some embodiments, the methylation level of a certain read is the ratio of the number of methylated CpG sites in the read to the number of all CpG sites in the read. Therefore, in some embodiments, the methylation level is represented by any fraction from 0 to 1 (including 0 and 1). In some embodiments, the "specific numerical range" is any sub-range from 0 to 1 (including a sub-range with an endpoint of 0 or 1), including but not limited to a sub-range containing or not containing an endpoint obtained by dividing the numerical range from 0 to 1 into 2 equal parts, 3 equal parts, 4 equal parts, 5 equal parts, 6 equal parts, 7 equal parts, 8 equal parts, 9 equal parts, 10 equal parts and more equal parts. In some embodiments, the "specific numerical range" is a union of two or more adjacent sub-ranges obtained by dividing the numerical range of 0 to 1 into 2 equal parts, 3 equal parts, 4 equal parts, 5 equal parts, 6 equal parts, 7 equal parts, 8 equal parts, 9 equal parts, 10 equal parts and more equal parts. In some embodiments, the "specific numerical range" is, for example, 0 to 25% (including the endpoint), 25% to 75% (excluding the endpoint), or 75% to 1 (including the endpoint), etc. In some embodiments, the specific numerical range refers to a sub-range or a union of sub-ranges containing an endpoint value of 0 or 1. In some embodiments, the specific numerical range refers to a sub-range or a union of sub-ranges containing an endpoint value of 0, then the read whose methylation level value is located in the specific numerical range is called U read, and the second target data is called URA. In some embodiments, the specific numerical range refers to a sub-range or a union of sub-ranges containing an endpoint value of 1, then the read whose methylation level value is located in the specific numerical range is called M read, and the second target data is called MRA. In some embodiments, the specific numerical range refers to a subrange or a union of subranges that does not include endpoint values 0 and 1, and the read whose methylation level value is within the specific numerical range is called Xread, and the second target data is called XRA.
[0147] In some embodiments, the first population is a population suffering from the target disease, and the second population is a population not suffering from the target disease. In some embodiments, the target disease is a tumor. In some embodiments, the sample is a sample of tumor-derived tissue, that is, the sample of the first population is a healthy tissue sample corresponding to the tumor-derived tissue, and the sample of the second population is a tumor tissue sample. For example, the sample of the first population is an esophageal epithelial tissue cell sample, and the sample of the second population is an esophageal cancer malignant tumor tissue sample. In some embodiments, the sample is a cfDNA or blood sample, that is, the sample of the first population is a cfDNA or blood sample of a subject who does not suffer from esophageal cancer, and the sample of the second population is a cfDNA or blood sample of a subject suffering from esophageal cancer. In some embodiments, the blood sample is whole blood, plasma or serum.
[0148] In this aspect, the number of samples of the first population and the second population is not limited. Those skilled in the art can select appropriate samples and sample sizes to train the model based on the accuracy and specificity of distinguishing the first population and the second population.
[0149] Probes, primers and combinations thereof
[0150] The third aspect of the present application provides probes, primers, and combinations thereof targeting methylation detection markers, and libraries composed of the probes and / or primers. Wherein, the methylation detection marker is a marker determined according to the aforementioned method for screening markers. The probes and primers targeting methylation detection markers contain a polynucleotide sequence complementary to the complementary sequence of the methylation detection marker, the basicization detection marker, or the complementary sequence of the methylation detection marker or the basicization detection marker after bisulfite treatment. The probes and primers can be DNA, RNA, or a hybrid molecule of DNA and RNA.
[0151] As used herein, the "complementarity" of a nucleic acid or sequence refers to the ability of a nucleic acid or sequence to form hydrogen bonds with another nucleic acid or sequence by traditional Watson-Crick base pairing. In some embodiments, the complementarity is fully complementary or substantially complementary, or about 50%, 60%, 70%, 80%, 90% and 100% complementary. Percent complementarity (e.g., the aforementioned about 50%, 60%, 70%, 80%, 90% and 100% complementary) represents the percentage of residues in a nucleic acid molecule that can form hydrogen bonds (i.e., Watson-Crick base pairing) with another nucleic acid molecule. "Fully complementary" refers to that all continuous residues of a nucleic acid sequence form hydrogen bonds with the same number of continuous residues in a second nucleic acid sequence. "Substantially complementary" refers to a degree of complementarity of at least about 70%, 75%, 80%, 85%, 90%, 95%, 97%, 98%, 99% or 100% over a region of about 40, 50, 60, 70, 80, 100, 150, 200, 250 or more nucleotides, or refers to two nucleic acids that hybridize under stringent conditions. For a single base or a single nucleotide, according to the Watson-Crick base pairing rule, when A and T or U, C and G or I pair, they are called complementary or matching, and vice versa; and base pairings other than this are called non-complementary or non-matching.
[0152] In some embodiments, the methylation detection marker is a methylation detection marker for detecting a tumor or cancer. In some embodiments, the methylation detection marker is a methylation detection marker for detecting esophageal cancer. In some embodiments, the methylation detection marker is used to detect cfDNA or a blood sample. In some embodiments, the blood sample is whole blood, plasma or serum. In some embodiments, the methylation detection marker is selected from any two, more or all of the gene sites or their corresponding nucleic acid sequence fragments in the following table:
[0153]
[0154]
[0155] It should be understood that the present application includes various aspects, embodiments and combinations of the aspects and / or embodiments described herein. The above description and subsequent examples are intended to illustrate rather than limit the scope of the present application. Within the technical concept of the present application, the technical solution of the present application can be subjected to a variety of simple modifications, including that each technical feature is combined in any other suitable manner, and these simple modifications and combinations should also be regarded as the contents disclosed in the present application, and all belong to the protection scope of the present application.
[0156] The practice of the present application will employ, unless otherwise indicated, conventional techniques of cell biology, cell culture, molecular biology, transgenic biology, microbiology, recombinant DNA, and immunology. The conventional techniques are described in the prior art literature, see, for example, Current Protocols in Molecular Biology (Frederick M. AUSUBEL, 2000, Wiley and son Inc, Library of Congress, USA); Molecular Cloning: A Laboratory Manual, Third Edition, (Sambrook et al, 2001, Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press); Oligonucleotide Synthesis (MJ Gaited., 1984); Mullis et al. US Pat. No. 4, 683, 195; Nucleic Acid Hybridization (BD Harries & S. J. Higgins eds. 1984); Transcription And Translation (BD Hames & S. J. Higgins eds. 1984); Culture Of Animal Cells (RI Freshney, Alan R. Liss, Inc., 1987); Immobilized Cells And Enzymes (IRL Press, 1986); B. Perbal, A Practical Guide To Molecular Cloning (1984); the series, Methods In ENZYMOLOGY (J. Abelson and M. Simon, eds.-in-chief, Academic Press, Inc., New York), Vols. 154 and 155 (Wuetal. eds.) and Vol. 185, "Gene Expression Technology" (D. Goeddel, ed.); GeneTransfer Vectors For Mammalian Cells (JHMiller and MPCaloseds., 1987, Cold Spring Harbor Laboratory); Immunochemical Methods In Cell And Molecular Biology (Mayer and Walker, eds., Academic Press, London, 1987); Hand book Of Experimental Immunology, Volumes I-IV (DMWeir and CC Blackwell, eds., 1986); and Manipulating the Mouse Embryo (Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, 1986). .
[0157] Example
[0158] The inventors used epigenomic and bioinformatics techniques, and through the above feature extraction scheme, analyzed high-throughput methylation data of cancers such as esophageal cancer, and found a methylation pattern associated with cancer (e.g., esophageal cancer). Through the target sequence of this methylated gene, the methylation state of the gene can be sensitively and specifically detected, so that it can be used for the detection of peripheral blood free DNA.
[0159] Example 1 Identification of candidate intervals
[0160] (1) Obtain the public data GSE186458, which contains WGBS (whole genome bisulfite sequencing) data of purified cell types from various major tissues and organs, and select 39 common major cell types to ensure that there is at least one representative cell type in each major organ of the human body, with an average of 3 biological replicates for each cell type;
[0161] (2) Using the secondary data provided in the dataset (including the beta value of the methylation level of each CpG site in each sample), an internal self-built process was used to identify HMB intervals, so that the methylation level of each HMB interval between each CpG site in each cell type is uniform, that is, either uniformly high methylation or uniformly low methylation. A total of 7,264,350 HMB intervals were identified, with an average length of 224 bp and an average of 3.8 CpG sites;
[0162] (3) Based on the HMB intervals identified in the previous step, the intervals containing at least 4 CpG sites were screened, and the intervals whose methylation levels had a certain degree of difference in various cell types were further screened, resulting in 21K HMB intervals with a total length of 14.5 Mb;
[0163] (4) Based on the intervals obtained in (3), Twist designed and synthesized probes to obtain a panel for specifically capturing and enriching cfDNA molecules in the candidate intervals, with a ratio of 92% of the effectively designed intervals;
[0164] Example 2 Quantitative URA eigenvalues of candidate intervals
[0165] (1) A 10 ml blood sample was drawn from the test subject, and a cfDNA sample was extracted from the test sample. Panel capture and methylation library sequencing were then performed to obtain sequence reads from the cfDNA sample. An average of 243M reads were obtained from each sample sequencing.
[0166] (2) Through a series of basic downstream analyses, the alignment results and methylation site information of the reads were obtained. First, the fastp (v0.23.2) quality control software was used to check the sequencing quality and remove low-quality sequencing bases or reads. Then, the bismark (v0.23.0) alignment software was used to align the quality-controlled data to the reference genome hg19 and remove sequencing duplications to obtain the deduplicated alignment results (bam file), which contained detailed alignment results of the reads and methylation information of each site on the reads.
[0167] (3) Since the alignment results obtained in the previous step include detailed alignment results of reads and methylation information of each site on the reads, further analysis is required to obtain the reads falling within the candidate interval and the methylation status information of the CpG sites therein, which is called readMeth. The internal self-built analysis process is used to complete the task of extracting readMeth information. The extraction results are as follows:
[0168] 1_10484_10572HWY5HDSX2110483,10488,10492,10496,10524,10541,10562,10570Z,Z,Z,Z,Z,Z,Z,Z
[0169] From left to right, each element corresponds to the HMB interval index, read index, CpG site coordinates and the methylation coding status of each site. In the methylation coding status element of each site, "Z" and "z" correspond to the methylation status and unmethylation status respectively;
[0170] (4) Identify UXM reads. For each candidate interval, only retain the reads that contain at least 4 CpG sites. Calculate the methylation level u of each read, which is the proportion of methylated CpG sites in all CpG sites, i.e., u = C / (C + T), where C is the number of methylated CpG sites and T is the number of unmethylated CpG sites. If u ≤ 1 / 4, classify this read as a U read (mostly unmethylated read); if u ≥ 3 / 4, classify this read as an M read (mostly methylated read); otherwise, if 1 / 4 < u < 3 / 4, classify this read as an X read (mixed methylated read). For example, for the read in (3), u = 1 belongs to an M read;
[0171] (5) For each candidate interval, calculate URA, which is the proportion of the number of U reads in all reads within this interval, i.e., U / (U + X + M). To improve the accuracy of URA quantification, set the minimum effective number of reads to 50, i.e., require U + X + M ≥ 50. For example, for a certain HMB interval of a certain sample, the numbers of U, X, and M reads are 3, 1, and 84 respectively, then the value of URA is 0.0341;
[0172] Example 3 Feature Interval Screening and Modeling Evaluation
[0173] (1) Divide 84 healthy human plasma samples and 43 esophageal cancer plasma samples into a training set and a test set according to a certain ratio, requiring the numbers of healthy human samples and cancer samples in the training set to be close, that is, to ensure sample balance in the training set. The training set contains 30 healthy human plasma samples and 30 esophageal cancer plasma samples, and the test set contains 54 healthy human plasma samples and 13 esophageal cancer plasma samples;
[0174] (2) In the training set, two groups of candidate intervals with discriminative URA features were selected. The screening method is that for each candidate interval, the cancer sample is set as the positive group and the healthy person sample is set as the negative group. Under the premise of setting the specificity of the negative group at 0.9, the sensitivity of the positive group is calculated. The calculation method is that if the median of the negative group is smaller than that of the positive group, the 90% quantile of the negative group is calculated as the threshold, the value of the positive group is compared with the threshold, and the proportion of the number of positive groups with values not lower than the threshold in all positive groups is the sensitivity value; if the median of the negative group is larger than that of the positive group, the 10% quantile of the negative group is calculated as the threshold, the value of the positive group is compared with the threshold, and the proportion of the number of positive groups with values not greater than the threshold in all positive groups is the sensitivity value. The 20 candidate intervals with the largest sensitivity values are selected as the feature intervals for post-selection modeling. The detailed candidate intervals, i.e., the markers for methylation detection, are shown in Table 1 below.
[0175] Table 1:
[0176] Serial number chromosome Starting position End position Target genes 1 chr1 224804716 224805088 CNIH3 2 chr2 38302118 38302190 CYP1B1 3 chr3 27766032 27766367 EOMES 4 chr4 113432549 113432657 NEUROG2 5 chr4 134073584 134073906 PCDH10 6 chr5 3594738 3595978 IRX1 7 chr5 16179012 16179268 MARCHF11 8 chr6 100055731 100056756 PRDM13 9 chr6 105388800 105388961 LIN28B-AS1 10 chr10 22623460 22623607 SPAG6 11 chr11 20181456 20182036 DBX1 12 chr12 113917865 113918236 LHX5 13 chr13 28500781 28501299 PDX1 14 chr13 53775490 53775596 OLFM4 15 chr14 29237294 29237532 FOXG1 16 chr14 37132008 37132598 PAX9 17 chr14 60977414 60977962 SIX6 18 chr18 13868635 13869240 MC5R 19 chr18 25755415 25756043 CDH2 20 chr19 54481832 54482102 MIR935
[0177] (3) Modeling evaluation. The URA values of the 20 feature intervals of the training set samples were used as input, and a binary classification discriminant model was constructed using the random forest algorithm. The test set samples were predicted and the model prediction performance was evaluated. The evaluation results showed that the model had good discrimination between healthy people and esophageal cancer patients. The AUCs in the training set and test set were 1 and 0.976 (>0.85, 0.75), respectively. The ROC curves are shown in Figure 2. Figure 1 .
Claims
1. A method for screening markers for methylation detection, comprising: (1) The step of determining the candidate area comprises: Acquire first target data based on tissue, wherein the first target data is methylation sequencing data, which includes methylation level information of each CpG site in the tissue cell type; Determine a candidate region based on the first target data, wherein the candidate region includes more than x CpG sites and the methylation levels of all CpG sites in the candidate region in the tissue cell type are comparable; (2) The step of selecting a marker from the candidate region comprises: Obtaining second target data of each candidate region in each sample of the first population and the second population, wherein the second target data is data obtained after statistical calculation based on the methylation sequencing data of each candidate region in each sample, The candidate regions of the second target data that differ to a certain extent between the first population and the second population are used as markers based on methylation detection. The first population and the second population are two groups of people with different characteristics that do not overlap.
2. The method according to claim 1, wherein the value of x is 3 to 6, preferably 4.
3. The method according to claim 1, in the step of selecting a marker from a candidate region, For each candidate region, obtain z methylation sequencing reads with different sequences from each sample, and each read contains data of at least y CpG sites, and obtain the methylation level of each read; For z reads of each candidate region of each sample, the ratio of the number of reads in which the methylation level is within a specific numerical range to z is calculated, which is the second target data of the candidate region of the sample.
4. A method according to claim 3, wherein y=x.
5. The method according to claim 3, wherein the methylation level of the read is the ratio of the number of methylated CpG sites in the read to the number of all CpG sites in the read.
6. The method according to claim 3, wherein the z reads cover 50% to 100% of the CpG sites in the candidate region, preferably at least 75% of the CpG sites.
7. The method according to claim 3, wherein z is at least 30 to 50.
8. A method for constructing a model for distinguishing a first population from a second population, comprising: Obtaining methylation sequencing statistics of markers from samples of the first population and the second population; Using the statistical data as input to construct a binary classification model; in, The markers are: A marker obtained by the method of any one of claims 1 to 7, or Two or more or all of the markers selected from the following table:
9. A probe composition comprising probes targeting one or more markers selected from the following table:
10. A methylation sequencing composition comprising primers targeting one or more markers selected from the following table:
