Lung cancer methylation marker combination based on machine learning, construction method and application of risk model and lung cancer diagnosis system

Abstract

The invention relates to the field of molecular biomedicine, and discloses a lung cancer methylation marker combination based on machine learning, a risk model construction method, application and a lung cancer diagnosis system. The marker comprises one or a combination of more of the following gene segments: Shox2, HOXD9, OTX2, CLEC14A, CDH13, PAX3, OSR1, ZIC4, FOXI2, DMRTA2, SOD3, IRX1 and TBR1. The invention further discloses a method for preparing the marker. According to the method, three groups of different methylation markers are obtained by using a machine learning method based on different risk crowds, the three groups of models based on different methylation marker combinations realize coverage from low-risk crowds to high-risk crowds, screening efficiency can be greatly improved accurately for different risk crowds, and the accuracy of screening is improved. And meanwhile, the three groups of models are excellent in performance and have wide application prospects and high industrial value.

Description

Technical Field

[0001] The present application relates to the field of molecular biomedicine, and more specifically, to a lung cancer methylation marker combination based on machine learning, a method for constructing a risk model, its application, and a lung cancer diagnosis system. Background Art

[0002] According to data released by GLOBOCAN2020, approximately 2.2 million people are newly diagnosed with lung cancer each year worldwide, and 1.76 million people die from the disease. Early detection and treatment of lung cancer can significantly improve the survival rate of patients. According to the 9th edition of the AJCC (American Joint Committee on Cancer) Tumor Staging Manual in 2023, for non-small cell lung cancer, the 5-year survival rate of stage IA / IB patients can reach 82% / 69%; the 5-year survival rate of stage II patients is about 54% to 62%; the 5-year survival rate of stage III patients drops to about 30%, and the 5-year survival rate of stage IV patients is only less than 20%, among which the median survival time of stage IVB patients is less than 1 year.

[0003] Traditional lung cancer screening methods include imaging examinations, tumor markers, sputum cytology, etc. Imaging examinations are mainly LDCT, which has the advantages of high sensitivity and wide application range, but CT examinations will bring ionizing radiation, a high false positive rate, and high requirements for the level of equipment and film readers. Tumor markers mainly include CEA (carcinoembryonic antigen), NSE (neuron-specific enolase), etc., but their sensitivity and specificity are limited, making them difficult to use for large-scale screening. Sputum cytology is a method of screening for lung cancer by examining the morphology of sputum cells. Although this method is easy to operate and has high specificity, its sensitivity is too low and it has certain application value only in some lung cancers with central or diffuse distribution in imaging. .

[0004] Liquid biopsy based on cfDNA methylation is a new type of early screening method for lung cancer. Compared with traditional screening methods, it has the following advantages: 1. Non-invasive and less invasive: It is only tested through the patient's peripheral blood sample, without the need for invasive examination and without additional radiation; 2. High sensitivity: It can detect changes in the methylation level of related markers in very early lung cancer patients; 3. Convenient and easy: It is simple and fast to operate, and does not require special equipment and professional skills. In summary, liquid biopsy based on cfDNA methylation provides new directions and possibilities for early diagnosis and personalized treatment of lung cancer.

[0005] Developing highly sensitive and specific products based on cfDNA methylation requires a large amount of methylation sequencing data to screen for methylation markers with excellent performance, as well as a rigorous machine learning model construction and verification process, which is a very challenging task. Summary of the invention

[0006] The purpose of this invention is to screen methylation markers with lung cancer signals, integrate these methylation markers using machine learning models, and develop lung cancer methylation early screening products suitable for different tumor risk groups. The development of this product aims to achieve the following goals:

[0007] 1. Early lung cancer detection: By detecting abnormal changes in DNA methylation, potential problems can be discovered in the early stages of tumor development, which helps to diagnose lung cancer at an early stage and improve the success rate of treatment.

[0008] 2. Personalized medicine: By analyzing the individual's DNA methylation pattern, methylation early screening products are expected to provide individuals with personalized tumor risk assessments, thereby guiding personalized prevention and treatment plans.

[0009] 3. Improve screening efficiency: Compared with traditional tumor screening methods, methylation early screening products can improve the accuracy and efficiency of screening and reduce misdiagnosis and missed diagnosis rates.

[0010] 4. Reduce medical burden: The cost of treating early lung cancer is relatively low. Early lung cancer detection through early screening products can reduce patients' medical expenses and reduce the burden on the medical system.

[0011] Based on the above invention objectives, the present application provides a lung cancer methylation marker combination based on machine learning, a method for constructing a risk model, an application and a lung cancer diagnosis system.

[0012] In a first aspect, the present application discloses the use of a reagent for detecting ctDNA methylation segment markers in the preparation of a lung cancer diagnosis product, wherein the markers include a combination of one or more of the following gene segments:

[0013] chr3:157814948-157824292; chr2:176987451-176987651; chr14:57267425-57277197; chr1 4:38724555-38724755;chr16:82660460-82660774;chr2:223163395-223163595;chr2:1955 6785-19556985;chr3:147109862-147110062;chr10:129534693-129534935;chr1:50884536 -50884746;chr4:24801872-24802072;chr5:3602162-3602362;chr2:162280397-162280597.

[0014] Furthermore, the markers include a combination of the following gene segments:

[0015] Combination 1: chr3:157814948-157824292; chr14:57267425-57277197; chr16:82660460-82660774; chr2:223163395-223163595; chr2:19556785-19556985; chr3:147109862-147110062; chr10:129534693-129534935; chr1:50884536-50884746; chr5:3602162-3602362;

[0016] Combination 2: chr3:157814948-157824292; chr14:57267425-57277197; chr14:38724555-38724755; chr2:223163395-223163595; chr2:19556785-19556985; chr3:147109862-147110062; chr10:129534693-129534935; chr1:50884536-50884746; chr5:3602162-3602362; chr2:162280397-162280597;

[0017] Combination three: chr3:157814948-157824292; chr2:176987451-176987651; chr14:38724555-38724755; chr16:82660460-82660774; chr1:50884536-50884746; chr4:24801872-24802072; chr5:3602162-3602362.

[0018] Furthermore, the reagents include the following combination of one or more primer pairs:

[0019] The primer pair used to detect the gene segment chr3:157814948-157824292 has a sequence as shown in SEQ ID NOs.1-2; the primer pair used to detect the gene segment chr2:176987451-176987651 has a sequence as shown in SEQ ID NOs.3-4; the primer pair used to detect the gene segment chr14:57267425-57277197 has a sequence as shown in SEQ ID NOs.5-6; the primer pair used to detect the gene segment chr14:38724555-38724755 has a sequence as shown in SEQ ID NOs.7-8; the primer pair used to detect the gene segment chr16:82660460-82660774 has a sequence as shown in SEQ ID NO.9~10; a primer pair for detecting gene segment chr2:223163395-223163595, whose sequence is shown in SEQ ID NO.11~12; a primer pair for detecting gene segment chr2:19556785-19556985, whose sequence is shown in SEQ ID NO.13~14; ​​a primer pair for detecting gene segment chr3:147109862-147110062, whose sequence is shown in SEQ ID NO.15~16; a primer pair for detecting gene segment chr10:129534693-129534935, whose sequence is shown in SEQ ID NO.17~18; a primer pair for detecting gene segment chr1:50884536-50884746, whose sequence is shown in SEQ ID NO. IDNO.19~20; the primer pair used to detect the gene segment chr4:24801872-24802072, its sequence is shown in SEQ IDNO.21~22; the primer pair used to detect the gene segment chr5:3602162-3602362, its sequence is shown in SEQ IDNO.23~24; the primer pair used to detect the gene segment chr2:162280397-162280597, its sequence is shown in SEQ IDNO.25~26.

[0020] Furthermore, the reagent also includes a combination of one or more of the following probes:

[0021] The probe for detecting gene segment chr3:157814948-157824292 has a sequence as shown in SEQ ID NO.27; the probe for detecting gene segment chr2:176987451-176987651 has a sequence as shown in SEQ ID NO.28; the probe for detecting gene segment chr14:57267425-57277197 has a sequence as shown in SEQ ID NO.29; the probe for detecting gene segment chr14:38724555-38724755 has a sequence as shown in SEQ ID NO.30; the probe for detecting gene segment chr16:82660460-82660774 has a sequence as shown in SEQ ID NO.31; the probe for detecting gene segment chr2:223163395-223163595 has a sequence as shown in SEQ ID NO.32; a probe for detecting gene segment chr2:19556785-19556985, whose sequence is shown in SEQ ID NO.33; a probe for detecting gene segment chr3:147109862-147110062, whose sequence is shown in SEQ ID NO.34; a probe for detecting gene segment chr10:129534693-129534935, whose sequence is shown in SEQ ID NO.35; a probe for detecting gene segment chr1:50884536-50884746, whose sequence is shown in SEQ ID NO.36; a probe for detecting gene segment chr4:24801872-24802072, whose sequence is shown in SEQ ID NO.37; a probe for detecting gene segment chr5:3602162-3602362, whose sequence is shown in SEQ ID NO.38; the probe used to detect gene segment chr2:162280397-162280597, its sequence is shown in SEQ ID NO.39.

[0022] In a second aspect, the present application discloses a lung cancer diagnosis product, comprising the reagent for detecting ctDNA methylation segment markers as described in the first aspect above.

[0023] In a third aspect, the present application discloses a method for constructing a risk population screening model, comprising the following steps:

[0024] Step 1, collecting plasma samples for building a risk population screening model, and dividing the plasma samples into a training set and a test set;

[0025] Step 2, performing qPCR detection on the plasma samples collected in step 1 to obtain the Ct value of the methylation segment marker of each sample;

[0026] The methylation segment markers include: chr3: 157814948-157824292; chr2: 176987451-176987651; chr14: 57267425-57277197; chr14: 38724555-38724755; chr16: 82660460-82660774; chr2: 223163395-223163595; chr 2:19556785-19556985; chr3:147109862-147110062; chr10:129534693-129534935; chr1:50884 536-50884746;chr4:24801872-24802072;chr5:3602162-3602362;chr2:162280397-162280597;

[0027] Step 3: Use the Ct value obtained in step 2 to build a machine learning model. Use the logistic regression model to build a model in the training set data. The formula of the model is as follows: x is the Ct value vector of the target marker in the sample, w is the coefficient vector of different markers, b is the intercept value, and y is the model prediction score. After obtaining the score, the relevant threshold can be set according to the samples in the training set:

[0028]

[0029] Furthermore, in step 3, the target data used is the Ct value data of qPCR of marker combination 1, and a low-risk population screening model is obtained;

[0030] Combination 1: chr3:157814948-157824292; chr14:57267425-57277197; chr16:82660460-82660774; chr2:223163395-223163595; chr2:19556785-19556985; chr3:147109862-147110062; chr10:129534693-129534935; chr1:50884536-50884746; chr5:3602162-3602362.

[0031] Alternatively, in step 3, the target data used is the Ct value data of qPCR of marker combination 2, and a screening model for a moderate-risk population is obtained;

[0032] Combination 2: chr3:157814948-157824292; chr14:57267425-57277197; chr14:38724555-38724755; chr2:223163395-223163595; chr2:19556785-19556985; chr3:147109862-147110062; chr10:129534693-129534935; chr1:50884536-50884746; chr5:3602162-3602362; chr2:162280397-162280597.

[0033] Alternatively, in step 3, the target data used is the Ct value data of qPCR of marker combination three, and a high-risk population screening model is obtained;

[0034] Combination three: chr3:157814948-157824292; chr2:176987451-176987651; chr14:38724555-38724755; chr16:82660460-82660774; chr1:50884536-50884746; chr4:24801872-24802072; chr5:3602162-3602362.

[0035] In a fourth aspect, the present application discloses an application of a model obtained by the modeling method described in the third aspect in the preparation of a lung cancer diagnosis product.

[0036] In a fifth aspect, the present application discloses an application of using a model obtained by the modeling method described in the third aspect in combination with CT image data in the preparation of a lung cancer diagnosis product.

[0037] In a sixth aspect, the present application discloses a lung cancer diagnosis system, comprising:

[0038] Data module: used to obtain the Ct value data of qPCR of the methylation segment marker of the sample to be tested, wherein the methylation segment marker includes: chr3:157814948-157824292; chr2:176987451-176987651; chr14:57267425-57277197; chr14:38724555-38724755; chr16:82660460-82660774; chr2:223163 395-223163595;chr2:19556785-19556985;chr3:147109862-147110062;chr10:129534693-129534935;ch r1:50884536-50884746; chr4:24801872-24802072; chr5:3602162-3602362; chr2:162280397-162280597;

[0039] Scoring module: obtaining the score of the sample to be tested through the low-risk population screening model, the medium-risk population screening model or the high-risk population screening model described in the third aspect;

[0040] The low-risk population was scored using the low-risk population screening model;

[0041] The moderate-risk population is scored using the moderate-risk population screening model. The moderate-risk population refers to people with a long history of smoking, a family history of lung cancer, long-term exposure to polluted environments, and lung-related diseases such as lung inflammation;

[0042] The high-risk population is scored using the high-risk population screening model. The high-risk population refers to people who have been diagnosed with symptoms such as chronic lung inflammation, lung infection, lung nodules, fibrous foci, and calcification foci;

[0043] Judgment module: Use the threshold of the corresponding risk population screening model to judge the score of the sample to be tested. If it is greater than the threshold, it is judged as positive, and if it is less than the threshold, it is judged as negative.

[0044] The principles of tumor methylation marker detection, machine learning and model building of the present invention are:

[0045] Tumor cells often release some fragmented DNA into the blood, which is called circulating tumor DNA or ctDNA. These ctDNA contain methylation patterns associated with tumor cells. By detecting methylation markers in these ctDNA, signs of tumor presence can be found even when the tumor is small or has not yet caused symptoms. Blood methylation testing has the following advantages:

[0046] 1. Non-invasive: Blood methylation testing is a non-invasive method that only requires the collection of a small blood sample, without the need for invasive surgery or tissue biopsy. This reduces patient discomfort and risk, and also makes screening more acceptable.

[0047] 2. Easily repeatable: Blood samples are relatively easy to collect and can be repeated at different time points. This means that regular monitoring can be performed to detect any changes in methylation patterns early, which may indicate potential tumors.

[0048] 3. Multiple tumor types: Blood methylation testing can be used for many different types of tumors. Although different types of tumors have different methylation patterns, this method can screen for different methylation markers and patterns.

[0049] 4. High sensitivity and specificity: Blood methylation tests are usually highly sensitive and specific, and can detect methylation changes in even trace amounts of ctDNA. This helps improve the accuracy of early screening.

[0050] Machine learning aims to enable computer systems to learn from data, improve their performance, and perform specific tasks without being explicitly programmed. Machine learning systems are able to automatically identify and respond to patterns in data in order to make decisions or predict future events. In the field of methylation tumor early screening, machine learning can be used to select the most relevant DNA methylation markers and can identify specific patterns in DNA methylation, helping to achieve more accurate, efficient, and personalized methylation tumor early screening and improve patients' survival rate and quality of life.

[0051] In summary, this application has the following beneficial effects:

[0052] The present invention obtains three groups of different methylation markers using machine learning methods based on different risk groups. These three groups of models based on different combinations of methylation markers achieve coverage from low-risk groups to high-risk groups, accurately targeting different risk groups and greatly improving the efficiency of screening. At the same time, the three groups of models have excellent performance, broad application prospects and high industrial value. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 : ROC curve of the low-risk screening model of Example 2 of the present application;

[0054] Figure 2 : The low-risk screening model score distribution of Example 2 of the present application;

[0055] Figure 3 : ROC curve of the medium-risk screening model of Example 3 of the present application;

[0056] Figure 4 : The medium risk screening model score distribution of Example 3 of the present application;

[0057] Figure 5 : ROC curve of the high-risk screening model of Example 4 of the present application;

[0058] Figure 6 : The high-risk screening model score distribution of Example 4 of the present application;

[0059] Figure 7 : The risk diagnosis model of Example 5 of the present application is combined with CT image data. DETAILED DESCRIPTION

[0060] The technical scheme and effects of the present application are further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the invention, rather than to limit the invention.

[0061] Example 1: Three groups of methylation marker combinations, primer and probe design

[0062] By querying gene annotation databases such as NCBI, Ensembl, and UCSC, the target gene and its upstream and downstream sequences within 5kb were obtained. Then, the CpG island-enriched regions were screened for primer and probe design. According to the design principles of methylation-specific PCR primers and probes, the C in the CpG island was considered to be methylated, while the C in the non-CpG region was not methylated. After obtaining the simulated bisulfite treatment sequence, conventional primer and probe design tools were used for design and chemical synthesis. The methylation marker PCR regions and specific primer and probe sequence information contained in the three different combinations are shown in Tables 1 and 2.

[0063] Table 1 PCR detection segments of methylation markers included in three different combinations

[0064]

[0065]

[0066] Table 2 Specific primers and probe sequences for methylation markers

[0067]

[0068] Example 2: Low-risk population screening model—methylation marker combination model

[0069] Plasma samples from 263 healthy people with no lung abnormalities and 284 lung cancer patients were collected and divided into two groups: training set and test set. The age of the training set samples was balanced between cancer patients and healthy controls to eliminate the interference of age factors.

[0070] Table 3. Relevant information of samples used in the combined model

[0071]

[0072] The collected plasma samples were subjected to the following experimental treatments. The specific steps are as follows:

[0073] S1, cfDNA was extracted from plasma samples using the commercial kit Qiagen QIAamp Circulating Nucleic Acid Kit. Then, the extracted cfDNA was subjected to sulfite conversion using the commercial kit MethylCodeTM Bisulfite Conversion Kit.

[0074] S2, pre-amplify the converted cfDNA, use the premix consisting of the target-specific primers and internal reference (ACTB) primers listed in Table 2 as the primer pool, and perform PCR amplification at a final concentration of 100nM. The PCR reaction system contains 10μL converted DNA, 2.5μL premix and 12.5μL PCR reagent Probe qPCR Master Mix (NEB). PCR reaction conditions were: 95°C for 5 minutes, 95°C for 30 seconds, 56°C for 60 seconds, and 15 cycles.

[0075] S3, dilute the pre-amplification product 10 times and use fluorescent PCR for detection. Use the primer and probe sequences listed in Table 2 to detect the internal reference gene ACTB at the same time. In the premix, the final concentration of the primer is 500nM, and the final concentration of the probe is 200nM. The PCR reaction system contains 10μL pre-amplification dilution, 2.5μL primer and probe premix, and 12.5μL PCR reagent Probe qPCR Master Mix (NEB). PCR reaction conditions are: 95℃ for 5 minutes; 95℃ for 15 seconds, 56℃ for 40 seconds (collecting fluorescence), and 50 cycles. Since the fluorescence modification of different gene probes is different, it is necessary to select the corresponding detection fluorescence channel according to the fluorescence modification of different gene probes. If no amplification signal is detected, the Ct (Cycle threshold) value of the target point is set to 50.

[0076] The Ct value detected by the target is used to build a machine learning model. The detailed steps of modeling are as follows:

[0077] S1, the logistic regression model (LogisticRegression) in the sklearn (V1.0.1) package in python (V3.9.7) was used to model the training set data. The target data used were the qPCR Ct value data of the 9 genes (SHOX2, OTX2, CDH13, PAX3, OSR1, ZIC4, FOXI2, DMRTA2, IRX1) in combination 1 marker in Table 2. The formula of the model is as follows: where x is the Ct value vector of the target marker in the sample, w is the coefficient vector of different markers, b is the intercept value, and y is the model prediction score. After obtaining the score, the relevant threshold can be set according to the samples in the training set:

[0078]

[0079] S2, the model constructed in the training set is brought into the test set data for model evaluation. After obtaining the score of each sample in the test set, the threshold determined by the training set is used for judgment. If it is greater than the threshold, it is positive, and if it is less than the threshold, it is negative.

[0080] S3, AUC, sensitivity, specificity and other indicators of the statistical model in the training set and test set.

[0081] The ROC curve of the low-risk screening model is shown in Figure 1 The model achieved an AUC of 0.942 (95% CI, 0.929-0.954) in training, and achieved similar performance in an independent test set (AUC = 0.933 (95% CI, 0.922-0.947)). The model score distribution is shown in Figure 2 , we can see that the scores of lung cancer samples in both the training set and the test set are significantly higher than those of healthy control samples. In the training set, the threshold is set to 0.1559, and values ​​greater than this threshold are judged as positive, and values ​​less than this threshold are judged as negative. Then, at a specificity of 60.5%, a high sensitivity of 98.5% can be achieved; in the test set, using the same threshold as the training set, a sensitivity of 90.7% can be achieved at a high specificity of 73.9%. It can better distinguish between lung cancer patients and healthy control samples.

[0082] Example 3: Screening model for moderate-risk population—Methylation marker combination model II

[0083] The target population of the screening model for medium-risk population has been expanded to include benign control samples based on Table 3. Detailed sample information is shown in Table 4.

[0084] Table 4. Relevant information of samples used in combination 2 models

[0085]

[0086]

[0087] The experimental steps and machine learning modeling process of the moderate risk screening model are consistent with those in Example 2, but the combination of methylation markers is changed to the 10 genes of combination 2 in Table 2 (SHOX2, OTX2, CLEC14A, PAX3, OSR1, ZIC4, FOXI2, DMRTA2, IRX1, TBR1).

[0088] The ROC curve of the moderate risk screening model is shown in Figure 3 . In the training set, the AUC can reach 0.925 (95% CI, 0.913-0.939), and in the independent test set, a similar AUC (0.919 (95% CI, 0.905-0.934)) can be achieved. The lung cancer sample scores of this model in the training set and test set were significantly higher than those of the healthy and benign control samples ( Figure 4 ). In the training set, the threshold was set to 0.1650. Under this threshold, the specificity of all controls was 58.9% (the specificity of healthy controls was 60.5% and the specificity of benign controls was 55.6%), and a sensitivity of 98.5% could be achieved. This model can better distinguish lung cancer patients with higher specificity.

[0089] Example 4: High-risk population diagnostic model - methylation marker combination three models

[0090] The high-risk population diagnostic model modeling samples included the high-risk control samples (benign controls) and lung cancer samples in Table 4. The detailed experimental steps and modeling steps were consistent with Example 2, except that the combination of methylation markers was changed to the 7 genes of combination three in Table 3 (SHOX2, HOXD9, CLEC14A, CDH13, DMRTA2, SOD3, IRX1).

[0091] The ROC curve of the diagnostic model for high-risk population is shown in Figure 5 . In the training set, the AUC can reach 0.897 (95% CI, 0.876-0.920), and in the independent test set, a similar AUC (0.885 (95% CI, 0.860-0.913)) can be achieved. The lung cancer sample scores of this model in the training set and the test set were significantly higher than those of the benign control samples ( Figure 6 ). In the training set, the threshold was set to 0.4758. Under this threshold, the specificity of the benign control was 79.4%, and the sensitivity was 82.0%. This model can better distinguish high-risk people from lung cancer patients.

[0092] Example 5: Combined diagnosis of methylation model and imaging data

[0093] Compare the diagnostic results of chest CT images with the results of the model, such as Figure 7 As shown, the sensitivity and specificity of the methylation model are not lower than those of imaging diagnosis. After the two are used in combination, the overall sensitivity of lung cancer diagnosis is improved, especially the sensitivity of stage I / II lung cancer is significantly improved, which illustrates the complementarity of this patented model with existing imaging. The combination of the two can achieve better screening effects.

[0094] The present invention is based on blood, extracts cfDNA from blood plasma, and uses the primers in Example 1 to target the amplification of the target methylation markers, and then uses a qPCR instrument to measure the CT value of each marker after sulfite conversion. The CT value is subsequently substituted into the following three different models according to the risk level of the subject to be tested for tumor discrimination: Combination 1 (including 9 genes: SHOX2, OTX2, CDH13, PAX3, OSR1, ZIC4, FOXI2, DMRTA2, IRX1) model is used for lung cancer screening in low-risk populations; Combination 2 (including 10 genes: SHOX2, OTX2, CL The model of SHOX2, HOXD9, CLEC14A, CDH13, DMRTA2, SOD3, and IRX1 is used for lung cancer screening in medium-risk groups (such as those with a long history of smoking, a family history of lung cancer, long-term exposure to polluted environments, and lung-related diseases such as lung inflammation); combination three (including 7 genes: SHOX2, HOXD9, CLEC14A, CDH13, DMRTA2, SOD3, and IRX1) is used to identify high-risk groups, and to identify patients with malignant tumors from those who have been diagnosed with symptoms such as chronic lung inflammation, lung infection, lung nodules, fibrous foci, and calcification foci. Positive patients are recommended to undergo LDCT, biopsy puncture, and other methods for further verification.

[0095] In summary, the present invention uses a machine learning-based method for modeling and screens three different methylation marker combination models, covering the screening of all populations from low risk to high risk, and achieving relatively excellent performance. At the same time, the three models of the present invention show complementarity with existing imaging methods, and the combination of the two can achieve more excellent performance, which illustrates the broad application prospects and high industrial value of the invention.

[0096] This specific embodiment is merely an explanation of the present application and is not a limitation of the present application. After reading this specification, those skilled in the art may make modifications to the present embodiment without any creative contribution as needed. However, as long as it is within the scope of the claims of the present application, it shall be protected by the patent law.

Claims

1. Use of a reagent for detecting ctDNA methylation segment markers in the preparation of a lung cancer diagnosis product, characterized in that: The markers include one or more combinations of the following gene segments: chr3:157814948-157824292; chr2:176987451-176987651; chr14:57267425-57277197; chr1 4:38724555-38724755;chr16:82660460-82660774;chr2:223163395-223163595;chr2:1955 6785-19556985;chr3:147109862-147110062;chr10:129534693-129534935;chr1:50884536 -50884746;chr4:24801872-24802072;chr5:3602162-3602362;chr2:162280397-162280597.

2. The use according to claim 1, characterized in that: The markers include a combination of the following gene segments: Combination 1: chr3:157814948-157824292; chr14:57267425-57277197; chr16:82660460-82660774; chr2:223163395-223163595; chr2:19556785-19556985; chr3:147109862-147110062; chr10:129534693-129534935; chr1:50884536-50884746; chr5:3602162-3602362; Combination 2: chr3:157814948-157824292; chr14:57267425-57277197; chr14:38724555-38724755; chr2:223163395-223163595; chr2:19556785-19556985; chr3:147109862-147110062; chr10:129534693-129534935; chr1:50884536-50884746; chr5:3602162-3602362; chr2:162280397-162280597; Combination three: chr3:157814948-157824292; chr2:176987451-176987651; chr14:38724555-38724755; chr16:82660460-82660774; chr1:50884536-50884746; chr4:24801872-24802072; chr5:3602162-3602362.

3. The use according to claim 1, characterized in that: The reagents include the following combination of one or more primer pairs: The primer pair used to detect the gene segment chr3:157814948-157824292 has a sequence as shown in SEQ ID NOs.1-2; the primer pair used to detect the gene segment chr2:176987451-176987651 has a sequence as shown in SEQ ID NOs.3-4; the primer pair used to detect the gene segment chr14:57267425-57277197 has a sequence as shown in SEQ ID NOs.5-6; the primer pair used to detect the gene segment chr14:38724555-38724755 has a sequence as shown in SEQ ID NOs.7-8; the primer pair used to detect the gene segment chr16:82660460-82660774 has a sequence as shown in SEQ ID NO.9~10; a primer pair for detecting gene segment chr2:223163395-223163595, whose sequence is shown in SEQ ID NO.11~12; a primer pair for detecting gene segment chr2:19556785-19556985, whose sequence is shown in SEQ ID NO.13~14; ​​a primer pair for detecting gene segment chr3:147109862-147110062, whose sequence is shown in SEQ ID NO.15~16; a primer pair for detecting gene segment chr10:129534693-129534935, whose sequence is shown in SEQ ID NO.17~18; a primer pair for detecting gene segment chr1:50884536-50884746, whose sequence is shown in SEQ ID NO.19~20; the primer pair used to detect the gene segment chr4:24801872-24802072, its sequence is shown in SEQ ID NO.21~22; the primer pair used to detect the gene segment chr5:3602162-3602362, its sequence is shown in SEQ ID NO.23~24; the primer pair used to detect the gene segment chr2:162280397-162280597, its sequence is shown in SEQID NO.25~26.

4. The use according to claim 3, characterized in that: The reagents also include a combination of one or more of the following probes: The probe for detecting gene segment chr3:157814948-157824292 has a sequence as shown in SEQ ID NO.27; the probe for detecting gene segment chr2:176987451-176987651 has a sequence as shown in SEQ ID NO.28; the probe for detecting gene segment chr14:57267425-57277197 has a sequence as shown in SEQ ID NO.29; the probe for detecting gene segment chr14:38724555-38724755 has a sequence as shown in SEQ ID NO.30; the probe for detecting gene segment chr16:82660460-82660774 has a sequence as shown in SEQ ID NO.31; the probe for detecting gene segment chr2:223163395-223163595 has a sequence as shown in SEQ ID NO.32; a probe for detecting gene segment chr2:19556785-19556985, whose sequence is shown in SEQ ID NO.33; a probe for detecting gene segment chr3:147109862-147110062, whose sequence is shown in SEQ ID NO.34; a probe for detecting gene segment chr10:129534693-129534935, whose sequence is shown in SEQ ID NO.35; a probe for detecting gene segment chr1:50884536-50884746, whose sequence is shown in SEQ ID NO.36; a probe for detecting gene segment chr4:24801872-24802072, whose sequence is shown in SEQ ID NO.37; a probe for detecting gene segment chr5:3602162-3602362, whose sequence is shown in SEQ ID NO.38; the probe used to detect gene segment chr2:162280397-162280597, its sequence is shown in SEQ ID NO.

39.

5. A lung cancer diagnosis product, characterized in that: A reagent for detecting ctDNA methylation segment markers comprising any one of claims 1-4.

6. A method for constructing a risk population screening model, characterized in that: The following steps are involved: Step 1, collecting plasma samples for building a risk population screening model, and dividing the plasma samples into a training set and a test set; Step 2, performing qPCR detection on the plasma samples collected in step 1 to obtain the Ct value of the methylation segment marker of each sample; The methylation segment markers include: chr3: 157814948-157824292; chr2: 176987451-176987651; chr14: 57267425-57277197; chr14: 38724555-38724755; chr16: 82660460-82660774; chr2: 223163395-223163595; chr 2:19556785-19556985; chr3:147109862-147110062; chr10:129534693-129534935; chr1:50884 536-50884746;chr4:24801872-24802072;chr5:3602162-3602362;chr2:162280397-162280597; Step 3: Use the Ct value obtained in step 2 to build a machine learning model. Use the logistic regression model to build a model in the training set data. The formula of the model is as follows: x is the Ct value vector of the target marker in the sample, w is the coefficient vector of different markers, b is the intercept value, and y is the model prediction score. After obtaining the score, the relevant threshold can be set according to the samples in the training set:

7. The construction method according to claim 6, characterized in that: In step 3, the target data used is the Ct value data of qPCR of marker combination 1, and a low-risk population screening model is obtained; Combination 1: chr3:157814948-157824292; chr14:57267425-57277197; chr16:82660460-82660774; chr2:223163395-223163595; chr2:19556785-19556985; chr3:147109862-147110062; chr10:129534693-129534935; chr1:50884536-50884746; chr5:3602162-3602362; Alternatively, in step 3, the target data used is the Ct value data of qPCR of marker combination 2, and a screening model for a moderate-risk population is obtained; Combination 2: chr3:157814948-157824292; chr14:57267425-57277197; chr14:38724555-38724755; chr2:223163395-223163595; chr2:19556785-19556985; chr3:147109862-147110062; chr10:129534693-129534935; chr1:50884536-50884746; chr5:3602162-3602362; chr2:162280397-162280597; Alternatively, in step 3, the target data used is the Ct value data of qPCR of marker combination three, and a high-risk population screening model is obtained; Combination three: chr3:157814948-157824292; chr2:176987451-176987651; chr14:38724555-38724755; chr16:82660460-82660774; chr1:50884536-50884746; chr4:24801872-24802072; chr5:3602162-3602362.

8. Use of the model obtained by the modeling method according to claim 6 or 7 in the preparation of lung cancer diagnosis products.

9. Use of the model obtained by the modeling method according to claim 6 or 7 in combination with CT image data in the preparation of lung cancer diagnosis products.

10. A lung cancer diagnosis system, characterized in that: include: Data module: used to obtain the Ct value data of qPCR of the methylation segment marker of the sample to be tested, wherein the methylation segment marker includes: chr3:157814948-157824292; chr2:176987451-176987651; chr14:57267425-57277197; chr14:38724555-38724755; chr16:82660460-82660774; chr2:223163 395-223163595;chr2:19556785-19556985;chr3:147109862-147110062;chr10:129534693-129534935;ch r1:50884536-50884746; chr4:24801872-24802072; chr5:3602162-3602362; chr2:162280397-162280597; Scoring module: obtaining the score of the sample to be tested by the low-risk population screening model, the medium-risk population screening model or the high-risk population screening model described in claim 7; The low-risk population was scored using the low-risk population screening model; The moderate-risk population is scored using the moderate-risk population screening model. The moderate-risk population refers to people with a long history of smoking, a family history of lung cancer, long-term exposure to polluted environments, and lung-related diseases such as lung inflammation; The high-risk population is scored using the high-risk population screening model. The high-risk population refers to people who have been diagnosed with symptoms such as chronic lung inflammation, lung infection, lung nodules, fibrous foci, and calcification foci; Judgment module: Use the threshold of the corresponding risk population screening model to judge the score of the sample to be tested. If it is greater than the threshold, it is judged as positive, and if it is less than the threshold, it is judged as negative.

Images