Lung nodule diagnosis method based on plasma DNA methylation characteristics and application
By constructing the PulmoSeek model based on plasma DNA methylation characteristics, the problem of high detection rate and low diagnostic accuracy in the diagnosis of pulmonary nodules has been solved. This has enabled non-invasive and accurate differentiation between benign and malignant nodules, reducing the frequency and cost of unnecessary examinations.
Patent Information
- Application Number
- CN202610016202.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-07
- Publication Date
- 2026-04-21
AI Technical Summary
Existing technologies present a contradiction between high detection rates and low diagnostic rates in the diagnosis of pulmonary nodules. Current methods cannot accurately, conveniently, and non-invasively distinguish between benign and malignant nodules, resulting in a large number of patients with benign nodules suffering unnecessary radiation exposure and economic and psychological burdens. Furthermore, existing liquid biopsy techniques have insufficient sensitivity in early-stage lung cancer.
By collecting peripheral blood from subjects, cfDNA was extracted using magnetic bead extraction or column extraction. After bisulfite conversion treatment, a PulmoSeek classification diagnostic model was constructed by combining a customized targeted sequencing panel and an autoencoder neural network. The plasma DNA methylation characteristics were analyzed, and a gradient boosting decision tree was used for supervised learning to output a malignancy risk score.
It achieves non-invasive diagnosis with high sensitivity and high negative predictive value, effectively avoiding unnecessary invasive examinations, improving the accuracy of early lung cancer detection and the universality of nodule types, and reducing medical costs and psychological trauma to patients.
Smart Images

Figure CN121905485A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of pulmonary nodule diagnosis technology, and particularly relates to a method and application for pulmonary nodule diagnosis based on plasma DNA methylation characteristics. Background Technology
[0002] Lung cancer is the leading cause of cancer-related death worldwide, and its prognosis is closely related to the disease stage at diagnosis. Data shows that the 5-year survival rate for early-stage (e.g., stage IA) non-small cell lung cancer patients can reach over 85%, while the rate for late-stage (stage IV) patients drops sharply to around 6%, highlighting the decisive significance of early detection and diagnosis in saving lives and improving prognosis. Against this backdrop, low-dose CT screening has been widely adopted and promoted due to its effectiveness in reducing lung cancer mortality in high-risk populations. However, the widespread use of this technology has also brought a significant clinical dilemma: while increasing the detection rate, it also identifies a large number of incidentally discovered lung nodules, the vast majority of which (approximately 80%-90%) are subsequently confirmed to be benign. This contradiction between "high detection rate" and "low diagnostic rate" has created a huge clinical management challenge. A large number of patients carrying benign nodules are involved in long-term imaging follow-ups, suffering unnecessary radiation exposure, continuous economic and psychological burdens, and facing the risk of overtreatment due to diagnostic uncertainty. Therefore, clinical practice urgently needs an auxiliary diagnostic tool that can accurately, conveniently, and non-invasively distinguish between benign and malignant nodules in order to optimize decision-making processes and achieve precise management.
[0003] Liquid biopsy, especially based on the analysis of circulating tumor DNA in the blood, offers a promising solution to this challenge. It achieves non-invasive "molecular insight" by capturing tumor-derived information released into the bloodstream. Among various ctDNA biomarkers, DNA methylation, as a stable epigenetic modification, shows abnormalities in the very early stages of lung cancer development and progression. It also exhibits high tissue specificity and relatively low background noise, making it an ideal target for achieving highly sensitive liquid biopsies in early-stage lung cancer.
[0004] Currently, while various methods exist for assessing the benign or malignant nature of pulmonary nodules in clinical practice, all have significant limitations. The most widely used clinical prediction models, which integrate limited variables such as age, smoking history, and nodule size, have limited discriminative ability, particularly in differentiating early lesions with atypical presentations, such as subsolid nodules. PET-CT functional imaging, as an important supplementary method, has high value in assessing solid nodules, but its sensitivity decreases significantly for nodules predominantly composed of ground-glass opacities, and it also suffers from radiation exposure and high costs. Traditional serum tumor markers have extremely low positive rates in early-stage lung cancer, limiting their diagnostic value. Detection technologies based on ctDNA somatic mutations are limited by the fundamental bottleneck of extremely low ctDNA abundance in early-stage lung cancer, making it difficult to improve sensitivity. Newly explored multi-omics integration strategies (such as combining multiple blood markers with PET-CT) demonstrate conceptual potential, but their complex procedures, high costs, and potential for false positives in early stages make them unsuitable for large-scale population screening and triage. In summary, existing technologies are insufficient in terms of accuracy, universality, accessibility, and cost-effectiveness, and cannot fully meet the urgent clinical need for accurate and non-invasive differential diagnosis of pulmonary nodules, especially early and atypical nodules. This creates clear space for technological improvement and necessity for the application of this invention. Summary of the Invention
[0005] The purpose of this invention is to provide a diagnostic method and application for pulmonary nodules based on plasma DNA methylation characteristics, in order to solve the technical problems mentioned in the background art.
[0006] To achieve the above objectives, the specific technical solution of the present invention is as follows: A method for diagnosing pulmonary nodules based on plasma DNA methylation characteristics, comprising the following steps:
[0007] S1. Peripheral blood of the subject was collected in a special cfDNA preservation tube, and plasma was obtained by centrifugation. cfDNA was extracted using an optimized magnetic bead method or column extraction method and quality controlled. Subsequently, the cfDNA was subjected to a complete bisulfite conversion treatment to convert unmethylated cytosine to uracil, while methylated cytosine remained unchanged.
[0008] S2. The transformed DNA was used for library construction and target region enrichment using a customized targeted sequencing panel;
[0009] S3. After obtaining the sequencing data, an innovative bioinformatics and machine learning analysis pipeline is used. After the raw data is removed by adapters and quality filtered, it is aligned to the reference genome that has undergone bisulfite conversion. Then, the initial methylation level matrix is constructed by calculating the methylation ratio of each target CpG site.
[0010] S4. Using the gradient boosting decision tree learning algorithm, and then supervising learning on a training set consisting of a large number of samples diagnosed by the gold standard of pathology, a classification and diagnostic model called "PulmoSeek" was constructed. Through the recursive feature elimination model optimization technique, a set of core methylation features were determined as the cornerstone of the model's judgment.
[0011] Preferably, the quality control in step S1 includes concentration determination and integrity assessment.
[0012] Preferably, the concentration measured is not less than 5 ng.
[0013] Preferably, in step S2, the panel covers 12,899 screened and validated specific methylation regions associated with lung cancer, involving a total of 105,844 CpG sites.
[0014] Preferably, the enriched library in step S2 is subjected to high-throughput sequencing on a next-generation sequencing platform to generate raw data.
[0015] Preferably, step S3 employs a feature engineering strategy, including the following steps:
[0016] First, by identifying differentially methylated regions and calculating the proportion of "comethylation reads" within these regions, an initial feature set is constructed. Then, an unsupervised deep learning model, an autoencoder neural network, is introduced to perform nonlinear dimensionality reduction and feature abstraction on the high-dimensional initial features, automatically learning and extracting low-dimensional, high-information feature vectors that can represent the essence of complex methylation patterns.
[0017] This invention also relates to the application of a diagnostic method for pulmonary nodules based on plasma DNA methylation characteristics, comprising the following steps:
[0018] After the patient's sample undergoes the above process, the resulting methylation data is input into the trained PulmoSeek model. The model outputs a continuous malignancy risk score between 0 and 1. By using statistical methods such as the Youden index on a large-scale validation set to determine an optimal decision threshold, this score can be transformed into a clear clinical interpretation: a score above the threshold is considered "suspected malignancy," and further clinical diagnostic intervention is strongly recommended; a score below the threshold is considered "possibly benign," with a very high negative predictive value. Clinicians can then adopt a safer and more proactive follow-up strategy for the patient, thereby effectively avoiding unnecessary invasive examinations.
[0019] The diagnostic method and application of pulmonary nodules based on plasma DNA methylation characteristics of the present invention have the following advantages:
[0020] 1. This invention originates from the selection of DNA methylation, an early and stable epigenetic marker, to replace traditionally relied-upon clinical parameters or late-stage mutation signals. This directly solves the problem of insufficient discrimination ability (AUC < 0.60) in existing clinical prediction models (such as the Mayo Clinic model) due to limited input information. By analyzing a set of specific methylation feature combinations in plasma cfDNA that have undergone large-scale screening and validation, and using an autoencoder neural network for in-depth feature abstraction and dimensionality reduction, the classification model constructed in this invention can efficiently extract lung cancer-specific signals from complex background noise, thereby achieving an AUC value as high as 0.843 in independent validation, demonstrating a significantly better overall discrimination accuracy than traditional models.
[0021] 2. This invention prioritizes achieving extremely high sensitivity (0.99), directly addressing the urgent clinical need to avoid missed diagnoses of early-stage lung cancer. Combined with its validation performance in high-risk populations, its negative predictive value reaches 0.997 at a clinically relevant prevalence setting of 10%. This characteristic makes it an extremely reliable "exclusionary" diagnostic tool. In terms of application procedures, this means that for most benign nodules detected by LDCT screening, this test can provide a high-confidence "likely benign" result, potentially safely avoiding over 40% of unnecessary invasive examinations (such as biopsies or surgeries), effectively solving the problem of overdiagnosis caused by the high false-positive rate of LDCT.
[0022] 3. This invention also overcomes a key bottleneck in liquid biopsy for early cancer detection. By employing methylation markers that undergo widespread alteration in the early stages of tumor development, rather than somatic mutations with extremely low abundance, this invention achieves a sensitivity of up to 0.971 for stage I lung cancer, successfully enabling accurate detection of extremely low tumor burden. Simultaneously, this invention overcomes the limitations of existing imaging techniques (such as PET-CT) regarding nodule types. Since methylation signals originate from the intrinsic biological characteristics of tumor cells, rather than morphological or metabolic manifestations, this approach maintains excellent and stable high sensitivity (1.000, 0.947, and 0.964, respectively) for solid, partially solid, and pure ground-glass nodules, providing unprecedented universality across nodule types. Attached Figure Description
[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.
[0024] Figure 1 This is a flowchart of the model construction process in this invention;
[0025] Figure 2 This is a schematic diagram comparing the performance of the PulmoSeek and Mayo Clinic / VA models across all nodule sizes in this invention.
[0026] Figure 3 This is a schematic diagram comparing the performance of the PulmoSeek and Mayo Clinic / VA models in 6-20mm nodules in this invention;
[0027] Figure 4 This is a schematic diagram illustrating the performance of PulmoSeek in early-stage lung cancer in this invention.
[0028] Figure 5 This is a schematic diagram illustrating the performance of PulmoSeek in different types of nodules and its comparison with PET-CT. Detailed Implementation
[0029] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the embodiments of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.
[0030] In the description of the embodiments of the present invention, it should be understood that the terms "length", "vertical", "horizontal", "top", "bottom", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings. They are only for the convenience of describing the embodiments of the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of the present invention.
[0031] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of embodiments of the present invention, "a plurality of" means two or more, unless otherwise explicitly specified.
[0032] In this embodiment of the invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection, an electrical connection, or a communication connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this embodiment of the invention according to the specific circumstances.
[0033] The following disclosure provides many different implementations or examples for carrying out different structures of the embodiments of the present invention. To simplify the disclosure of the embodiments of the present invention, specific examples of components and arrangements are described below. Of course, these are merely examples and are not intended to limit the embodiments of the present invention. Furthermore, reference numerals and / or reference letters may be repeated in different examples of the embodiments of the present invention; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various implementations and / or arrangements discussed.
[0034] To better understand the purpose, structure, and function of this invention, the following detailed description, in conjunction with the accompanying drawings, provides a method for diagnosing pulmonary nodules based on plasma DNA methylation characteristics and its application.
[0035] like Figure 1-5 As shown, this invention provides a method for diagnosing pulmonary nodules based on plasma DNA methylation characteristics. First, peripheral blood from the subject must be collected in a dedicated cfDNA storage tube to ensure nucleic acid stability. After centrifugation to obtain plasma, cfDNA is extracted using an optimized magnetic bead or column extraction method, and subjected to strict quality control, including concentration determination (typically requiring no less than 5 ng) and integrity assessment. Subsequently, the cfDNA undergoes thorough bisulfite conversion, converting unmethylated cytosine to uracil, while methylated cytosine remains unchanged. This is a crucial step for subsequent differentiation of methylation status.
[0036] The transformed DNA was used for library construction and target region enrichment using a meticulously designed and validated custom-designed targeted sequencing panel. The core value of this panel lies in its coverage of 12,899 specific methylation regions highly associated with lung cancer development, validated through large-scale screening, encompassing a total of 105,844 CpG sites. This design ensures that the detection system comprehensively captures lung cancer signals while avoiding the high cost and noise issues associated with whole-genome sequencing. The enriched library was then subjected to high-throughput sequencing on next-generation sequencing platforms such as Illumina NovaSeq to generate raw data.
[0037] After obtaining the sequencing data, this invention employs an innovative bioinformatics and machine learning analysis pipeline. The raw data, after adapter removal and quality filtering, is aligned to a bisulfite-converted reference genome. Subsequently, an initial methylation level matrix is constructed by calculating the methylation ratio of each target CpG site. To extract the most discriminative information from the high-dimensional, sparse data, this invention employs an advanced feature engineering strategy: First, an initial feature set is constructed by identifying differentially methylated regions (DMRs) and calculating indicators such as the proportion of "co-methylated reads" (i.e., patterns of simultaneous methylation of multiple adjacent CpG sites on the same DNA fragment) within these regions. Then, an unsupervised deep learning model, an autoencoder neural network, is introduced to perform nonlinear dimensionality reduction and feature abstraction on the high-dimensional initial features, automatically learning and extracting low-dimensional, high-information feature vectors that represent the essence of complex methylation patterns.
[0038] Based on these refined feature vectors, this invention employs high-performance machine learning algorithms such as Gradient Boosting Decision Tree (LightGBM) to conduct supervised learning on a training set consisting of a large number of samples diagnosed by the gold standard of pathology (including benign and malignant cases), constructing a classification and diagnostic model called "PulmoSeek". Through model optimization techniques such as recursive feature elimination, a core set of methylation features (e.g., a feature set containing 100 key features) was ultimately determined as the cornerstone of the model's decision-making, ensuring the model's high robustness and interpretability.
[0039] In actual clinical applications, after the patient's sample undergoes the above process, the resulting methylation data is input into a trained PulmoSeek model. The model outputs a continuous malignancy risk score between 0 and 1. By determining an optimal decision threshold (e.g., 0.96) using statistical methods such as the Youden index on a large-scale validation set, this score can be translated into a clear clinical interpretation: a score above the threshold indicates "suspected malignancy," strongly recommending further clinical diagnostic intervention (such as biopsy or surgery); a score below the threshold indicates "probably benign," with a very high negative predictive value. Clinicians can then adopt a safer and more proactive follow-up strategy for the patient, effectively avoiding unnecessary invasive examinations.
[0040] This invention has undergone rigorous retrospective and prospective validation, demonstrating its particular suitability for addressing three major challenges in current clinical management: determining the benign or malignant nature of "uncertain" pulmonary nodules with a diameter of 6-20 mm; early non-invasive detection of stage I (especially stage IA) lung cancer; and universal differentiation of various subtypes, including solid, partially solid, and pure ground-glass nodules. Its superior sensitivity and negative predictive value characteristics allow for seamless integration into existing lung cancer screening and management pathways. This invention proposes and implements a novel management model of "LDCT imaging initial screening + PulmoSeek blood test triage": nodules detected by LDCT, especially those of low to intermediate risk, are preferentially triaged using this non-invasive blood test, with invasive diagnostic procedures initiated only for individuals with high-risk blood test results. This model is expected to maintain a high early detection rate while significantly reducing overall medical costs, minimizing patient trauma, and propelling pulmonary nodule management into a new era of precision and personalization.
[0041] This invention utilizes high-throughput, targeted methylation profiling analysis of cfDNA extracted from patient plasma to capture specific molecular signals of early-stage lung cancer at the epigenetic level. Compared to detecting gene mutations, DNA methylation alterations occur earlier and are more prevalent in tumorigenesis, exhibiting higher tissue specificity. This provides a more stable and sensitive biomarker basis for identifying early-stage microtumors against a background of trace amounts of cfDNA in the blood.
[0042] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.
Claims
1. A diagnostic method for pulmonary nodules based on plasma DNA methylation characteristics, characterized in that: Includes the following steps: S1. Peripheral blood of the subject was collected in a special cfDNA preservation tube, and plasma was obtained by centrifugation. cfDNA was extracted using an optimized magnetic bead method or column extraction method and quality controlled. Subsequently, the cfDNA was subjected to a complete bisulfite conversion treatment to convert unmethylated cytosine to uracil, while methylated cytosine remained unchanged. S2. The transformed DNA was used for library construction and target region enrichment using a customized targeted sequencing panel; S3. After obtaining the sequencing data, an innovative bioinformatics and machine learning analysis pipeline is used. After the raw data is removed by adapters and quality filtered, it is aligned to the reference genome that has undergone bisulfite conversion. Then, the initial methylation level matrix is constructed by calculating the methylation ratio of each target CpG site. S4. Using the gradient boosting decision tree learning algorithm, and then supervising learning on a training set consisting of a large number of samples diagnosed by the gold standard of pathology, a classification and diagnostic model called "PulmoSeek" was constructed. Through the recursive feature elimination model optimization technique, a set of core methylation features were determined as the cornerstone of the model's judgment.
2. The method for diagnosing pulmonary nodules based on plasma DNA methylation characteristics according to claim 1, characterized in that: The quality control in step S1 includes concentration determination and integrity assessment.
3. The method for diagnosing pulmonary nodules based on plasma DNA methylation characteristics according to claim 2, characterized in that: The concentration measured is not less than 5 ng.
4. The method for diagnosing pulmonary nodules based on plasma DNA methylation characteristics according to claim 1, characterized in that: In step S2, the panel covers 12,899 screened and validated specific methylation regions associated with lung cancer, involving a total of 105,844 CpG sites.
5. The method for diagnosing pulmonary nodules based on plasma DNA methylation characteristics according to claim 1, characterized in that: The enriched library in step S2 is subjected to high-throughput sequencing on a next-generation sequencing platform to generate raw data.
6. The method for diagnosing pulmonary nodules based on plasma DNA methylation characteristics according to claim 1, characterized in that: The feature engineering strategy employed in step S3 includes the following steps: First, by identifying differentially methylated regions and calculating the proportion of "comethylation reads" within these regions, an initial feature set is constructed. Then, an unsupervised deep learning model, an autoencoder neural network, is introduced to perform nonlinear dimensionality reduction and feature abstraction on the high-dimensional initial features. This automatically learns and extracts low-dimensional, high-information feature vectors that can represent the essence of complex methylation patterns.
7. The application of a diagnostic method for pulmonary nodules as described in any one of claims 1-6, characterized in that: Includes the following steps: After the patient's sample undergoes the above process, the resulting methylation data is input into the trained PulmoSeek model. The model outputs a continuous malignancy risk score between 0 and 1. By using statistical methods such as the Youden index on a large-scale validation set to determine an optimal decision threshold, this score can be transformed into a clear clinical interpretation: a score above the threshold is considered "suspected malignancy," and further clinical diagnostic intervention is strongly recommended; a score below the threshold is considered "probably benign," with a very high negative predictive value. Clinicians can then adopt a safer and more proactive follow-up strategy for the patient, thereby effectively avoiding unnecessary invasive examinations.
