Peripheral blood leukocyte dna methylation gene network model for lung nodule benign and malignant, invasive risk assessment and construction method thereof

Abstract

The application relates to the field of biomedical technology and discloses a peripheral blood leukocyte DNA methylation gene network model for lung nodule benignity and malignancy and infiltration risk assessment and a construction method thereof. The method is based on an epigenome network medical algorithm, specific methylation characteristics are screened from whole blood leukocytes of sub-centimeter lung nodule patients, and a diagnostic model is constructed. Research proves that, by using peripheral blood leukocyte DNA methylation combined with network topology analysis, the low cfDNA content, the PBMC separation complexity and the cell heterogeneity interference can be effectively overcome, the systemic immune remodeling signal of early lung cancer can be accurately captured, benignity and malignancy classification and infiltration risk stratification can be performed, the model has good sensitivity and specificity, can be used as an important supplement of LDCT screening, can improve early risk assessment accuracy, can guide high-risk patient treatment and reduce over-treatment of low-risk patients, and realizes precise management of lung small nodules.

Description

Technical Field

[0001] This invention relates to the field of biomedical technology, and more specifically, to a peripheral blood leukocyte DNA methylation gene network model and its construction method for assessing the risk of benign, malignant, and invasive pulmonary nodules. Background Technology

[0002] Low-dose computed tomography (CT) screening is an effective method for early cancer detection, which can reduce lung cancer mortality in high-risk populations by at least 20% [3]. With the widespread application of low-dose CT and artificial intelligence-assisted technology, a large number of pulmonary nodules have been detected [4-5], with a very high false positive rate. Although some benign nodules can be excluded by combining imaging features, about 30% to 50% of patients are confirmed to have benign lesions after surgery [6]. Studies have shown that only patients with preinvasive lesions (especially adenocarcinoma in situ AIS and minimally invasive adenocarcinoma MIA) can achieve 100% overall survival at 5 and 10 years after surgical resection [7-8]. However, once it develops into invasive adenocarcinoma, even at stage IA1, the overall survival rate drops to 96.9% after 5 years and 90.7% after 10 years [9]. Therefore, the accurate assessment of the benignity, malignancy and invasiveness of small pulmonary nodules is a major challenge in clinical practice, requiring frequent CT follow-up or even invasive biopsy or surgery, which increases radiation exposure and may cause the best treatment opportunity to be missed [10-12]. There is an urgent need for a new predictive model to improve the accuracy of early lung cancer diagnosis.

[0003] DNA methylation is one of the most studied epigenetic modifications, with the addition of a methyl group at the C5 position of cytosine, mainly occurring at CpG dinucleotides. Under normal physiological conditions, DNA methylation is dynamically balanced by DNA methyltransferases and demethylases, maintaining heterochromatin structure and regulating gene expression

[13] . Numerous studies have shown that DNA methylation dysregulation exists during tumorigenesis and development, including genome-wide hypomethylation, local hypermethylation in specific gene regions (such as CpG islands), and methylation-induced mutations [13-17]. In addition, studies have found that DNA methylation changes occur even before the onset of atypical adenomatous hyperplasia (AAH) in the progression of lung adenocarcinoma

[18] . These findings suggest that DNA methylation can serve as a powerful biomarker for the early detection of lung cancer.

[0004] Existing studies have shown that cancer-specific DNA methylation changes can be detected in the plasma, peripheral blood PBMCs, sputum, saliva, and pleural effusion of lung cancer patients. Some studies have attempted to use the methylation status of a single or a few genes as biomarkers to differentiate between benign and malignant lung nodules. However, since abnormal methylation of CpG islands can silence hundreds of genes associated with the development and progression of lung cancer, relying solely on the methylation of a few genes is insufficient to fully reflect tumor characteristics. Although several studies on DNA methylation in lung cancer patients have provided early risk assessment tools, their balance between sensitivity and specificity is poor, and they are insufficient for validation in sub-centimeter lung nodules—the smaller the lesion, the more drastically the diagnostic sensitivity of traditional methylation biomarkers decreases [19-24].

[0005] Network biology plays a crucial role in biomarker discovery and intervention strategy development. Analysis based on protein-protein interaction (PPI) networks shows that functionally similar genes tend to cluster in adjacent regions within the network. Compared with traditional single-gene analysis, network biology methods are more effective in predicting complex phenotypic outcomes [25-27]. Summary of the Invention

[0006] This invention addresses the technical problems in existing non-invasive diagnostic techniques for pulmonary nodules, such as uneven sensitivity and specificity of circulating cell-free DNA (cfDNA) methylation markers, cumbersome sample processing of peripheral blood mononuclear cells (PBMCs), and significant interference from cellular heterogeneity. This invention uses DNA extracted from peripheral whole blood cells after plasma removal, including all white blood cells such as granulocytes, and provides a pulmonary nodule risk assessment scheme based on a peripheral blood leukocyte (PBLC) DNA methylation gene network model.

[0007] The primary objective of this invention is to provide a peripheral blood leukocyte DNA methylation gene network model and its construction method for assessing the risk of benign, malignant, and invasive pulmonary nodules. This method creatively employs a Selective Epigenomic Module Optimization (SEMO) model, based on pre-trained PPI sub-networks, to accurately capture network topological features from peripheral whole blood cells that reflect systemic immune remodeling, epigenetic changes, and tissue developmental abnormalities.

[0008] Another objective of this invention is to provide methylation markers with high diagnostic efficacy obtained based on the above-described model and their application in the preparation of diagnostic kits.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a methylation biomarker for differentiating between benign and malignant pulmonary nodules. The methylation biomarker is a key node gene (a key node gene in the immune response, tumor development, invasion, and epigenetic regulation network) with network topological significance, obtained by screening based on a pre-trained protein-protein interaction network epigenetic network remodeling optimization model. The methylation biomarker is selected from at least one of the following genes: APOE, BAD, BCL6, BRCA2, CABP1, CREB1, CXCL12, DNMT3B, DNTT, FLT1, FOXB1, FOXG1, FUS, GHSR, GLI3, HES1, HMGB2, IL6R, IRS2, ITPR3, KLF5, LAMA3, LCK, LEF1, MAP2K1, MEIS2, NEUR OD1, NFATC1, PAX6, PAXIP1, PIM1, PINK1, PLCG2, PPP1CB, PPP1R14A, PRKG1, PRKG2, PTK2, PTK6, PTPN1, RAC3, RARG, RGS2, RUNX3, SLC11A1, SRF, THRB, TYK2, UBB, UBE2K, UNG, VCAM1.

[0010] Furthermore, the present invention provides a combination of methylation biomarkers for differentiating benign and malignant pulmonary nodules. The combination comprises the aforementioned 52 genes, which are functionally significantly enriched in the following benign / malignant differentiation-specific signaling pathways, including but not limited to: GO_REGULATION_OF_GROWTH (growth regulation), GO_REGULATION_OF_NEUROGENESIS (neurogenesis regulation), GO_FOREBRAIN_DEVELOPMENT (forebrain development), GO_TRANSCRIPTION_COREGULATOR_BINDING (transcriptional co-regulatory factor binding), GO_CATION_TRANSMEMBRANE_TRANSPORT (cation transmembrane transport), GO_GROWTH (growth), GO_NEGATIVE_REGULATION_OF_TRANSCRIPTION_BY_RNA_POLYMERASE_II (RNA polymerase II transcriptional negative regulation), GO_TRANSFERASE_ACTIVITY_TRANSFERRING_PHOSPHORUS_CONTAINING_GROUPS (phosphotransferase activity), KEGG_PATHWAYS_IN_CANCER (cancer pathway), and GO_IMMUNE_EFFECTOR_PROCESS (immune effector processes).

[0011] Secondly, the present invention provides a methylation marker for identifying the infiltrative nature of pulmonary nodules, wherein the methylation marker is selected from at least one of the following genes: ADORA1, ADRA1A, APAF1, ASPM, ATG12, BCR, FLCN, GANAB, GNAL, GNAS, GSTP1, HRH1, IL12B, IL12, RB1, IL15, ITPR1, MAVS, MLEC, MLST8, NAG LU, PAWR, PENK, PLCB2, PRKAA2, PRKAG3, PRKCG, PRKCZ, RBCK1, SCNN1A, SCNN1G, SEC24B, TIRAP, TLR2, TLR9, TNFAIP3, TRHR, TSC1, WNT5A.

[0012] Furthermore, the present invention provides a methylation biomarker combination for identifying the invasiveness of pulmonary nodules. This combination comprises the aforementioned 38 genes, which reflect tumor invasion-related microenvironmental interactions and are functionally significantly enriched in the following invasiveness-specific signaling pathways, including but not limited to: KEGG_CALCIUM_SIGNALING_PATHWAY (calcium signaling pathway), GO_DEFENSE_RESPONSE_TO_SYMBIONT (defense response to symbionts), GO_TUMOR_NECROSIS_FACTOR_SUPERFAMILY_CYTOKINE_PRODUCTION (tumor necrosis factor superfamily cytokine production), GO_EPITHELIAL_TUBE_FORMATION (epithelial tube formation), GO_GLYCOPROTEIN_METABOLIC_PROCESS (glycoprotein metabolism), GO_NEURAL_TUBE_FORMATION (neural tube formation), GO_TUBEFORMATION (tube formation), GO_DENDRI TIC_TREE (dendritic structure), GO_POSITIVE_REGULATION_OF_INNATE_IMMUNE_RESPONSE (positive regulation of innate immune response), GO_POSITIVE_REGULATION_OF_RESPONSE_TO_BIOTIC_STIMULUS (positive regulation of biostimulation response), GO_TUBE_FORMATION (tube formation), GO_TUMOR_NECROSIS_FACTOR_SUPERFAMILY_CYTOKINE_PRODUCTION (tumor necrosis factor superfamily cytokine production), and GO_POSITIVE_REGULATION_OF_INNATE_IMMUNE_RESPONSE (positive regulation of innate immune response).

[0013] Thirdly, the present invention provides a comprehensive methylation marker for jointly differentiating between benign and malignant pulmonary nodules and their invasiveness, wherein the methylation marker is selected from at least one of the following genes: ADCY8, AKT1, BMP4, BMP7, BRAF, CALM1, CALM2, CASP8, CD36, CDKN1A, CYBA, DRD2, EP300, ERCC1, FGF8, HRAS, IGF1, KR AS, LEP, MAP3K5, MAPK3, NOD2, PSEN1, PTK2B, RAC1, SHC1, SIRT1, SMAD7, SQSTM1, TGFBR1, TGFBR2, TNF, TSPO, UBE2N.

[0014] Furthermore, this invention provides a combination of methylation markers for the combined differentiation of benign and malignant pulmonary nodules and their invasiveness. The combination consists of the aforementioned 34 genes, which are functionally significantly enriched in the following common signaling pathways shared by benign and malignant pulmonary nodules and their invasiveness, including but not limited to: GO_REGULATION_OF_TRANSFERASE_ACTIVITY (regulation of transferase activity), GO_POSITIVE_REGULATION_OF_TRANSFERASE_ACTIVITY (positive regulation of transferase activity), GO_CARTILAGE_DEVELOPMENT (cartilage development), GO_CONNECTIVE_TISSUE_DEVELOPMENT (connective tissue development), GO_INTRINSIC_APOPTOTIC_SIGNALING_PATHWAY (intrinsic apoptosis signaling pathway), GO_POSITIVE_REGULATION_OF_CELL_POPULATION_PROLIFERATION (positive regulation of cell population proliferation), GO_SKELETAL_SYSTEM_DEVELOPMENT (skeleton system development), KEGG_PATHWAYS_IN_CANCER (cancer pathway), GO_KINASE_ACTIVITY (kinase activity), and GO_PROTEIN_KINASE_ACTIVITY (protein kinase activity).

[0015] Fourthly, the present invention provides the application of the above-mentioned methylation markers or combinations of markers.

[0016] Specifically, this involves the application of reagents (such as primers, probes, capture chips, etc.) for detecting the methylation level of the methylation markers in genomic regions in the preparation of auxiliary diagnostic kits for identifying benign and malignant and / or invasive pulmonary nodules in biological samples (especially peripheral whole blood cell samples).

[0017] Fifthly, the present invention provides the use of reagents for detecting the methylation level of said methylation markers or combinations of said methylation markers in the preparation of auxiliary diagnostic kits for benign, malignant, or invasive pulmonary nodules.

[0018] Sixthly, the present invention provides a method for constructing a peripheral blood leukocyte DNA methylation gene network model for assessing the risk of benign or malignant pulmonary nodules, comprising the following steps: (1) Collect peripheral blood leukocyte samples from the subjects and divide them into training set and test set; the peripheral blood samples are from benign and malignant lung nodule samples; the lung nodules are sub-centimeter lung nodules, preferably sub-centimeter lung nodules with a diameter ≤8mm; (2) Extracting DNA from peripheral blood leukocytes in peripheral blood samples; (3) Detect the methylation level of the DNA of the peripheral blood leukocytes to obtain methylation data; (4) Construct a feature matrix and a SEMO algorithm model from the methylation data, and screen out benign and malignant methylation markers of lung nodules based on the training set sample data; (5) Validate the model’s effectiveness using test set sample data and determine the methylation markers to be used for predicting the benign or malignant nature of lung nodules. The SEMO algorithm model (see the paper Utilizing Pre-trained Network Medicine Models for Generating Biomarkers, Targets, Re-purposing Drugs, and Personalized Therapeutic Regimes: COVID-19 Applications, doi: https: / / doi.org / 10.1101 / 2023.02.21.527754) is as follows: in, The average gene score of genome X. S represents the average gene score of genome Y. x 2 S represents the score variance of genome X. Y 2 Let be the score variance of genome Y, n be the number of genes in genome X, and m be the number of genes in genome Y.

[0019] Seventhly, the present invention provides a risk assessment model for benign or malignant pulmonary nodules, the risk assessment model being constructed according to the above method, and comprising: 1) Data acquisition module, at least used to acquire sample datasets; 2) Sequencing module, used at least to obtain sequencing data of methylation marker genes; 3) A data alignment module, used at least to align the sequencing data with a reference sequence and determine the methylation result of the markers in the sequencing data based on the alignment result; 4) Result determination module, which is used at least to calculate the predicted score threshold through statistical model analysis (such as logistic regression analysis) and determine whether the sample to be tested is a benign or malignant lung nodule.

[0020] Eighthly, the present invention provides a method for constructing a peripheral blood leukocyte DNA methylation gene network model for assessing the risk of pulmonary nodule infiltration, comprising the following steps: (a) Collect peripheral blood leukocyte samples from the subjects and divide them into training set and test set; the peripheral blood samples are from malignant lung nodule samples; the lung nodules are sub-centimeter lung nodules, preferably sub-centimeter lung nodules with a diameter ≤10mm; (b) Extracting DNA from peripheral blood leukocytes in peripheral blood samples; (c) Detect the methylation level of the DNA in the peripheral blood leukocytes to obtain methylation data; (d) Construct a feature matrix from the methylation data, build a SEMO algorithm model, and screen out invasive methylation markers of lung nodules based on the training set sample data; (e) Validate the effectiveness of the model using test set sample data and determine the methylation markers to be used for the final assessment of pulmonary nodule infiltrative prediction; The SEMO algorithm model is as follows: in, The average gene score of genome X. S represents the average gene score of genome Y. x 2 S represents the score variance of genome X. Y 2 Let be the score variance of genome Y, n be the number of genes in genome X, and m be the number of genes in genome Y.

[0021] Ninthly, the present invention provides a risk assessment model for invasive pulmonary nodules, the risk assessment model being constructed according to the above method, comprising: a) A data acquisition module, used at least to acquire sample datasets; b) A sequencing module, used at least to obtain sequencing data for methylation marker genes; c) A data alignment module, at least used to align the sequencing data with a reference sequence, and determine the methylation result of the markers in the sequencing data based on the alignment result; and d) Result determination module, which is used at least to calculate the predicted score threshold through statistical model analysis (such as logistic regression analysis) and determine whether the sample to be tested is a non-invasive or invasive lung nodule.

[0022] By employing the above technical solution, the present invention has at least the following advantages and beneficial effects: (i) By detecting key methylation sites in peripheral whole blood leukocytes that construct the above-mentioned specific signaling pathway network, this invention can effectively overcome the interference of low cfDNA content, heterogeneity of cell components, and complexity of PBMC separation in peripheral blood samples, and accurately capture systemic abnormal signals of the "immune-tumor development and infiltration" axis, thereby achieving high sensitivity and high specificity in the determination of the nature of lung nodules.

[0023] (II) This invention utilizes a gene network-based method to screen for lung cancer-specific DNA methylation features from blood cell samples of patients with sub-centimeter (≤8 mm diameter) malignant lung nodules and benign lesions, and constructs a diagnostic model to differentiate between benign and malignant lung nodules and their degree of infiltration. Results show that this model exhibits good sensitivity and specificity in sub-centimeter lung cancer patients and has certain predictive value for patient prognosis. This novel blood cell methylation-based diagnostic strategy is expected to become an important supplement to LDCT (low-dose spiral CT) screening, improving the accuracy of risk assessment in the early, curable stages of lung cancer, guiding follow-up and treatment strategies for high-risk patients, reducing unnecessary surgery and radiation exposure for low-risk patients, and ultimately achieving precise management of small lung nodules and long-term survival benefits for patients. Attached Figure Description

[0024] Figure 1 This is the SEMO model algorithm of the present invention.

[0025] Figure 2 This is a schematic diagram of the research process of this invention.

[0026] Figure 3 This invention provides a preferred embodiment of the efficacy in predicting the benign or malignant nature of sub-centimeter pulmonary nodules.

[0027] Figure 4 This invention provides a preferred embodiment of the predictive efficacy for sub-centimeter pulmonary nodule infiltration.

[0028] Figure 5 This is a Venn diagram illustrating differential gene characteristics of DNA methylation in a preferred embodiment of the present invention. Detailed Implementation

[0029] Lung cancer is one of the leading causes of cancer death worldwide. Although low-dose spiral CT (LDCT) screening can reduce lung cancer mortality by approximately 20% in high-risk populations, it also leads to a significant increase in the detection rate of pulmonary nodules and a high false-positive rate, making the differentiation between benign and malignant nodules, as well as invasive nodules, a major clinical challenge. This invention aims to construct a gene network biomarker model based on peripheral blood leukocyte DNA methylation to determine the benign, malignant, and invasive nature of pulmonary nodules, and to evaluate the predictive performance of this model.

[0030] This invention retrospectively included 149 patients suspected of having lung cancer who underwent surgery at the Strategic Support Force Characteristic Medical Center between November 2019 and November 2020, and were followed up until November 2023. Venous blood was collected from all patients preoperatively, and cell-free DNA was extracted for methylation detection. Selective remodeling of protein-protein interaction networks (SEMO) was used to generate biomarkers of gene network imbalance. Two predictive models were constructed for different clinical discrimination tasks: a benign / malignant classification model was built using 60% of cases as the training set for ≤8 mm lung nodules (benign + malignant); an invasive classification model was built using 60% of cases as the training set for ≤10 mm lung cancer (invasive + non-invasive). After model establishment, their performance was validated in the ≤30 mm nodule subgroup and the entire sample.

[0031] The results showed that the area under the ROC curve (AUC) of the benign / malignant classification model was 0.94 in nodules ≤30 mm, while the AUC of the invasive classification model was 0.92 in nodules ≤30 mm. The sensitivities of the benign / malignant model in the internal test set and the ≤30 mm nodule subgroup were 84.21% and 85.62%, respectively, with specificities of 100% and 84.25%. The invasive model's sensitivity in the ≤10 mm lung cancer subgroup and the entire sample was 72.73% and 68.48%, respectively, with specificities of 91.67% and 92.68%, respectively.

[0032] This invention demonstrates excellent accuracy in differentiating benign and malignant pulmonary nodules and their invasiveness based on a gene network model of peripheral blood leukocyte DNA methylation. It can even effectively identify the benign and malignant nature of subcentimeter pulmonary nodules. This is the first report of using peripheral blood leukocyte DNA methylation to predict the benign and malignant classification of pulmonary nodules and stratify their invasiveness risk. This non-invasive and economical blood testing method holds promise as a new tool to effectively improve the accuracy of assessing the benign and malignant nature and invasiveness risk of pulmonary nodules, reduce patient anxiety, avoid unnecessary radiation exposure, and achieve precise management of pulmonary nodules. The correlation of gene network nodes can indicate the molecular characteristics of subcentimeter pulmonary nodules, and this method helps to understand the mechanisms of pulmonary nodule development.

[0033] The following examples are used to illustrate the present invention, but are not intended to limit the scope of the invention. Unless otherwise specified, the technical means used in the examples are conventional means well known to those skilled in the art, and the raw materials used are all commercially available products.

[0034] Example: A method for differentiating benign, malignant, and infiltrative subcentimeter pulmonary nodules based on a gene network model of peripheral blood leukocyte DNA methylation. I. Materials and Methods 1. Research Subjects From November 2019 to January 2021, this study included 149 patients who underwent surgery for suspected lung cancer. Preoperatively, 10 mL of venous blood was collected from each patient. Blood cells and plasma were separated within 4 hours (10 mL EDTA anticoagulant tube, centrifuged twice, 1500×g, 10 minutes). Separated blood cells were stored at -80℃ until DNA extraction. Postoperative pathological diagnoses of all patients conformed to the 2015 WHO histological classification of lung cancer. Staging was performed according to the 8th edition of the NCCN lung cancer staging system.

[0035] 2. DNA methylation analysis After nucleic acid extraction and quality control, 500 ng of genomic DNA was collected and purified using the ZYMO Research EZ DNAMethylation Gold™ Kit according to the kit instructions. DNA methylation was then tested using the proprietary P500 capture sequencing kit. This system contains 500 genes selected from the GEO (Gene Expression Omnibus) blood cell DNA methylation database and ZYMO's proprietary 850k microarray blood cell DNA methylation data, identifying the top 500 genes associated with age and aging. Previous studies have confirmed that these genes are significantly associated with aging and tumor immunity. This system captures promoter region sequences from these 500 genes for high-throughput sequencing analysis. After thawing the pre-constructed probe solution, library hybridization, capture, elution, and purification were performed. The DNB rapid preparation kit was used to prepare the sequencing reagents, and sequencing was conducted using the DNBSEQ-T7 sequencer developed by BGI Genomics. Sequencing data was visualized using FastQC software for quality control, including removal of sequencing adapters and low-quality sequencing bases. After quality control, the data was aligned with a reference genome, and alignment efficiency was calculated using bismark. PCR redundancy was removed using the deduplicate_bismark tool, and the redundancy of each sample was calculated. For each gene, the average β value of all CpG sites within its promoter region was calculated as the methylation characteristic value for that gene; a higher β value indicates a higher degree of methylation.

[0036] This study was approved by the Ethics Committee of the Special Medical Center of the Strategic Support Force (K2018 Lun Shen No. (01)) and the Ethics Committee of Peking University Cancer Hospital (2024YJZ75-GZ01).

[0037] 3. Construction of gene network biomarkers This study employs the Protein Network Selective Remodeling (SEMO) algorithm to construct biomarkers of gene network imbalance. For detailed implementation information, please refer to the paper "Utilizing Pre-trained Network Medicine Models for Generating Biomarkers, Targets, Re-purposing Drugs, and Personalized Therapeutic Regimes: COVID-19 Applications," doi: https: / / doi.org / 10.1101 / 2023.02.21.527754. First, protein-protein interaction data were integrated from the Human Protein Reference Database (HPRD, www.hprd.org), dividing the entire PPI network into different central gene sets. The "in" gene set was defined as the set of all genes interacting with protein i, with a size equal to the number of genes interacting with protein i. This study only considered central gene sets with at least 20 interacting genes for subsequent analysis. To identify phenotype-related SEMO features, we traversed various functional gene set resources, including Gene Ontology, KEGG Pathway, Reactome Pathway, BIOCARTA Pathway, WikiPathway, Hallmark gene sets, and the Pathway Interaction Database, etc. Figure 1 ).

[0038] The SEMO algorithm model is as follows: Considering the PPI gene set i There are multiple genes, the gene pathway target gene set k1 and the PPI gene set. i The intersection of these sets is x1, and the distinct gene sets (belonging to PPI gene set i but not in gene pathway target gene set k1) are Y: in, The average gene score of genome X. S represents the average gene score of genome Y. x 2 S represents the score variance of genome X. Y 2 Let be the score variance of genome Y, n be the number of genes in genome X, and m be the number of genes in genome Y.

[0039] Therefore, for any combination of PPI genome i and gene pathway k, the calculated Tik (SEMO index Tik) is used to represent the state of SEMO (PPI-gene pathway). For example, "CCK.n-GODEVELOPMENTAL GROWTH" represents SEMO associated with PPI genome "CCK" and "GO DEVELOPMENTAL GROWTHpathway".

[0040] Then, all combinations of PPIs and gene pathways are iterated. For each patient's DNA methylation profile data, the SEMO index Tik for each PPI set i and each gene pathway k is calculated. A t-test is then used to check for significant differences between phenotypes for each SEMO feature. The SEMO features are then ranked according to the t-test p-values ​​to form a phenotype-related SEMO list.

[0041] 4. Causal Emergence Calculation For each phenotype associated with the SEMO feature, the p-value from the t-test above was transformed into an association index AI = -LOG10(P). This means that the smaller the p-value, the larger the AI ​​value. Since SEMO is a higher-level macro-feature built upon a set of genes, we observed the differences between SEMO and micro-features, where micro-features represent the correlation between a single gene and a phenotype. Taking the correlation between the SEMO feature APOA1.n-Niacin and the malignant phenotype as an example, we established the PPI involved in this SEMO feature, which includes all genes in the APOA1.n proteome network. We used a t-test to calculate the correlation between each gene and the phenotype (malignant - 0 - 8), obtaining the eigenvalue for each gene. From this, we calculated the AI ​​value for each gene, i.e., AI = -LOG10(P). The AI ​​values ​​of all genes in this SEMO form a distribution. We plotted the distribution of gene AI values ​​together with the AI ​​values ​​of SEMO (single values).

[0042] 5. Classification and prediction models The classification prediction model was developed based on the Lasso (minimum absolute contraction and selection operator) algorithm. For any patient population participating in the phenotypic test, 60% of the data was used for training and 40% for testing. Taking the establishment of a benign / malignant classification model based on a combination of protein-protein interaction (PPI) and gene pathway as an example, we selected the top 10 semo features significantly associated with benign / malignant classification. The model was built using the Lasso algorithm, trained on 60% of the cases, and tested on the remaining 40% of the data. The ROC curves and AUC (the area under the receiver operating characteristic curve) for the training and test sets are reported.

[0043] 6. Statistical Analysis Clinical characteristics of each group were compared, and the classification parameter was calculated using chi-square (χ²). 2 For continuous variables, the Wilcoxon rank-sum test was used. Sensitivity and specificity, along with 95% confidence intervals, were calculated for the entire sample and different subgroups. Logistic regression was used to assess the predictive power of the model score for outcomes, and the area under the receiver operating characteristic (AUC) was calculated. Positive likelihood ratios (LR+) and negative likelihood ratios (LR-) were also calculated. Multivariate logistic regression analysis was also used to examine the correlation between the model score and the benign / malignant or invasive nature of lung nodules. The odds ratio (OR) in the multivariate regression model was used to measure the contribution of each predictor to the classification results, and the AUC was used to evaluate the overall predictive performance of the model.

[0044] 7. Comparison with other traditional models In this cohort, the performance of several existing lung nodule malignancy risk prediction models (Mayo Clinic model, VA model, Brock model, and the externally validated Peking University People's Hospital PKUPH model) was evaluated. The AUC results for each model are detailed in the table. In contrast, the DNA methylation network model constructed in this invention exhibited a higher AUC on the independent validation set. The specific prediction formulas for the aforementioned traditional models are as follows: The Mayo model (see references Swensen SJ, Silverstein MD, Ilstrup DM) et alThe probability of malignancy in solitary pulmonary nodules. Application to small radiologically indeterminate nodules. Arch Intern Med, 1997, 157(8): 849-855) In the formula, X = -6.8272 + (0.0391 × age) + (0.7917 × smoking history) + (1.3388 × history of malignant tumor) + (0.1274 × nodule diameter) + (1.0407 × spiculation) + (0.7838 × upper lobe location). Previous smoking history = 1, no smoking history = 0; history of extrathoracic tumors more than 5 years ago = 1, no history of extrathoracic tumors = 0; nodule diameter in millimeters; spiculation present = 1, no spiculation absent = 0; nodule location in the left or right upper lobe = 1, other locations = 0.

[0045] VA model (see Gould MK, Ananth L, Barnett PG, ...) et al In the study "A clinical model to estimate the pretest probability of lung cancer in patients with solid pulmonary nodules. Chest, 2007, 131(2): 383-388", X = -8.404 + (2.061 × smoking history) + [0.779 × age (10 years)] + (0.112 × diameter) - [0.567 × smoking cessation time (10 years)]. A history of smoking = 1, no smoking history = 0; age (10 years) represents the age at which the nodule was discovered, divided by 10; nodule diameter represents the maximum diameter of the nodule as measured; smoking cessation time (10 years) represents the number of years since quitting smoking, divided by 10 (0 indicates not having quit).

[0046] Brock model (see literature McWilliams A, Tammemagi MC, Mayo JR, et alIn the study "Probability of cancer in pulmonary nodules detected on first screening CT," NEngl J Med, 2013, 369(10): 910-919, X = -6.6144 + (0.6467 × sex) + (-5.5537 × diameter) + (0.9309 × spiculation) + (0.6009 × upper lobe). Here, female = 1, male = 0; nodule diameter represents the maximum diameter (mm) measured for the nodule; spiculation presence = 1, no spiculation absence = 0; nodule located in the upper lobe = 1, located elsewhere = 0.

[0047] In the PKUPH model (see the literature Li Yun, Chen Kezhong, Sui Xichao, et al. Establishment of a mathematical prediction model for judging the benignity and malignancy of solitary pulmonary nodules. Peking University Journal (Medical Edition), 2011, 43(3): 450-454), X = -4.496 + (0.07 × age) + (0.676 × diameter) + (0.736 × spiculation) + (1.267 × family history of tumors) - (1.615 × calcification) - (1.408 × boundary). Among them, spiculation = 1, no spiculation = 0; family history of tumors = 1, no family history of tumors = 0; calcification = 1, no calcification = 0; clear nodule boundary = 1, blurred boundary = 0; nodule diameter represents the maximum diameter (mm) of the nodule as measured.

[0048] Infiltrative determination model: There are few reports on the determination of invasiveness of pulmonary nodules. In previous studies, Chen Haiquan from Fudan University Cancer Hospital reported a model for determining the invasiveness of pulmonary nodules (see the literature YE T, WU HX, WANG SP, et al. Radiologic identification of pathologic tumor invasion in patients with lungadenocarcinoma[J]. JAMA Netw Open, 2023, 6(10): e2337889). This model was validated in the data of this study and compared with the DNA methylation gene network model.

[0049] The technical route of this invention is as follows: Figure 2The nodules were stratified according to pathological characteristics and diameter. High-throughput methylation sequencing was used to obtain features, differential methylation sites were mined based on the SEMO framework, and LASSO regression was used to complete feature compression and modeling, resulting in two types of models (benign / malignant and invasive). The methylation modeling workflow is as follows: stratified comparison → high-throughput sequencing → SEMO differential mining → LASSO modeling → dual-task output (benign / malignant / invasive).

[0050] II. Research Results 1. Clinical characteristics This study included 178 patients from November 2019 to January 2021 at the Department of Thoracic Surgery, Special Medical Center of the Strategic Support Force, who underwent surgical treatment and had confirmed lung cancer due to pulmonary nodules (maximum diameter ≤30mm) detected by chest CT scans. One case of small cell lung cancer and 28 cases with unsuitable blood samples were excluded, resulting in a total of 149 patients (Table 1). Of these, 123 (82.56%) were pathologically confirmed as lung cancer postoperatively, all of which were adenocarcinomas: 10 cases of carcinoma in situ, 31 cases of minimally invasive adenocarcinoma, and 81 cases of invasive adenocarcinoma. 26 cases (17.44%) were benign lung lesions, including 7 cases of atypical adenomatous hyperplasia, 11 cases of alveolar epithelial hyperplasia, 5 cases of meningioma-like nodules, and 3 cases of sclerosing hemangioma. Of the 118 patients (79.19%) with lung cancer, 118 were in stage 0-I, including 69 in stage Ia1, 20 in stage Ia2, 2 in stage Ia3, 17 in stage Ib, 2 in stage IIB, and 3 in stage IIIa. There were 51 males (34.23%) and 101 females (65.78%). The mean age of the entire group was (range), with a mean age of 60.5 years for benign cases (median age 60 years, range: 27-83 years) and a mean age (range) for malignant cases. The mean age for minimally invasive adenocarcinoma was 54.34 years, with a median age of 50.5 years (range: 34-80 years), and the mean age for adenocarcinoma in situ was 63.11 years, with a median age of 64 years (range: 47-81 years).

[0051] Stratified by nodule diameter: 102 cases (≤10 mm) were nodules, accounting for 68.46%, of which 78 were malignant (38 invasive and 40 pre-invasive lesions) and 24 were benign; 84 cases were nodules ≤8 mm, of which 60 were malignant and 24 were benign. 90 patients had a single pulmonary nodule, and 59 patients had multiple pulmonary nodules. 88 patients had a single lung cancer, and 35 patients had multiple lung cancers (Table 1).

[0052] Table 1 Patient Clinical Information

[0053] 2. SEMO Model Construction and Performance Assessment To accommodate different clinical decision-making needs, we constructed two models for benign / malignant differentiation and invasive differentiation, respectively. First, 60% of lung nodule cases (including both benign and malignant nodules) with a diameter ≤8 mm were randomly selected as the training set to build a benign / malignant classification model, with the remaining 40% used as the test set. Second, 60% of lung cancer cases (including both invasive and non-invasive cases) with a diameter ≤10 mm were randomly selected as the training set to build an invasive classification model, with the remaining 40% used as the test set. After model construction, we validated and evaluated the model performance in a nodule cohort ≤30 mm.

[0054] 2.1 Benign / Malignant Model The American College of Chest Physicians guidelines recommend that clinicians use effective predictive models to assess the risk of malignant transformation of pulmonary nodules detected by LDCT screening, such as the Mayo Clinic model, the VA model, the Brock model, and the externally validated Peking University People's Hospital (PKUPH) model. We calculated the predictive efficacy of the aforementioned Mayo Clinic, VA, Brock, and PKUPH models in this study's patient cohort.

[0055] For nodules 8 mm or less, the AUC of SEMO was 0.98, with a sensitivity of 84.21% and a specificity of 1; the AUC of the 1997 Mayo Clinic model was 0.538, with a sensitivity of 31.75% and a specificity of 78.26%; the AUC of the 2005 VA model was 0.449, with a sensitivity of 6.35% and a specificity of 1; the AUC of the 2013 Brock model was 0.604, with a sensitivity of 19.05% and a specificity of 86.96%; and the AUC of the 2011 PKUPH model was 0.471, with a sensitivity of 26.98% and a specificity of 82.61%. In nodules ≤3cm, the SEMO model had an auc value of 0.94, with a sensitivity of 85.62% and a specificity of 84.25%; the 1997 Mayo Clinic model had an auc value of 0.695, with a sensitivity of 56.43% and a specificity of 84.62%; the 2005 VA model had an auc value of 0.606, with a sensitivity of 48.57% and a specificity of 76.92%; the 2013 Brock model had an auc value of 0.748, with a sensitivity of 47.14% and a specificity of 92.31%; and the 2011 PKUPH model had an auc value of 0.655, with a sensitivity of 48.57% and a specificity of 84.62%. In this study cohort, compared with data using the Mayo Clinic model, VA model, Brock model, and PKUPH model, the methylation model showed better performance on the independent validation set. Figure 3 (Table 2).

[0056] Table 2 Comparison of benign and malignant cases Note: The 8mm model was validated on a dataset with nodule diameters ≤30mm. The parameters required for the classic models were determined through discussion by two associate senior thoracic surgeons and one associate senior radiology specialist. The classic models are the 1997 Mayo Clinic model, the 2005 VA model, the 2013 Brock model, and the 2011 Peking University People's Hospital model.

[0057] 2.2 Infiltrative Model There are few published studies on the benign or malignant assessment of small pulmonary nodules, and the prediction of sub-centimeter pulmonary nodule invasiveness using DNA methylation is a first-time report. The model was compared with one published in JAMA OPEN by Chen Haiquan of Fudan University Cancer Hospital, and with a radiomics model published in Radiology. For nodules 10 mm or smaller, the SEMO model had an AUC of 0.98, accuracy of 82.61%, sensitivity of 72.73%, and specificity of 91.67%. Validating Chen Haiquan's model in this study cohort yielded an AUC of 0.6737, accuracy of 67.95%, sensitivity of 44.74%, and specificity of 90%, lower than the data reported in his article (accuracy 96.0%, sensitivity 62.5%, specificity 97.2%), which did not report its AUC. For nodules 30 mm or smaller, the SEMO model had an AUC of 0.92, accuracy of 75.94%, sensitivity of 68.48%, and specificity of 92.68%. The Chen Haiquan model was validated in this study cohort, with an AUC of 0.76, accuracy of 70.68%, sensitivity of 63.04%, and specificity of 87.80%. These figures are lower than those reported in the article, which showed accuracy of 83.0%, sensitivity of 82.4%, and specificity of 83.3%. The article did not report the AUC of that model. The model of this invention is superior to the Chen Haiquan model and has stronger discriminative ability. Figure 4 (Tables 3 and 4).

[0058] Table 3 Comparison of wettability and non-wetting properties Table 4. Wetting vs. Non-wetting 3. DNA methylation characteristics based on gene networks 3.1 Genetic characteristics Based on a systematic analysis of the DNA methylation characteristics of peripheral blood leukocytes, the gene set associated only with benign or malignant lesions contained 52 genes, the gene set associated only with invasiveness contained 38 genes, and the gene set shared by both contained 34 genes. Figure 5Genes appearing in the lung nodule benign / malignant prediction model are referred to as the "benign / malignant related gene set." Genes appearing in the invasiveness prediction model are referred to as the "invasiveness related gene set." The overlapping part of the two is the benign / malignant and invasiveness related gene set. Genes only related to benign / malignantness showed significant differences only in the benign / malignant classification of lung nodules. Representative genes include APOE, BRCA2, FOXB1, PIM1, and VCAM1 (Table 5). These genes are mostly involved in tumorigenesis-related signaling pathways, such as cell cycle regulation, DNA damage response, and proliferation and apoptosis regulation (Table 7). Genes only related to invasiveness are mainly regulated during the evolution of lung nodules from non-invasive to invasive. Representative genes include ADORA1, APAF1, FLCN, ITPR1, and TLR9 (Table 5). They are mainly enriched in biological processes such as cell migration, intercellular adhesion, and immune response, suggesting that they are associated with nodule invasiveness (Table 7). Genes shared by benign / malignantness and invasiveness showed significant differences in methylation in both benign / malignantness and invasiveness discrimination, and have a dual role. Representative genes include AKT1, BMP4, BRAF, IGF1, and KRAS (Table 5), involving multiple classic oncogenic pathways such as PI3K-Akt, MAPK, and Wnt (Table 7). These genes may play important regulatory roles throughout the entire process of pulmonary nodule development and progression.

[0059] Table 5. Gene sets associated only with benign or malignant characteristics, gene sets associated only with invasiveness, and gene sets shared by both. The accession numbers of each gene in NCBI are shown in Table 6: Table 6. Accession numbers of each gene in NCBI 3.2 Gene signaling pathways are shown in Table 7: Table 7 Gene signaling pathways With the widespread use of low-dose spiral CT (LDCT) in lung cancer screening, the detection rate of sub-centimeter (≤10 mm) lung nodules has increased significantly. However, the high false positive rate, the risk of over-intervention, and the psychological burden on patients make the scientific management of these small nodules a significant clinical challenge.

[0060] There are currently two main management strategies for sub-centimeter pulmonary nodules. One is to regularly follow up with imaging to observe the dynamic changes of the nodules and avoid unnecessary surgical intervention for lesions that may be benign. Studies have shown that about 80% of pulmonary nodules do not progress during the follow-up period and can remain stable for a long time [2]. Therefore, follow-up observation helps to reduce the surgical treatment rate and alleviate the burden on medical resources. However, the limitation of this strategy is that the sensitivity of imaging diagnosis is limited, and some invasive lesions may be delayed in treatment during long follow-up. Another view advocates timely intervention for nodules with early malignant characteristics within the "time window for curative surgery for lung cancer". Studies have shown that patients with pathological types of adenocarcinoma in situ (AIS) or minimally invasive adenocarcinoma (MIA) can achieve a 5-year and 10-year survival rate of 100% after surgical resection [7-8]; however, once the nodule develops into invasive adenocarcinoma, even if it is in stage IA1, its survival rate drops significantly [9]. Therefore, it is crucial to grasp the best time for surgical intervention to improve the prognosis of patients. However, overly aggressive interventions may waste medical resources and increase the risk of unnecessary postoperative complications. Therefore, it is necessary to use precise risk stratification tools to guide decision-making and balance benefits and risks.

[0061] This study developed a gene network model based on peripheral blood leukocyte DNA methylation characteristics, providing a non-invasive and highly sensitive tool for the early identification of sub-centimeter pulmonary nodules. The model utilizes a protein-protein interaction network to construct various phenotype-related methylation imbalance features (SEMOs), enabling efficient differentiation between benign and malignant pulmonary nodules and their invasive status. In this study, the DNA methylation model achieved an AUC of 0.94 for differentiating benign from malignant nodules ≤30 mm and an AUC of 0.92 for predicting the invasiveness of nodules ≤10 mm, significantly outperforming traditional clinical models such as Mayo, VA, and Brock. Further analysis showed that the methylation model score, as an independent predictor, was unaffected by conventional risk factors such as nodule size, smoking status, or gender, demonstrating good independence and generalization ability. Furthermore, we explored the use of blood cell DNA methylation scores for preoperative surgical pathway decision-making—those with lower scores were recommended for wedge resection, while those with higher scores were recommended for lobectomy combined with lymph node dissection. Preliminary results showed a high degree of consistency between the methylation model score and the postoperative pathological infiltration classification, suggesting that this model has important clinical value in preoperative risk assessment and classification diagnosis. Our next step is to apply the blood cell DNA methylation model score to the formulation of preoperative surgical strategies: wedge resection for patients with lower scores, and lobectomy with regional lymph node dissection for patients with higher scores.

[0062] DNA methylation is a key epigenetic modification that plays a crucial role in tumorigenesis and development by regulating gene expression. Especially in the early stages of pulmonary nodules, peripheral blood leukocyte DNA methylation analysis can effectively differentiate between benign and malignant sub-centimeter pulmonary nodules and their invasiveness, greatly improving the accuracy of early diagnosis.

[0063] This study identified three gene sets: genes associated only with benign or malignant transformation, genes associated only with invasiveness, and gene sets shared by both. Functional analysis and pathway enrichment (GO / KEGG) of these genes revealed their potential mechanisms of action in the early diagnosis of lung cancer: gene sets associated only with benign or malignant transformation, such as APOE, BRCA2, FOXB1, PIM1, and VCAM1, are mainly involved in cell cycle regulation, cell proliferation, and apoptosis signaling pathways. GO analysis may focus on pathways related to cell growth regulation, DNA damage repair, and apoptosis. KEGG pathway analysis may involve key cancer-related pathways such as the PI3K-Akt signaling pathway and the p53 signaling pathway. These genes play a crucial regulatory role in the early stages of malignant transformation of lung nodules. Gene sets associated only with invasiveness, such as ADORA1, APAF1, FLCN, ITPR1, and TLR9, are mainly involved in cell migration, invasion, and metastasis. GO enrichment analysis likely reveals biological processes related to cell adhesion, extracellular matrix remodeling, and epithelial-mesenchymal transition (EMT). KEGG analysis may involve Wnt, MAPK, and TGF-β signaling pathways, which play important roles in tumor invasion and metastasis, suggesting the key regulatory functions of these genes in the transformation of subcentimeter pulmonary nodules from non-invasive to invasive carcinogenesis. Among the shared gene set, AKT1, BMP4, BRAF, IGF1, and KRAS are involved in both of these processes simultaneously, reflecting the continuous pathological process of pulmonary nodules progressing from benign to malignant and further to invasive stages. These genes exhibit functional overlap, potentially simultaneously regulating cell proliferation, apoptosis, and invasion / migration, revealing a broader and more comprehensive biological mechanism in the early stages of pulmonary nodules. Protein-protein interaction (PPI) network analysis can further clarify the interaction patterns and core regulatory nodes among these genes, potentially identifying key hub genes. These hub genes may play an important role in the early diagnosis and personalized treatment of lung cancer, suggesting possible future therapeutic targets. By analyzing expression patterns and mutations in the TCGA database, the differences in gene expression or mutation status can further validate their diagnostic and prognostic value in the early stages of lung cancer, thereby clarifying the clinical application potential of methylation modification in assessing the benignity, malignancy, and invasiveness of sub-centimeter pulmonary nodules. The important role of specific gene sets identified through DNA methylation analysis in the diagnosis, development mechanisms, and early intervention of pulmonary nodules provides new molecular evidence for the formulation of early precision diagnosis and treatment strategies for lung cancer.

[0064] Compared to traditional statistical association models that only provide a "list of differentially expressed genes," the SEMO model can interpret more systematic biological meanings from these functional annotations. Through functional enrichment analysis of three characteristic gene sets (benign / malignancy-specific, invasiveness-specific, and common gene sets) selected by the SEMO model, we have profoundly revealed the core biological pathways that coordinate and drive the occurrence and evolution of pulmonary nodules. Benign / malignancy-specific genes (such as APOE and BRCA2) are significantly enriched in cell cycle, DNA damage response, and p53 signaling pathways. These genes mainly regulate the early event of malignant transformation of cells, and their methylation disorders may be the "molecular switch" for normal lung epithelial cells to acquire unlimited proliferative capacity and escape apoptosis. Invasiveness-specific genes (such as ADORA1 and ITPR1) dominate cell adhesion, epithelial-mesenchymal transition (EMT), and Wnt signaling pathways. These genes act on the "next stage" of malignant transformation, enabling tumor cells to acquire invasive and migratory abilities, which is key to nodules breaking through the basement membrane and infiltrating surrounding tissues. Shared genes (such as AKT1 and KRAS) act as the "core engines" driving lung cancer development and progression, enriched in broad-spectrum oncogenic pathways such as PI3K-Akt and MAPK. They continuously play a role in the process from malignant transformation to invasive invasion, and their methylation changes may simultaneously affect the proliferative and invasive characteristics of nodules. Beyond this, the deeper value of this model lies in its ability to provide potential mechanistic interpretations and treatment guidance for individual patients. For example, for a specific subcentimeter lung nodule, if its SEMO characteristics suggest that pathways regulating the cell cycle (such as those associated with BRCA2) and pathways regulating epithelial-mesenchymal transition (EMT) (such as those associated with certain invasive genes) have significant cross-communication through pivotal genes such as AKT1, it can be inferred at the molecular level that this nodule not only has a high risk of malignancy but may also benefit from targeted therapies such as PI3K / Akt / mTOR pathway inhibitors.

[0065] This study aimed to address the clinical challenge of differentiating between benign and malignant pulmonary nodules. By innovatively combining peripheral whole blood DNA methylation sequencing with the SEMO network medical model, we successfully identified a set of methylation network biomarkers comprising 52 genes. This model demonstrated exceptional efficacy in distinguishing between benign and malignant nodules. Its core breakthrough lies in demonstrating that utilizing peripheral whole blood leukocyte DNA methylation combined with network topology analysis can overcome the interference of cellular heterogeneity in traditional liquid biopsies, accurately capturing systemic immune remodeling signals caused by early-stage lung cancer. Its value in early diagnosis provides a novel non-invasive diagnostic strategy for lung cancer screening that is more sensitive than cfDNA, ctDNA, and circulating tumor cell detection, and simpler than PBMC isolation. Although the results of this study show high clinical translational potential, several limitations remain. First, this study was a single-center retrospective design with a limited sample size; therefore, it is necessary to conduct large-scale, multi-center prospective studies in the future to further validate the stability and applicability of the model. Second, the current model is based solely on leukocyte DNA methylation characteristics. In the future, we can consider combining it with artificial intelligence radiomics, clinical data, and other omics data to build a multimodal fusion model to improve the comprehensiveness and accuracy of early pulmonary nodule diagnosis.

[0066] In summary, the gene network model based on peripheral blood leukocyte DNA methylation not only demonstrates excellent performance in differentiating between benign and malignant, and invasive, subcentimeter-sized lung nodules, but also provides a reliable basis for precise risk stratification and preoperative decision-making for lung nodules. This method holds promise as an important supplement to imaging screening, propelling early lung cancer management from "imaging discovery" to "biomarker-driven" precision management.

[0067] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

[0068] References: 1. Siegel RL, Miller KD, Wagle NS, Jemal A. Cancer statistics, 2023. CACancer J Clin. 2023;73(1):17-48. 2、Allemani C, Matsuda T, Di Carlo V, Harewood R, Matz M, Nikšić M, etal. Global surveillance of trends in cancer survival 2000-14 (CONCORD-3):analysis of individual records for 37 513 025 patients diagnosed with one of18 cancers from 322 population-based registries in 71 countries. Lancet.2018;391(10125):1023-75. 3、Adams SJ, Stone E, Baldwin DR, Vliegenthart R, Lee P, FintelmannFJ.Lung cancer screening. Lancet. 2023;401(10374):390-408. 4、Mazzone PJ, Lam L.Evaluating the Patient With a Pulmonary Nodule: AReview. Jama. 2022;327(3):264-73. 5、Aberle DR, Adams AM, Berg CD, Black WC, Clapp JD, Fagerstrom RM, etal. Reduced lung-cancer mortality with low-dose computed tomographicscreening. N Engl J Med. 2011;365(5):395-409. 6、de Koning HJ, Meza R, Plevritis SK, ten Haaf K, Munshi VN, Jeon J,et al. Benefits and harms of computed tomography lung cancer screeningstrategies: a comparative modeling study for the U.S. Preventive ServicesTask Force. Ann Intern Med. 2014;160(5):311-20. 7、Yotsukura M, Asamura H, Motoi N, Kashima J, Yoshida Y, Nakagawa K,et al. Long-Term Prognosis of Patients With Resected Adenocarcinoma In Situand Minimally Invasive Adenocarcinoma of the Lung. J Thorac Oncol. 2021;16(8):1312-20. 8、Li D, Deng C, Wang S, Li Y, Zhang Y, Chen H.Ten-year follow-up oflung cancer patients with resected adenocarcinoma in situ or minimallyinvasive adenocarcinoma: Wedge resection is curative. J Thorac CardiovascSurg. 2022;164(6):1614-22.e1. 9、Ye T, Deng L, Xiang J, Zhang Y, Hu H, Sun Y, et al. Predictors ofPathologic Tumor Invasion and Prognosis for Ground Glass Opacity FeaturedLung Adenocarcinoma. Ann Thorac Surg. 2018;106(6):1682-90. 10、Plaks V, Koopman CD, Werb Z.Cancer. Circulating tumor cells.Science. 2013;341(6151):1186-8. 11、Li S, Zhao B, Wang X, Yu J, Yan S, Lv C, et al. Overestimatedvalue of (18)F-FDG PET / CT to diagnose pulmonary nodules: Analysis of 298patients. Clin Radiol. 2014;69(8):e352-7. 12、Seemann MD, Beinert T, Fürst H, Fink U.An evaluation of the tumourmarkers, carcinoembryonic antigen (CEA), cytokeratin marker (CYFRA 21-1) andneuron-specific enolase (NSE) in the differentiation of malignant from benignsolitary pulmonary lesions. Lung Cancer. 1999;26(3):149-55. 13、Jones PA.Functions of DNA methylation: islands, start sites, genebodies and beyond. Nat Rev Genet. 2012;13(7):484-92. 14、Berman BP, Weisenberger DJ, Aman JF, Hinoue T, Ramjan Z, Liu Y, etal. Regions of focal DNA hypermethylation and long-range hypomethylation incolorectal cancer coincide with nuclear lamina-associated domains. Nat Genet.2011;44(1):40-6. 15、Bert SA, Robinson MD, Strbenac D, Statham AL, Song JZ, Hulf T, etal. Regional activation of the cancer genome by long-range epigeneticremodeling. Cancer Cell. 2013;23(1):9-22. 16、Pfeifer GP, Tang M, Denissenko MF.Mutation hotspots and DNAmethylation. Curr Top Microbiol Immunol. 2000;249:1-19. 17、Greenblatt MS, Bennett WP, Hollstein M, Harris CC.Mutations in thep53 tumor suppressor gene: clues to cancer etiology and molecularpathogenesis. Cancer Res. 1994;54(18):4855-78. 18、Kerr KM, Galler JS, Hagen JA, Laird PW, Laird-Offringa IA.The roleof DNA methylation in the development and progression of lung adenocarcinoma.Dis Markers. 2007;23(1-2):5-30. 19、Liang W, Zhao Y, Huang W, Gao Y, Xu W, Tao J, et al. Non-invasivediagnosis of early-stage lung cancer using high-throughput targeted DNAmethylation sequencing of circulating tumor DNA (ctDNA). Theranostics. 2019;9(7):2056-70. 20、Markou Α, Londra D, Tserpeli V, Kollias Ι, Tsaroucha E,Vamvakaris I, et al. DNA methylation analysis of tumor suppressor genes inliquid biopsy components of early stage NSCLC: a promising tool for earlydetection. Clin Epigenetics. 2022;14(1):61. 21、Zhao Y, O'Keefe CM, Hsieh K, Cope L, Joyce SC, Pisanic TR, et al.Multiplex Digital Methylation-Specific PCR for Noninvasive Screening of LungCancer. Adv Sci (Weinh). 2023;10(16):e2206518. 22、Ji XY, Li H, Chen HH, Lin J.Diagnostic performance of RASSF1A andSHOX2 methylation combined with EGFR mutations for differentiation betweensmall pulmonary nodules. J Cancer Res Clin Oncol. 2023;149(11):8557-71. 23、Gaga M, Chorostowska-Wynimko J, Horváth I, Tammemagi MC, ShitritD, Eisenberg VH, et al. Validation of Lung EpiCheck, a novel methylation-based blood assay, for the detection of lung cancer in European and Chinesehigh-risk individuals. Eur Respir J. 2021;57(1). 24、do Valle IF, Roweth HG, Malloy MW, Moco S, Barron D, Battinelli E,et al. Network medicine framework shows that proximity of polyphenol targetsand disease proteins predicts therapeutic effects of polyphenols. Nat Food.2021;2(3):143-55. 25、Nogales C, Mamdouh ZM, List M, Kiel C, Casas AI, Schmidt H.Networkpharmacology: curing causal mechanisms instead of treating symptoms. TrendsPharmacol Sci. 2022;43(2):136-50. 26、Hoel EP, Albantakis L, Tononi G.Quantifying causal emergence showsthat macro can beat micro. Proc Natl Acad Sci U S A. 2013;110(49):19790-5. 27、Ruiz C, Zitnik M, Leskovec J.Identification of disease treatmentmechanisms through the multiscale interactome. Nat Commun. 2021;12(1):1796.

Claims

1. A methylation marker for discriminating between benign and malignant lung nodules, characterized in that, The methylation biomarkers are key node genes with network topological significance, selected based on a Selective Epigenomic Module Optimization (SEMO) model of pre-trained PPI sub-networks; the methylation biomarkers are selected from at least one of the following genes: APOE, BAD, BCL6, BRCA2, CABP1, CREB1, CXCL12, DNMT3B, DNTT, FLT1, FOXB1, FOXG1, FUS, GHSR, GLI3, HES1, HMGB2, IL6R, IRS2, ITPR3, KLF5, LAMA3, LCK, LEF1, MAP2K1, MEIS2, NEUR OD1, NFATC1, PAX6, PAXIP1, PIM1, PINK1, PLCG2, PPP1CB, PPP1R14A, PRKG1, PRKG2, PTK2, PTK6, PTPN1, RAC3, RARG, RGS2, RUNX3, SLC11A1, SRF, THRB, TYK2, UBB, UBE2K, UNG, VCAM1.

2. A combination of methylation markers for discriminating between benign and malignant lung nodules, characterized in that, The methylation biomarker set comprises a gene set that constructs a specific signal transduction network, the gene set consisting of the 52 genes described in claim 1; the gene set is functionally significantly enriched in at least one of the following signaling pathways: GO_REGULATION_OF_GROWTH (growth regulation), GO_REGULATION_OF_NEUROGENESIS (neurogenesis regulation), GO_FOREBRAIN_DEVELOPMENT (forebrain development), GO_TRANSCRIPTION_COREGULATOR_BINDING (transcriptional co-regulatory factor binding), GO_CATION_TRANSMEMBRANE_TRA NSPORT (cation transmembrane transport), GO_GROWTH (growth), GO_NEGATIVE_REGULATION_OF_TRANSCRIPTION_BY_RNA_POLYMERASE_II (negative regulation of RNA polymerase II transcription), GO_TRANSFERASE_ACTIVITY_TRANSFERRING_PHOSPHORUS_CONTAINING_GROUPS (phosphotransferase activity), KEGG_PATHWAYS_IN_CANCER (cancer pathway), and GO_IMMUNE_EFFECTOR_PROCESS (immune effector process).

3. A methylation marker for discriminating the invasiveness of a lung nodule, characterized by, The methylation marker is selected from at least one of the following genes: ADORA1, ADRA1A, APAF1, ASPM, ATG12, BCR, FLCN, GANAB, GNAL, GNAS, GSTP1, HRH1, IL12B, IL12, RB1, IL15, ITPR1, MAVS, MLEC, MLST8, NAG LU, PAWR, PENK, PLCB2, PRKAA2, PRKAG3, PRKCG, PRKCZ, RBCK1, SCNN1A, SCNN1G, SEC24B, TIRAP, TLR2, TLR9, TNFAIP3, TRHR, TSC1, WNT5A.

4. A combination of methylation markers for identifying the infiltrative nature of pulmonary nodules, characterized in that, The methylation marker combination comprises the 38 genes from claim 3; the gene combination reflects tumor invasion-related microenvironmental interactions and is functionally significantly enriched in at least one of the following signaling pathways: KEGG_CALCIUM_SIGNALING_PATHWAY (calcium signaling pathway), GO_DEFENSE_RESPONSE_TO_SYMBIONT (defense response to symbiotic organisms), GO_TUMOR_NECROSIS_FACTOR_SUPERFAMILY_CYTOKINE_PRODUCTION (tumor necrosis factor superfamily cytokine production), GO_EPITHELIAL_TUBE_FORMATION (epithelial tubule formation), GO_GLYCOPROTEIN_METABOLIC_PROCESS (glycoprotein metabolism), and GO_NEURAL_TUBE_FORMATION (neural tube formation). The following are listed: GO_TUBE_FORMATION (tube formation), GO_DENDRITIC_TREE (dendritic structure), GO_POSITIVE_REGULATION_OF_INNATE_IMMUNE_RESPONSE (positive regulation of innate immune response), GO_POSITIVE_REGULATION_OF_RESPONSE_TO_BIOTIC_STIMULUS (positive regulation of biostimulation response), GO_TUBE_FORMATION (tube formation), GO_TUMOR_NECROSIS_FACTOR_SUPERFAMILY_CYTOKINE_PRODUCTION (tumor necrosis factor superfamily cytokine production), and GO_POSITIVE_REGULATION_OF_INNATE_IMMUNE_RESPONSE (positive regulation of innate immune response).

5. A methylation marker for differentiating benign, malignant, and infiltrative pulmonary nodules, characterized in that, The methylation marker is selected from at least one of the following genes: ADCY8, AKT1, BMP4, BMP7, BRAF, CALM1, CALM2, CASP8, CD36, CDKN1A, CYBA, DRD2, EP300, ERCC1, FGF8, HRAS, IGF1, KR AS, LEP, MAP3K5, MAPK3, NOD2, PSEN1, PTK2B, RAC1, SHC1, SIRT1, SMAD7, SQSTM1, TGFBR1, TGFBR2, TNF, TSPO, UBE2N.

6. A combination of methylation markers for differentiating benign, malignant, and invasive pulmonary nodules, characterized in that, The methylation marker combination comprises the 34 genes described in claim 5; the gene combination is functionally significantly enriched in at least one of the following signaling pathways: GO_REGULATION_OF_TRANSFERASE_ACTIVITY (regulation of transferase activity), GO_POSITIVE_REGULATION_OF_TRANSFERASE_ACTIVITY (positive regulation of transferase activity), GO_CARTILAGE_DEVELOPMENT (cartilage development), GO_CONNECTIVE_TISSUE_DEVELOPMENT (connective tissue development), and GO_INTRINS. IC_APOPTOTIC_SIGNALING_PATHWAY (intrinsic apoptosis signaling pathway), GO_POSITIVE_REGULATION_OF_CELL_POPULATION_PROLIFERATION (positive regulation of cell population proliferation), GO_SKELETAL_SYSTEM_DEVELOPMENT (skeletal system development), KEGG_PATHWAYS_IN_CANCER (cancer pathway), GO_KINASE_ACTIVITY (kinase activity), and GO_PROTEIN_KINASE_ACTIVITY (protein kinase activity).

7. The use of a reagent for detecting the methylation level of the methylation markers of claim 1, 3 or 5, or the combination of methylation markers of claim 2, 4 or 6, in the preparation of a kit for the auxiliary diagnosis of benign, malignant or invasive pulmonary nodules.

8. A method for constructing a peripheral blood leukocyte DNA methylation gene network model for assessing the risk of benign or malignant pulmonary nodules, characterized in that, Includes the following steps: (1) Collect peripheral blood samples from the subjects and divide them into training set and test set; the peripheral blood samples are from benign and malignant lung nodule samples; the lung nodules are sub-centimeter lung nodules with a diameter ≤8mm; (2) Extracting DNA from peripheral blood leukocytes in peripheral blood samples; (3) Detect the methylation level of the DNA of the peripheral blood leukocytes to obtain methylation data; (4) Construct a feature matrix and a SEMO algorithm model from the methylation data, and screen out benign and malignant methylation markers of lung nodules based on the training set sample data; (5) Validate the effectiveness of the model using test set sample data and determine the methylation markers as described in claim 1 for the final evaluation of benign and malignant lung nodules; The SEMO algorithm model is as follows: in, The average gene score of genome X. S represents the average gene score of genome Y. x 2 S represents the score variance of genome X. Y 2 Let be the score variance of genome Y, n be the number of genes in genome X, and m be the number of genes in genome Y.

9. A risk assessment model for benign and malignant pulmonary nodules, characterized in that, The risk assessment model is constructed according to the method described in claim 8, and includes: 1) Data acquisition module, at least used to acquire sample datasets; 2) Sequencing module, used at least to obtain sequencing data of methylation marker genes; 3) A data alignment module, used at least to align the sequencing data with a reference sequence and determine the methylation result of the markers in the sequencing data based on the alignment result; 4) Result determination module, which is used at least to calculate the predicted score threshold through statistical model analysis and determine whether the sample to be tested is a benign or malignant lung nodule.

10. A method for constructing a peripheral blood leukocyte DNA methylation gene network model for assessing the risk of pulmonary nodule infiltration, characterized in that, Includes the following steps: (a) Collect peripheral blood leukocyte samples from the subjects and divide them into training and testing sets; the peripheral blood samples are from malignant lung nodules; the lung nodules are sub-centimeter lung nodules with a diameter ≤10mm; (b) Extracting DNA from peripheral blood leukocytes in peripheral blood samples; (c) Detect the methylation level of the peripheral blood leukocytes to obtain methylation data; (d) Construct a feature matrix from the methylation data, build a SEMO algorithm model, and screen out invasive methylation markers of lung nodules based on the training set sample data; (e) Validate the effectiveness of the model using test set sample data and determine the methylation markers as described in claim 3 for the final assessment of lung nodule infiltrative prediction; The SEMO algorithm model is as follows: in, The average gene score of genome X. S represents the average gene score of genome Y. x 2 S represents the score variance of genome X. Y 2 Let be the score variance of genome Y, n be the number of genes in genome X, and m be the number of genes in genome Y.

11. A risk assessment model for invasive pulmonary nodules, characterized in that, The risk assessment model is constructed according to the method of claim 10, and includes: a) A data acquisition module, used at least to acquire sample datasets; b) A sequencing module, used at least to obtain sequencing data for methylation marker genes; c) A data alignment module, at least used to align the sequencing data with a reference sequence, and determine the methylation result of the markers in the sequencing data based on the alignment result; and d) Result determination module, which is used at least to calculate the predicted score threshold through statistical model analysis and determine whether the sample to be tested is a non-invasive or invasive lung nodule.

Images