Application of 5-hydroxymethylcytosine and prognosis model thereof in nasopharynx cancer prognosis evaluation

Abstract

The invention relates to the technical field of biology, in particular to 5-hydroxymethylcytosine and application of a prognosis model of 5-hydroxymethylcytosine in nasopharynx cancer prognosis evaluation. According to the method, the prognosis scoring model capable of layering the outcome of the patient is constructed, and the model shows strong distinguishing performance in a training set and a verification set. When it is integrated into a prognostic model together with tumor staging and EBV status, the 5hmC score results in higher calibration accuracy and net clinical benefit in decision curve analysis. According to the findings, a 5hmC prognosis model is determined as a noninvasive marker rich in biological information, can be used for prognosis risk layering of NPC patients, and has potential significance for individualized management and improvement of outcome prediction.

Description

Technical Field

[0001] This invention relates to the field of biotechnology, and in particular to the application of 5-hydroxymethylcytosine and its prognostic model in the prognostic assessment of nasopharyngeal carcinoma. Background Technology

[0002] Nasopharyngeal carcinoma (NPC) is an epithelial malignant tumor originating from the nasopharyngeal mucosa, primarily occurring in the roof and lateral walls of the nasopharynx, especially the pharyngeal recesses. In terms of symptoms, early-stage NPC often does not cause obvious symptoms. Advanced stages may present with tinnitus, hearing loss, nasal congestion, bloody nasal discharge, headache, neck masses, and cranial nerve palsy. Treatment primarily involves radiotherapy, combined with chemotherapy and targeted therapy. Despite advancements in radiotherapy and systemic therapy, a significant proportion of patients, especially those with locally advanced stages, experience recurrence or distant metastasis, leading to poor long-term prognosis.

[0003] Prognostic prediction for nasopharyngeal carcinoma (NPC) involves physicians assessing the condition of NPC patients, predicting and evaluating the future development of the disease or disease, and formulating a tailored treatment plan. Traditional prognostic models based on the tumor-lymph node-metastasis (TNM) staging system provide important risk stratification but fail to capture the biological heterogeneity affecting treatment response and survival. In recent years, various biomarkers, including plasma Epstein-Barr virus (EBV) DNA levels, circulating tumor cells (CTCs), and circulating microRNAs (MICs), have been investigated for prognostic assessment of NPC. Several circulating biomarkers have been studied for risk stratification of NPC, with plasma EBV DNA being the most well-established. While EBV DNA load is associated with tumor burden and treatment response, its prognostic performance is stage-dependent and limited in EBV-negative patients. CTCs and cell-free microRNAs have also been explored but face challenges in standardization, sensitivity, and clinical translation.

[0004] Therefore, there is an urgent need for a non-invasive, robust biomarker rich in biological information to better predict outcomes for NPC patients and guide individualized treatment.

[0005] Epigenetic alterations in cell-free DNA (cfDNA), particularly changes in DNA methylation and hydroxymethylation, play a crucial role in cancer development and progression. Among these, 5-hydroxymethylcytosine (5hmC), a 5-methylcytosine oxidation derivative mediated by the TET enzyme family, has emerged as a key epigenetic modification with unique regulatory functions. Studies have shown that global 5hmC deletion is a marker of various malignancies, including hematologic malignancies and solid tumors. Advances in technologies such as 5hmC-Seal have enabled sensitive analysis of genome-wide 5hmC patterns from circulating cell-free DNA (cfDNA), promoting the development of non-invasive cancer biomarkers. However, few studies have evaluated the prognostic value of circulating 5hmC, and there are no reports of its application in the prognostic assessment of nasopharyngeal carcinoma. Summary of the Invention

[0006] The technical problem to be solved by this invention is to provide the application of 5-hydroxymethylcytosine and its prognostic model in the prognostic assessment of nasopharyngeal carcinoma.

[0007] This invention is implemented as follows: This invention first provides the application of 5-hydroxymethylcytosine in the prognostic assessment of nasopharyngeal carcinoma.

[0008] Furthermore, the prognostic assessment includes prognostic risk stratification for NPC patients.

[0009] Furthermore, the risk stratification reflects the risks associated with total survival.

[0010] This invention also provides the application of the 5-hydroxymethylcytosine prognostic model in the prognostic assessment of nasopharyngeal carcinoma.

[0011] Furthermore, the prognostic model includes a 5hmC score.

[0012] Furthermore, the 5hmC score is obtained by constructing a LASSO-Cox regression model that includes the genes HBQ1, SPAT6L, C1QTNF1, CAPN2, FAM72C, HOXA2, and FTL, resulting in a weighted 5hmC score.

[0013] Specifically, the 5hmC score is obtained by multiplying the 5hmC CPM value by its corresponding Cox regression coefficient and then summing the results: 5hmC score = FAM72C×0.093 + CAPN2×0.007 + HOXA2×0.1015 + SPATA6L×(−0.0079) + HBQ1×0.2083 + C1QTNF1×0.002 + FTL×0.1422.

[0014] Furthermore, the prognostic model also includes a prognostic nomogram constructed by integrating 5hmC score, tumor stage, and EBV DNA status.

[0015] Furthermore, the method for constructing the prognostic model includes the following steps: (1) A LASSO-Cox regression model containing HBQ1, SPATA6L, C1QTNF1, CAPN2, FAM72C, HOXA2 and FTL genes was constructed in the training set, and a 5hmC score was established and the 5hmC score value of each patient was calculated. (2) The optimal cutoff value for the prognostic score was determined by analyzing the training set using X-Tile software. Patients were divided into high-risk and low-risk groups based on the optimal cutoff value. The mortality rate of patients in the high-risk group was significantly higher than that of the low-risk group, and the overall survival of patients in the high-risk group was significantly worse than that of patients in the low-risk group. When a patient's 5hmC score is greater than the optimal cutoff value, the patient belongs to the high group; If a patient's 5hmC score is less than or equal to the optimal cutoff value, the patient belongs to the low group; (3) A prognostic nomogram was constructed by integrating 5hmC score, tumor stage and EBV status, and a corresponding calibration plot was generated to assess the consistency between the predicted probability and the observed results.

[0016] Furthermore, the optimal cutoff value in step (2) is 1.21.

[0017] The present invention has the following advantages: This invention identified a seven-gene 5hmC feature using multivariate Cox and LASSO regression and constructed a prognostic scoring model capable of stratifying patient outcomes. This model demonstrated strong discriminative performance on both the training and validation sets. When integrated with tumor stage and EBV status into the prognostic model, the 5hmC score produced high calibration accuracy and net clinical benefit in decision curve analysis. These findings establish the 5hmC prognostic model as a non-invasive and biologically rich marker for prognostic risk stratification in NPC patients, with potential implications for individualized management and improved outcome prediction. Attached Figure Description

[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0019] Figure 1Differential 5hmC profiles in cfDNA from surviving and deceased NPC patients. (A) Volcano plot showing differentially hydroxymethylated genes (FDR < 0.05) between surviving and deceased NPC patients. (B) Heatmap showing the top 15 differentially hydroxymethylated genes. Box plot on the right illustrates the intergroup 5hmC level for each gene. (C) GO biological process enrichment analysis of high and low hydroxymethylated genes.

[0020] Figure 2 The establishment of a 5hmC-based prognostic model for NPC patients. (A) Flowchart depicts the randomization of 174 NPC patients to the training set (n = 105) and the test set (n = 69). (B) Pie chart illustrates the distribution of 673 favorable genes and 1,365 unfavorable genes identified by multivariate Cox regression (adjusted for sex, age, EBV status, and tumor stage) in the training set. (C) Partial likelihood bias curve for 10-fold cross-validation of the LASSO-Cox model. The vertical dashed line indicates the optimal lambda (0.1681327), at which point the seven genes retain non-zero Cox coefficients. (D) Weighted gene coefficients in the final seven-gene prognostic model (left plot) and a heatmap showing the 5hmC Z-scores for these genes in the training set (middle plot). The bottom plot shows the distribution of 5hmC scores, with a cutoff value of 1.21, dividing patients into high and low groups.

[0021] Figure 3 The performance of the 5hmC score in the training and test sets. (A) Distribution of 5hmC scores and corresponding survival status in the training set. The dashed line marks the cutoff value of 1.21. (B) Time-dependent ROC curves of 5hmC score prediction for 3-year (AUC = 0.83) and 5-year (AUC = 0.87) overall survival in the training set. (C) Kaplan-Meier survival curves comparing the high and low groups in the training set. Patients with a 5hmC score > 1.21 (high) had significantly worse overall survival than patients with a score ≤ 1.21 (low). (DF) A similar chart for the test set illustrating the distribution of the 5hmC score (D), the AUC time-dependent ROC curves for 3-year and 5-year survival (E), and the Kaplan-Meier survival curves stratified by the 1.21 cutoff value (F).

[0022] Figure 4The prognostic value of the 5hmC score in clinical subgroups. (A, B) Kaplan-Meier survival curves for all patients with stage III (A) or IV (B) tumors, grouped by the optimal cutoff value of 1.21 for 5hmC scores higher (high) or lower / equal (low). (C, D) Kaplan-Meier curves for EBV-negative (C) and EBV-positive (D) patients, stratified by 5hmC score: high group vs. low group.

[0023] Figure 5 Development and evaluation of nomograms integrating 5hmC score, tumor stage, and EBV status. (A) Nonograph predicting 5-year overall survival (OS) in NPC patients. (B) Calibration plot of the nomogram demonstrating good agreement between the OS probability predicted by the nomogram (x-axis) and the observed OS (y-axis). (C) Decision curve analysis (DCA) comparing the net clinical benefit of the nomogram with models based on single predictors in terms of 5-year OS. Detailed Implementation

[0024] The technical solution of the present invention will now be clearly and completely described in conjunction with the accompanying drawings and specific embodiments. Unless otherwise specified in the embodiments, conditions are performed according to conventional conditions or conditions recommended by the manufacturer. Reagents or instruments used, unless otherwise specified, are all commercially available conventional products.

[0025] 1. Method 1.1 Study participants, sample collection, and EBV DNA testing This study included 174 newly diagnosed, histologically confirmed patients with non-paraneoplastic angiogenesis (NPC) recruited at Fujian Cancer Hospital between September 2017 and September 2018. Eligible participants were required to be treatment-free and have no history of other malignancies at the time of diagnosis. Patients with diabetes mellitus were excluded to avoid potential confounding effects of hyperglycemia on the DNA 5-hydroxymethylome. All participants provided written informed consent, and the study protocol was approved by the institution's ethics committee. Disease staging was determined according to the American Joint Committee on Cancer (AJCC) Stages 8, published in 2017. Peripheral blood samples (10 mL) were collected into K2-EDTA vacuum blood collection tubes (BD, USA; two tubes per patient, if possible) and processed within 2 hours of venipuncture (median approximately 60 minutes). Plasma was separated by two-step centrifugation (1,600 × g, 10 min, 4°C; then 16,000 × g, 10 min, 4°C) to remove cellular debris. Plasma was aliquoted (1.0 mL) to minimize future freeze-thaw cycles and stored at -80°C; hemolyzed samples were excluded. DNA for EBV detection was extracted from the same plasma aliquot and quantified by reverse transcription quantitative PCR (RT-qPCR). The detection limit for plasma was set at 500 EBV genome copies / mL. Patients were subsequently classified as EBV-positive or EBV-negative based on the detectability of plasma EBV DNA levels. The demographic and clinical characteristics of the study cohort, including sex, age, EBV DNA status, and tumor stage, were well-balanced between the training and test sets, with no statistically significant differences.

[0026] 1.2 Sample preparation, 5hmC-Seal spectral analysis and data processing Methods for cfDNA extraction, 5hmC-Seal library construction, sequencing, and data processing are referenced in the literature (Li W, Zhang X, Lu X, You L, Song Y, Luo Z, et al. 5-Hydroxymethylcytosine signatures in circulating cell-free DNA as diagnostic biomarkers for human cancers. Cell Research. 2017;27:1243-57.). For each subject, cfDNA was extracted using the Quick-cfDNA Serum and Plasma Kit (ZYMO) using 2–4 mL double-centrifuged plasma aliquots. Samples meeting the preset QC thresholds were retained: read depth ≥30M paired ends, Q30 ≥85%, alignment ≥92%, repetition ≤60%, gene region proportion ≥50%, fragment length mode -170 bp, and number of detected genes ≥10,000 (CPM>0). Libraries with an alignment below 90% or a repetition above 70% were considered hard failures; intermediate deviations were marked and included in sensitivity analysis. Mitochondrial / blacklist and chrX / Y chromosome peaks were excluded from the common peak set. The 5hmC-Seal library was prepared using a chemical labeling technique targeting 5-hydroxymethylcytosine (Song CX, Szulwach KE, Fu Y, Dai Q, Yi C, Li X, et al. Selective chemical labeling reveals the genome-wide distribution of 5-hydroxymethylcytosine. Nature Biotechnology. 2011;29:68-72.). Paired-end 38 bp sequencing was performed on an Illumina NextSeq 500 platform. Sequencing reads were aligned to the human reference genome (hg19), and 5hmC signaling was quantified by summarizing reads overlapping with the annotated genome. For genome localization summaries, the promoter proximal window (transcription start site ± 2 kb) and enhancer proximal regions (gene distality, supported by H3K27ac in the reference annotation) were annotated post-hoc.The raw counts were normalized to counts per million (CPM) using the "cpm" function in the "edgeR" package (Robinson MD, McCarthy DJ, Smyth GK. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics (Oxford, England). 2010;26:139-40.).

[0027] 1.3 Treatment regimen, follow-up, and outcome assessment Stage I patients receive radiotherapy alone, while Stage II patients receive concurrent chemoradiotherapy. For Stage III-IV disease, radiotherapy is administered after platinum-based induction chemotherapy. Intensity-modulated radiotherapy (IMRT) is used with concurrent booster doses: 66-70 Gy / 31-35 fractions (2.1-2.0 Gy / fraction) for gross tumor volume (primary lesion and affected lymph nodes), 54-56 Gy for high-risk clinical target areas, and 50-54 Gy for low-risk / selective lymph node areas; daily image guidance is used, and organ-at-risk restraint follows the QUANTEC guidelines. Concurrent chemoradiotherapy (Stage II) includes cisplatin 80 mg / m² every 3 weeks during IMRT (as determined by the physician). Induction chemotherapy (stages III-IV) typically consists of 2-3 cycles of gemcitabine 800-1,000 mg / m² (days 1 and 8) plus cisplatin 80-100 mg / m² (day 1), every 21 days (GP). Dosage adjustments follow institutional guidelines based on CTCAE.

[0028] Following initial treatment, patients were systematically followed up through regular outpatient visits. Assessments included physical examination, nasopharyngoscopy, MRI of the nasopharynx and neck, CT scan of the chest, ultrasound or CT scan of the abdomen, and bone scan (if clinically indicated). Follow-up was conducted every 3 months for the first two years, every 6 months from years 3 to 5, and annually thereafter. The primary endpoint was overall survival (OS), defined as the interval from diagnosis to death from any cause or the last follow-up. Event-free survival (EFS) was also assessed, calculated from diagnosis to the occurrence of recurrence, progression, a second malignancy, or death (whichever occurred first). The study follow-up cutoff date was February 18, 2024, and all patient outcomes were assessed up to this date. Patients who did not experience events were censored at their last known follow-up. As of the censoring date, 52 deaths and 62 EFS events had occurred. The median follow-up time was 68.9 months. Clinical data and survival outcomes were independently reviewed by two investigators to ensure data integrity.

[0029] 1.4 Differential Hydroxymethylation Analysis Patients were stratified based on survival status (survival vs. death) at the last follow-up, baseline EBV status (negative vs. positive), and baseline clinical stage (stage I / II vs. stage III / IV). Differential hydroxymethylation between groups was analyzed using the Wilcoxon rank-sum test, and p-values ​​were corrected using multiple tests via the Benjamini and Hochberg procedure (Li Y, Ge X, Peng F, Li W, Li JJ. Exaggerated false positives by popular differential expression methods when analyzing human population samples. Genome biology. 2022;23:79.). Genes with a false discovery rate (FDR) < 0.05 were considered significantly differentially hydroxymethylated genes. Gene Ontology (GO) enrichment analysis was performed using the clusterProfiler package to characterize the biological functions associated with differentially hydroxymethylated genes. Pathways with corrected p-values ​​< 0.05 were considered significantly enriched.

[0030] 1.5 Establishment of OS-weighted prognostic score A total of 174 NPC patients were randomly assigned to the training and test sets in a 6:4 ratio. To develop a prognostic model focused on overall survival (OS), the raw counts were independently converted to CPM for each set to prevent unintentional data leakage. In the training set, a multivariate Cox proportional hazards model was fitted for each gene, incorporating sex, age, EBV status, and tumor stage as covariates. Genes associated with patient OS and with p-values ​​less than 0.05 were retained for subsequent feature selection. Subsequently, these candidate genes were subjected to 10-fold cross-validation using the LASSO (Laminate Absolute Shrinkage and Selection) Cox model with the cv.glmnet function from the glmnet R package (Friedman J, Hastie T, Tibshirani R. Regularization Paths for Generalized Linear Models via Coordinate Descent. Journal of statisticalsoftware. 2010;33:1-22.). This step determined the optimal lambda value, striking a balance between model complexity and predictive accuracy regarding OS. Third, using a determined lambda value, the LASSO-Cox regression model containing all selected candidate genes was retrained on the entire training set to build the final model. This model was constructed as a weighted prognostic score, called the 5hmC score. The score for each patient was obtained by summing the 5hmC CPM value multiplied by its corresponding Cox regression coefficient, thus reflecting the risk associated with overall survival (OS). The optimal cutoff value for the prognostic score was determined by analysis within the training set using X-Tile software (Camp RL, Dolled-Filhart M, Rimm DL. X-tile: a new bio-informatics tool for biomarker assessment and outcome-based cut-point optimization. Clinicalcancer research: an official journal of the American Association for Cancer Research. 2004;10:7252-9.). The cutoff value derived from X-Tile (1.21) was then pre-locked and applied unchanged to the test set and all subgroup analyses to avoid overfitting. Finally, time-dependent receiver operating characteristic (ROC) curve analysis was performed on the test set using the "timeROC" package to evaluate the performance of the prognostic score and determine its effectiveness in stratifying patients based on OS.

[0031] 1.6 Statistical Analysis All statistical analyses and data visualizations were performed using R software (version 4.4.2). The Kaplan-Meier method was used to estimate the survival distributions of subgroups, and the log-rank test was used to assess differences. A multivariate Cox proportional hazards regression model was constructed incorporating 5hmC score groups, tumor stage, and EBV DNA status. Prognostic nomograms were then constructed using the Cox regression coefficients, and corresponding calibration plots were generated to assess the consistency between predicted probabilities and observed outcomes. Both nomograms and calibration plots were created using the "rms" package. Time-dependent ROC curves and corresponding area under the curve (AUC) values ​​were generated using the "timeROC" package to evaluate the discriminative performance of the prognostic models. Decision curve analysis (DCA) was performed using the "dcurves" package (https: / / www.danieldsjoberg.com / dcurves / ) to quantify the net clinical benefit of different modeling approaches.

[0032] 2 Results 2.1 5hmC in cfDNA reflects patient survival status To determine whether cfDNA 5hmC profiles reflect patient survival status, we compared the 5hmC profiles of patients who were alive (“survivors”) and those who died (“deceased”) at the last follow-up. Differential hydroxymethylation analysis showed that, compared with the survivor group, the deceased group had 2,763 genes with high hydroxymethylation and 2,444 genes with low hydroxymethylation. Figure 1 A). Heatmap of the top 15 differentially hydroxymethylated genes ( Figure 1 (B) shows the different 5hmC patterns between the two survival groups. Functional enrichment analysis indicated that hyperhydroxymethylated genes were associated with processes such as angiogenesis, positive regulation of inflammatory responses, cell fate determination, extracellular matrix organization, and epithelial morphogenesis, while hypohydroxymethylated genes were enriched in terms of negative regulation of the cell cycle G2 / M phase transition, organelle division, chromosome segregation, and cytoskeleton-dependent intracellular transport. Figure 1 C). In contrast, no significant differences in 5hmC were observed when patients were stratified by EBV DNA status or tumor stage. Overall, these findings suggest that cfDNA 5hmC profiles differ significantly based on NPC survival outcomes.

[0033] 2.2 Establishment of the NPC 5hmC prognostic model Considering the correlation between cfDNA 5hmC modification and patient survival, we developed a prognostic model based on cfDNA 5hmC to stratify NPC patients. A total of 174 patients were randomly assigned in a 6:4 ratio to the training set (n = 105) and the test set (n = 69). Figure 2 A). In the training set, we fitted a multivariate Cox proportional hazards model for each gene, adjusted for sex, age, EBV status, and tumor stage. This initial screening identified 673 genes with favorable prognostic outcomes and 1,365 genes with unfavorable prognostic outcomes. Figure 2 B). Next, we applied the 10-fold cross-validation LASSO-Cox procedure and determined the optimal lambda value to be 0.1681327 (B). Figure 2 C). Using this optimal lambda, we constructed a final LASSO-Cox regression model containing seven genes, obtaining a weighted 5hmC score (C). Figure 2 D). This seven-gene 5hmC score includes HBQ1, SPATA6L, C1QTNF1, CAPN2, FAM72C, HOXA2, and FTL. We defined the optimal cutoff value for the 5hmC score as 1.21, with values ​​above 1.21 assigned to the high group and values ​​equal to or below 1.21 assigned to the low group.

[0034] Then, we evaluated the prognostic performance of this 5hmC score on the training and test sets. Figure 3 In the training set, the mortality rate in the high-risk group was significantly higher than that in the low-risk group (75% vs. 14.3%). Figure 3 A). Time-dependent ROC curve analysis showed that the AUCs for predicting 3-year and 5-year overall survival by the 5hmC score were 0.83 and 0.87, respectively. Figure 3 B). Consistently, Kaplan-Meier analysis showed that the overall survival of the high-survival group was significantly worse than that of the low-survival group. Figure 3 C). A similar pattern was observed in the test set, where the mortality rates in the high-risk group were 58.3% vs. 13.3% (C). Figure 3 D), and the corresponding 3-year and 5-year time-dependent ROC AUCs were 0.78 and 0.80, respectively. Figure 3 E). Kaplan-Meier analysis on the test set also confirmed the strong predictive power of the 5hmC score (E). Figure 3 (F). Furthermore, consistent patterns were observed in both OS and EFS when a comprehensive analysis was performed on the entire cohort of 174 patients. Overall, these findings confirm that our 5hmC-based prognostic model accurately stratifies NPC patients based on overall survival and demonstrates robust predictive performance on both the training and test sets.

[0035] 2.3 Performance of 5hmC score in different clinical subgroups To further validate the clinical applicability of the 5hmC score, we evaluated its prognostic performance in different patient subgroups stratified by tumor stage and EBV status. Throughout the entire cohort of 174 NPC patients, the seven-gene 5hmC score consistently differentiated high-scoring patients with significantly worse overall survival from low-scoring patients, regardless of stage III ( Figure 4 A) or is it in stage IV disease ( Figure 4 B). Similarly, in EBV negative ( Figure 4 C) and EBV positive ( Figure 4 In patients (D), a high 5hmC score was associated with poorer survival outcomes. These findings suggest that the prognostic information provided by the 5hmC score is independent of disease stage and EBV status. Indeed, 5hmC-based models achieved higher prognostic accuracy compared to using tumor stage or EBV status alone, demonstrating their potential to add prognostic value to traditional clinicopathological features.

[0036] 2.4 Construction of a nomogram integrating 5hmC score and clinical characteristics To provide a quantitative method for predicting overall survival in NPC patients, we constructed a nomogram integrating 5hmC score grouping, tumor stage, and EBV status. Figure 5 A). The calibration plot shows good agreement between the predicted probabilities of the nomogram and the observed results. Figure 5 B). Notably, this multivariate nomogram achieved a 5-year AUC of 0.843 and a 5-year C-index of 0.796, superior to any single clinical feature. Furthermore, DCA results indicated that combining 5hmC score grouping with tumor stage and EBV status yielded a higher net benefit compared to using either feature alone. Figure 5 C). These findings highlight the strong prognostic capability of the 5hmC score when integrated into clinically applicable predictive models.

[0037] 3. Summary The above studies demonstrate that cfDNA 5hmC profiling is significantly associated with overall survival in NPC patients. Using a well-defined cohort, we identified a seven-gene 5hmC signature using multivariate Cox and LASSO regressions and constructed a prognostic score capable of stratifying patient outcomes. The model exhibited strong discriminative performance on both the training and validation sets, with AUC values ​​approaching or exceeding 0.80 for 3-year and 5-year survival predictions. Notably, these survival-related 5hmC alterations were independent of baseline EBV DNA status or tumor stage, suggesting that 5hmC modifications provide additional prognostic information beyond traditional clinical indicators. When integrated with stage and EBV status into a nomogram, the 5hmC score yielded high calibration accuracy and net clinical benefit in decision curve analysis. These findings establish the practicality of cfDNA 5hmC profiling analysis as a non-invasive and biologically rich method for prognostic risk stratification in NPC patients.

[0038] Studies have shown that the cfDNA 5hmC profile not only complements traditional biomarkers but also independently predicts survival outcomes, even within the EBV-positive and early-stage subgroups. The seven-gene 5hmC score constructed in this invention provides fine prognostic stratification across clinical subtypes and outperforms models based on individual clinical stages or EBV status, particularly when integrated into a nomogram framework.

[0039] This invention introduces a clinically relevant tool for non-invasive risk stratification of NPC (tumor necrosis) through the development of a cfDNA 5hmC-based prognostic score. Unlike traditional prognostic indicators that reflect anatomical extent or viral load (such as TNM stage or EBV DNA load), the 5hmC atlas captures transcriptional and epigenetic alterations relevant to tumor biology. The score identified in this study complements existing prognostic systems and has demonstrated utility in clinical subgroups, including EBV-negative and early-stage patients. Its integration into a composite nomogram improves survival prediction and suggests its potential to inform decisions regarding treatment intensity or monitoring frequency. Importantly, the 5hmC-Seal assay can be performed on peripheral blood, is highly technically reproducible, and minimally invasive, supporting its suitability for clinical translation.

[0040] In practical applications, using a 1.21 high / low cutoff value combined with individualized nomogram probabilities allows for anticipation of decision points such as risk assessment, in-treatment adjustments, and post-treatment monitoring. At diagnosis, pre-treatment plasma sampling can guide intensive treatment for high-risk patients (e.g., induction chemotherapy or trial enrollment) or support standard treatment for low-risk patients. During treatment, repeated sampling can detect elevated 5hmC scores indicating resistance and prompt treatment adjustments, while a decrease in the score may justify a downgrade strategy. Post-treatment, longitudinal monitoring can optimize imaging intervals; elevated scores trigger early assessment, while stable low scores allow for standard follow-up. Because 5hmC reflects dynamic genetic regulation, continuous profiling analysis enables real-time monitoring of treatment response or early detection of relapse.

[0041] While specific embodiments of the present invention have been described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and not intended to limit the scope of the present invention. Equivalent modifications and variations made by those skilled in the art in accordance with the spirit of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

Application of 1,5-hydroxymethylcytosine in prognostic assessment of nasopharyngeal carcinoma.

2. The application according to claim 1, characterized in that: The prognostic assessment includes prognostic risk stratification for NPC patients.

3. The application according to claim 2, characterized in that: The risk stratification reflects the risks associated with overall survival. Application of the 4,5-hydroxymethylcytosine prognostic model in the prognostic assessment of nasopharyngeal carcinoma.

5. The application according to claim 4, characterized in that: The prognostic model includes the 5hmC score.

6. The application according to claim 5, characterized in that: The 5hmC score is obtained by constructing a LASSO-Cox regression model that includes the genes HBQ1, SPAT6L, C1QTNF1, CAPN2, FAM72C, HOXA2, and FTL, resulting in a weighted 5hmC score.

7. The application according to claim 6, characterized in that: The 5hmC score is obtained by multiplying the 5hmC CPM value by its corresponding Cox regression coefficient and then summing the results: 5hmC score = FAM72C×0.093 + CAPN2×0.007 + HOXA2×0.1015 + SPATA6L×(−0.0079) + HBQ1×0.2083 + C1QTNF1×0.002 + FTL×0.1422.

8. The application according to claim 5, characterized in that: The prognostic model also includes a prognostic nomogram, constructed by integrating 5hmC score, tumor stage, and EBV DNA status.

9. The application according to claim 5, characterized in that: The method for constructing the prognostic model includes the following steps: (1) A LASSO-Cox regression model containing HBQ1, SPATA6L, C1QTNF1, CAPN2, FAM72C, HOXA2 and FTL genes was constructed in the training set, and a 5hmC score was established and the 5hmC score value of each patient was calculated. (2) The optimal cutoff value for the prognostic score was determined by analyzing the training set using X-Tile software. Patients were divided into high-risk and low-risk groups based on the optimal cutoff value. The mortality rate of patients in the high-risk group was significantly higher than that of the low-risk group, and the overall survival of patients in the high-risk group was significantly worse than that of patients in the low-risk group. When a patient's 5hmC score is greater than the optimal cutoff value, the patient belongs to the high group; If a patient's 5hmC score is less than or equal to the optimal cutoff value, the patient belongs to the low group. (3) A prognostic nomogram was constructed by integrating 5hmC score, tumor stage and EBV status, and a corresponding calibration plot was generated to assess the consistency between the predicted probability and the observed results.

10. The application according to claim 9, characterized in that: The optimal cutoff value in step (2) is 1.21.

Images