Skip to main content
NCBI home page
Laryngoscope Investigative Otolaryngology logoLink to Laryngoscope Investigative Otolaryngology
. 2026 Jun 7;11(3):e70461. doi: 10.1002/lio2.70461

A DNA Methylation Signature Predicts Survival and Platinum Response in HNSCC

Riya Chhabra 1,2, Alfred Kao 1,2, Omar Mokhashi 1,2, Wei Tse Li 1,2,3, Daniel John 1,2, Jessica Wang‐Rodriguez 4,5, Weg M Ongkeko 1,2,✉
PMCID: PMC13243778  PMID: 42266439

ABSTRACT

Objective

We aimed to develop and validate a DNA methylation–based risk score to identify patients at higher risk of poor survival and platinum non‐response.

Methods

Genome‐wide DNA methylation and clinical data, including treatment response and survival, were analyzed from platinum‐treated patients with head and neck squamous cell carcinoma in The Cancer Genome Atlas (n = 153). A weighted DNA methylation risk score was constructed from response‐associated methylation features and evaluated for associations with overall survival and platinum treatment response. Prognostic performance was assessed using Kaplan–Meier analysis and multivariable Cox regression. Predictive performance for platinum response was evaluated by comparing models incorporating methylation features with models based on clinical variables alone. External validation was performed in an independent oral rinse–based methylation cohort (GSE52793, n = 82).

Results

The methylation risk score significantly stratified overall survival (HR = 3.03, 95% CI: 1.20–7.65; p = 0.019). Incorporation of the risk score improved prediction of platinum response compared with clinical variables alone in both logistic regression (AUC 0.58 vs. 0.80) and random forest models (AUC 0.66 vs. 0.90). Prognostic performance was preserved in an independent oral rinse–based validation cohort (HR = 1.043; p = 0.019).

Conclusion

A DNA methylation–based risk score stratifies survival and improves prediction of platinum response in head and neck squamous cell carcinoma beyond standard clinical factors. Validation in an oral rinse–based cohort highlights the potential clinical applicability of methylation biomarkers for non‐invasive treatment risk stratification and supports prospective evaluation.

Level of Evidence

3.

Keywords: DNA methylation, epigenetic biomarkers, head and neck squamous cell carcinoma, platinum‐based chemotherapy, prognosis, treatment response prediction

We developed and validated a DNA methylation–based risk score that stratifies overall survival and improves prediction of platinum treatment response in head and neck squamous cell carcinoma (HNSCC). In a TCGA platinum‐treated cohort (n = 153), the methylation risk score independently predicted survival (HR = 3.03, p = 0.019) and significantly enhanced treatment response classification beyond clinical variables (AUC improved from 0.58 to 0.80), with consistent prognostic performance observed in an external oral rinse–based validation cohort (n = 82). These findings support the translational potential of epigenetic biomarkers for non‐invasive risk stratification and treatment personalization in HNSCC.

graphic file with name LIO2-11-e70461-g004.jpg

1. Introduction

Head and neck squamous cell carcinoma (HNSCC) is a major global health burden, resulting in 890,000 new cases and 450,000 deaths annually [1]. Despite advances in surgery, radiation, targeted therapy, and immunotherapy, outcomes for many patients remain unsatisfactory. Platinum‐based chemoradiotherapy continues to serve as a backbone of systemic treatment in HNSCC, especially for patients who are ineligible for surgery [2]. However, clinical benefit is highly variable. While some patients achieve durable responses, others derive little benefit and experience significant toxicity early in treatment [3]. This therapeutic heterogeneity underscores the urgent need for biomarkers that can identify patients most likely to benefit from platinum‐based therapy while sparing others from unnecessary morbidity.

Current approaches to treatment stratification in HNSCC rely largely on clinicopathologic features, HPV status, and selected genomic alterations [4, 5]. Although HPV status is a well‐established prognostic factor in oropharyngeal cancers, it has not consistently functioned as a predictive biomarker of treatment sensitivity [6, 7]. Similarly, genomic and transcriptomic predictors such as TP53 mutations and EGFR alterations have provided biological insights but lack robust and reproducible predictive performance in clinical practice [8, 9, 10]. As a result, reliable molecular biomarkers for platinum treatment risk stratification in HNSCC remain limited.

DNA methylation represents a promising class of biomarkers for treatment stratification. Methylation patterns are relatively stable, can be assessed using routinely collected formalin‐fixed tissue, and are detectable in minimally invasive biospecimens such as saliva and oral rinse [11, 12, 13]. In several malignancies, such as lung, ovarian, and colorectal cancer, methylation‐based signatures have been associated with platinum response and survival, suggesting potential applicability to HNSCC [14, 15, 16].

In this study, we developed a DNA methylation–based risk score using genome‐wide methylation data from platinum‐treated patients with HNSCC. We evaluated its association with overall survival, assessed its ability to improve prediction of platinum response beyond standard clinical factors, and examined its reproducibility in an independent oral rinse–based cohort. Our objective was to assess the potential clinical utility of methylation biomarkers for treatment risk stratification in HNSCC.

2. Methods

2.1. Data Sources and Study Cohorts

Genome‐wide DNA methylation and clinical data were obtained from The Cancer Genome Atlas Head and Neck Squamous Cell Carcinoma (TCGA‐HNSC) cohort, which includes tumors from the oral cavity, oropharynx, larynx, and hypopharynx. Among patients with available treatment and response annotations, 153 patients who received platinum‐based therapy (cisplatin, carboplatin, or oxaliplatin) were included in the primary analysis. Treatment response was classified using TCGA first‐course outcome annotations, with responders defined as “complete or partial response” or “stable disease,” and non‐responders defined as “progressive or persistent disease.”

External validation was performed using an independent cohort (GSE52793) consisting of 82 patients with oral squamous cell carcinoma (OSCC) and available survival outcomes, profiled using Illumina HumanMethylation450 arrays from oral rinse samples. The validation cohort was selected to assess the transportability of a tumor‐derived methylation signature to a non‐invasive biospecimen relevant to clinical application.

2.2. Institutional Review Board Statement

This study utilized publicly available, de‐identified data obtained from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) repositories. Because the data are fully de‐identified and do not contain protected health information, this study does not constitute human subjects research as defined by federal regulations and is therefore exempt from institutional review board (IRB) review.

2.3. Methylation Analysis and Risk Score Development

Differential methylation analysis was performed to identify CpG sites associated with platinum treatment response in the TCGA cohort. A DNA methylation–based risk score was constructed by weighting the top 300 response‐associated CpG sites by their effect sizes and aggregating these values at the patient level. Higher scores reflected increased similarity to the non‐responder methylation profile. All CpG‐specific logFC weights are listed in Table S1.

2.4. Survival and Prognostic Analysis Based on the Methylation Risk Score

Patients were stratified into low‐risk and high‐risk groups using the median methylation risk score.

(MRS), which was evaluated for association with overall survival (OS) and progression‐free interval (PFI) in the TCGA‐HNSC cohort. Kaplan–Meier survival analysis with log‐rank testing was used to compare survival between risk groups. Multivariable Cox proportional hazards regression was performed to estimate hazard ratios and assess the prognostic value of the MRS after adjustment for age, tumor stage, HPV status, smoking history, alcohol history, and tumor subsite.

For external validation, the TCGA‐derived MRS was applied to an independent oral rinse–based cohort (GSE52793) profiled on the same Illumina HumanMethylation450 platform. MRS values were computed using standardized methylation values and CpG weights. Prognostic performance was evaluated using Kaplan–Meier survival analysis based on available overall survival data.

2.5. Prediction of Platinum Treatment Response

2.5.1. Model 1: Methylation Risk Score

To evaluate the predictive value of the methylation risk score, Elastic Net–regularized logistic regression models were trained using a 70% training and 30% test split. Two models were constructed: a clinical‐only model incorporating age, cancer stage, HPV status, alcohol history, and smoking history, and a clinical plus MRS model that included the methylation risk score as an additional predictor. Tumor subsites were excluded from this model due to sparse representation of certain subsites, which can reduce stability in penalized regression frameworks.

Model performance was evaluated in the held‐out test set using receiver operating characteristic (ROC) analysis to compare discrimination between responders and non‐responders.

2.5.2. Model 2: CpG‐Level Methylation Features

A second predictive model was developed using methylation M‐values derived from the top response‐associated CpG sites. β‐values from the top 60 differentially methylated CpG sites were transformed to M‐values prior to model training. A Random Forest classifier was trained using these methylation features in combination with clinical variables (age, stage, HPV status, alcohol history, smoking history) and tumor subsite to capture anatomic heterogeneity within HNSCC.

To address class imbalance, sampling parameters were tuned such that each decision tree was trained on equal numbers of responders and non‐responders. Model performance was assessed using ROC curves and partial area under the curve (AUC) in the test set [17, 18].

2.6. HPV Status‐Associated Differential Methylation of MRS Genes

To examine HPV status‐specific methylation patterns within the MRS, differential methylation analysis was performed comparing HPV‐positive and HPV‐negative tumors across the 300 MRS constituent CpG sites. Linear models were fitted using the limma package in R, with HPV status as the primary covariate. FDR‐adjusted p‐values were computed using the Benjamini‐Hochberg method. Results were summarized at the gene level by reporting the minimum FDR‐adjusted p‐value and mean logFC across all CpG probes mapping to each gene. The analysis was restricted to patients with available methylation and HPV status data from the TCGA‐HNSC cohort.

2.7. Exploratory Correlation of CpG Methylation With Protein Expression

To explore biological relevance, CpG methylation levels were correlated with matched protein expression data for selected resistance‐associated genes, including EGFR, NOTCH1, and NOTCH3, using Spearman correlation. CpG sites were annotated by genomic context, including promoter, enhancer, and 5′ untranslated regions, to evaluate context‐dependent associations.

3. Results

Differential Methylation Associated With Platinum‐Based Therapy Response in.

3.1. TCGA‐HNSC

Genome‐wide comparison of DNA methylation profiles between platinum responders and non‐responders in the TCGA‐HNSC cohort identified multiple CpG sites with significantly altered methylation. A representative volcano plot illustrates the distribution of response‐associated CpGs (Figure 1). Several differentially methylated CpGs mapped to genes with established roles in cancer therapy response, including NOTCH1, EGFR, E2F1, CSMD1, and HLA‐B, which have previously been implicated in DNA repair, apoptosis, immune regulation, and drug sensitivity [19, 20, 21, 22, 23].

FIGURE 1.

FIGURE 1

Open in a new tab

Volcano plot of differential methylation between platin responders and non‐responders in TCGA‐HNSC. Each point represents a CpG probe. Genes with established roles in therapy resistance (e.g., NOTCH1, EGFR, E2F1, CSMD1, HLA‐B) are labeled.

3.2. A Methylation‐Based Risk Score Stratifies Overall Survival

Patients stratified into high‐ and low‐risk groups by median MRS demonstrated significantly different overall survival outcomes. High‐risk patients experienced poorer overall survival compared with low‐risk patients (HR = 3.03, 95% CI: 1.20–7.65, p = 0.019; Figure 2A). In contrast, no statistically significant association was observed between the methylation risk score and progression‐free interval (Figure 2B).

FIGURE 2.

FIGURE 2

Open in a new tab

Kaplan–Meier survival analysis based on methylation risk score in TCGA‐HNSC. (A) Overall survival (OS) is significantly reduced in high‐risk patients (HR = 3.03, p = 0.019). (B) Progression‐free interval (PFI) shows a non‐significant trend (HR = 1.77, p = 0.24). Patients were stratified by median risk score (n = 76 low‐risk, n = 77 high‐risk).

3.3. Multivariable Analysis Confirms Independent Prognostic Value of the Methylation Risk Score

In multivariable analysis adjusting for age, tumor stage, HPV status, smoking history, alcohol history, and tumor subsite, the methylation risk score remained independently associated with overall survival (HR = 1.10 per unit increase, 95% CI: 1.04–1.17, p = 0.0015; Figure 3). None of the included clinical covariates demonstrated a statistically significant association with survival in this model.

FIGURE 3.

FIGURE 3

Open in a new tab

Multivariable Cox regression analysis of overall survival in TCGA‐HNSC. Forest plot shows hazard ratios for the methylation risk score and clinical covariates (age, HPV status, smoking and alcohol history, tumor stage, and anatomical subsite). The risk score remains independently predictive of survival after adjustment.

Inclusion of the methylation risk score improved overall model discrimination (C‐index = 0.73; global log‐rank p = 0.0007), indicating that the epigenetic signature contributed prognostic information beyond conventional clinicopathologic factors.

3.4. External Validation in an Oral Rinse–Based Cohort

To assess external transportability, the methylation risk score was applied to an independent oral rinse–based OSCC cohort (GSE52793, n = 82). The risk score remained significantly associated with overall survival (HR = 1.043 per unit increase, 95% CI: 1.007–1.081, p = 0.019). Kaplan–Meier analysis demonstrated consistent separation between risk groups, although survival probabilities remained above.

50%, reflecting the shorter follow‐up duration and higher baseline survival of this cohort (Figure 4). These findings indicate preservation of the prognostic signal across distinct biospecimen types and clinical contexts.

FIGURE 4.

FIGURE 4

Open in a new tab

External validation of the methylation risk score in GSE52793 (oral rinse cohort). Kaplan–Meier survival curves based on a median split of the logFC‐weighted methylation risk score. A trend toward poorer survival was observed in the high‐risk group compared with the low‐risk group (HR = 1.043 per unit increase, 95% CI: 1.007–1.081, p = 0.019; C‐index = 0.59). Kaplan–Meier analysis demonstrated consistent separation of risk groups, with survival probabilities remaining above 50% due to shorter follow‐up duration and higher baseline survival in this cohort. Patients were stratified by median risk score (n = 41 low‐risk, n = 41 high‐risk).

3.5. Predictive Performance of the Methylation Risk Score

The predictive utility of the methylation risk score for platinum treatment response was evaluated using an internal train–test framework. In the test cohort, a clinical‐only logistic regression model demonstrated limited discrimination between responders and non‐responders (AUC = 0.58). Incorporation of the methylation risk score substantially improved classification performance (AUC = 0.80; Figure 5A).

FIGURE 5.

FIGURE 5

Open in a new tab

Predictive performance of the methylation risk score for platin response. (A) Model using methylation risk scores. ROC curves in the TCGA test set comparing clinical covariates alone (AUC = 0.58) versus clinical covariates plus the methylation score (AUC = 0.80). (B) Random Forest model using methylation M values. ROC curves in the TCGA test set comparing clinical covariates (including tumor subsite) alone (AUC = 0.66) versus clinical covariates plus the methylation score (AUC = 0.90).

Consistent findings were observed using a complementary Random Forest classifier. Clinical covariates alone yielded modest discrimination (AUC = 0.66), whereas inclusion of methylation features markedly improved performance (AUC = 0.90; Figure 5B).

Together, these findings demonstrate that methylation features provide consistent and reproducible predictive value for platinum treatment response beyond established clinical variables.

3.5.1. HPV Status‐Specific Differences in MRS‐Associated Genes

To address the biological heterogeneity of HNSCC by HPV status, exploratory differential methylation analysis was performed across the 300 MRS constituent CpGs comparing HPV‐positive (n = 22) and HPV‐negative (n = 177) tumors in TCGA‐HNSC. Several genes demonstrated highly significant differential methylation by HPV status (Table S2). The most significantly differentially methylated genes included CCND1 (FDR = 6.32 × 10−11, 24 CpG probes), FGFR3 (FDR = 5.68 × 10−10, 19 probes), and NOTCH1 (FDR = 1.01 × 10−7, 30 probes), all of which were hypomethylated in HPV‐positive relative to HPV‐negative tumors. FAT1 (FDR = 1.50 × 10−6, 33 probes) and CSMD1 (FDR = 2.53 × 10−5, 69 probes) were hypermethylated in HPV‐positive tumors. Additional genes showing significant differential methylation included E2F1, EGFR, TP63, NOTCH3, TP53, and CDKN2A (all FDR < 0.01; Table S2). Because this analysis was restricted to CpGs included in the MRS, these results specifically describe HPV‐associated differences among methylation features incorporated into the risk score, rather than genome‐wide HPV‐associated methylation differences.

3.6. Exploratory Associations Between CpG Methylation and Protein Expression

EGFR, NOTCH1, and NOTCH3 were selected for exploratory analysis based on the presence of multiple response‐associated CpG sites identified in the differential methylation analysis. To assess whether genomic context influences methylation–protein relationships, we examined CpG sites located within promoter, 5′UTR, and enhancer regions. Significant CpG–protein associations are summarized in Table 1, with representative examples shown in Figure 6.

TABLE 1.

Spearman's coefficient values for individual CpGs in important regulatory elements with significant coefficient values.

Gene CpG ID Spearman's coefficient p Type Source Coordinate (in bp)
EGFR cg05064645 −0.125 1.826e‐02* 5′ UTR NCBI 55,019,174
EGFR cg20773588 0.158 9.224e‐03** Enhancer NCBI 55,066,407
EGFR cg04625338 −0.273 1.752e‐07*** Enhancer NCBI 55,074,277
EGFR cg01461514 −0.308 3.306e‐09*** Enhancer Ensembl 55,109,488
EGFR cg18809076 −0.301 7.344e‐09*** Enhancer Ensembl 55,109,929
EGFR cg05898452 0.254 1.304e‐06*** Enhancer NCBI 55,157,169
EGFR cg18071865 0.207 8.877e‐05*** Enhancer NCBI 55,157,201
EGFR cg02316066 0.260 6.915e‐07*** Enhancer NCBI 55,157,216
NOTCH1 cg13593436 −0.235 7.526e‐06*** Enhancer NCBI 136,498,168
NOTCH1 cg06338150 −0.170 1.341e‐03** Enhancer NCBI 136,498,442
NOTCH1 cg13468680 −0.229 1.405e‐05*** Enhancer NCBI 136,503,066
NOTCH1 cg05784201 −0.129 1.501e‐02* Enhancer Ensembl + NCBI 136,513,156
NOTCH3 cg03665287 −0.155 3.937e‐03** Promoter NCBI 15,201,074
NOTCH3 cg17944161 −0.248 3.544e‐06*** Promoter NCBI 15,201,178
NOTCH3 cg13404054 −0.265 6.650e‐07*** 5′ UTR NCBI 15,200,854
NOTCH3 cg13673837 −0.244 5.061e‐06*** Enhancer NCBI 15,195,858
Open in a new tab
*

p‐value ≤ 0.05.

**

p‐value ≤ 0.01.

***

p‐value ≤ 0.001.

FIGURE 6.

FIGURE 6

Open in a new tab

Correlation scatterplots for select CpG sites and genes from Table 1. cg01461514 methylation vs. EGFR expression (Spearman's ρ = −0.31, p = 3.3 × 10−9). cg13468680 methylation vs. NOTCH1 expression (Spearman's ρ = −0.24, p = 7.5 × 10−6). cg13404054 methylation vs. NOTCH3 expression (Spearman's ρ = −0.27, p = 6.7 × 10−7).

For EGFR, several enhancer‐associated CpGs demonstrated significant inverse correlations with protein expression, while promoter CpGs did not. NOTCH1 exhibited significant negative correlations exclusively at enhancer‐associated CpGs. In NOTCH3, significant CpGs were identified across promoters, enhancers, and the 5′UTR, all of which showed inverse associations with protein abundance.

Collectively, these findings highlight the site‐specific and context‐dependent nature of methylation–expression relationships, with regulatory elements–associated CpGs more frequently linked to reduced protein expression.

4. Discussion

This study demonstrates that a DNA methylation–based risk score (MRS) is associated with both overall survival and platinum treatment response in head and neck squamous cell carcinoma. By integrating differential methylation analysis with a weighted risk score, we identified an epigenetic signature that stratifies prognosis and improves prediction of platinum responsiveness beyond standard clinicopathologic factors. Importantly, this signal was preserved in an independent oral rinse–based cohort, supporting the potential clinical feasibility of methylation‐based risk stratification. These findings suggest that epigenetic profiling may provide clinically complementary information beyond traditional clinicopathologic risk stratification in HNSCC.

Epigenetic biomarkers have emerged as promising tools for treatment stratification across multiple malignancies. Prior studies in ovarian, lung, and colorectal cancers have demonstrated associations between DNA methylation patterns, platinum response, and survival outcomes [14, 15, 16, 24]. In HNSCC, however, methylation‐based investigations have largely focused on prognosis, HPV‐associated epigenetic changes, or immune‐related regulation rather than direct prediction of platinum treatment response [25, 26, 27]. Our findings extend this literature by demonstrating that methylation‐derived features can inform both prognostic and predictive modeling in platinum‐treated HNSCC.

Previous efforts to predict platinum sensitivity in HNSCC have primarily relied on transcriptomic or mutational biomarkers, including ERCC1 expression and TP53 mutation status [28, 29]. Although biologically informative, these markers have demonstrated inconsistent clinical performance and are limited by temporal variability and intratumoral heterogeneity [8, 9, 10]. In contrast, DNA methylation offers a more stable molecular signal that is compatible with routinely collected clinical specimens, supporting its suitability for translational applications [11, 12, 13, 14].

Exploratory analysis of the MRS‐associated CpGs by HPV status suggested that the risk score may capture biologically distinct methylation patterns across HPV‐positive and HPV‐negative disease. Several MRS‐associated genes, including CCND1, FGFR3, NOTCH1, FAT1, and CSMD1, showed HPV‐associated methylation differences, supporting the possibility that individual features within the composite score may reflect partially different molecular mechanisms across HNSCC subgroups. CCND1 amplification and FGFR3 alterations are well‐established oncogenic drivers enriched in HPV‐negative HNSCC [30, 31]. Their differential methylation within the MRS may therefore reflect broader HPV‐associated molecular differences, although functional validation is needed to determine whether these methylation changes correspond to altered gene expression or pathway activity.

The exploratory methylation–protein correlation analysis provided additional biological context for the MRS. When focusing on regulatory regions, CpG methylation within promoters, enhancers, and 5′UTRs was most often associated with reduced protein expression, consistent with established models of context‐dependent epigenetic regulation [32, 33]. Unlike transcriptomic or mutational markers that may fluctuate with tumor sampling or treatment exposure, DNA methylation may reflect more stable regulatory states that integrate intratumoral heterogeneity over time. These findings support the biological relevance of the identified methylation features and suggest links to pathways involved in DNA repair, apoptosis, and treatment resistance [19, 20, 21].

A notable strength of this study is external validation of the methylation risk score in an oral rinse–based cohort, demonstrating that the prognostic signal of the tumor‐derived signature is preserved across biospecimen types and clinical contexts. Oral rinse sampling offers a non‐invasive, outpatient‐compatible approach that could be readily integrated into routine otolaryngology practice, enabling longitudinal risk assessment without the need for repeat tissue biopsies [12, 34, 35]. This feasibility supports the potential for methylation‐based biomarkers to be incorporated into real‐world surveillance and treatment stratification workflows for patients with HNSCC.

Several limitations warrant consideration. First, external validation was limited to prognostic assessment because detailed platinum treatment‐response annotations were not available in the oral rinse validation cohort. The validation cohort was also modest in size and had shorter follow‐up duration, which may have attenuated the observed effect size. Second, HNSCC is biologically heterogeneous across anatomic subsites and HPV status, and the available sample size limited fully powered subgroup analyses within HPV‐positive, HPV‐negative, and subsite‐specific populations. This interpatient heterogeneity is associated with unique subgroup methylation patterns [36]. Although the methylation risk score remained associated with outcome after adjustment for HPV status and tumor subsite, larger cohorts will be needed to determine whether performance differs across clinically distinct HNSCC subgroups.

Additional limitations relate to biospecimen type and clinical translation. The TCGA methylation data used to develop the risk score were generated from bulk tumor specimens, meaning that the score may reflect an aggregate tumor‐level signal rather than the full diversity of epigenetic states within individual tumors. In contrast, the external validation cohort used oral rinse samples, a non‐invasive biospecimen that may capture tumor‐associated molecular alterations, although detection performance may vary depending on oral sampling method and assay characteristics [35]. HNSCC tumor anatomic sites also influence the method of DNA collection. Saliva‐based liquid biopsies are preferred for OSCC, while blood plasma methods work best with other subsites (e.g., oropharynx) [37, 38]. Broader implementation of methylation‐based biomarkers will also require assay standardization, cost‐effective workflows, and accessible bioinformatics pipelines for clinical interpretation [36, 39, 40].

Future studies should prioritize prospective validation in larger, multi‐institutional cohorts with detailed treatment annotation. Integrative approaches combining methylation with additional molecular modalities have demonstrated improved predictive performance in cancer and may further enhance treatment stratification in HNSCC [41, 42]. Ultimately, incorporation of methylation biomarkers into biomarker‐driven clinical trials may enable more personalized treatment strategies for patients with head and neck squamous cell carcinoma.

5. Conclusion

This study provides proof‐of‐concept that DNA methylation signatures can serve as predictive and prognostic biomarkers for platinum‐based therapy in head and neck squamous cell carcinoma. The methylation risk score stratified survival, improved classification of treatment response beyond clinical features, and demonstrated reproducibility across independent datasets, including a non‐invasive oral rinse cohort. These findings support the translational potential of methylation‐based biomarkers to guide treatment stratification and reduce unnecessary toxicity. Larger, prospectively annotated studies are warranted to validate these findings and to define the role of epigenetic biomarkers in personalized treatment strategies for head and neck cancer.

Funding

This study was funded by the University of California, Academic Senate grant: RG104647.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Differentially methylated CpG sites used in constructing the methylation risk score, including logFC weights, p‐values, and genomic annotations.

LIO2-11-e70461-s002.csv (23.2KB, csv)

Table S2: Differential methylation of MRS‐associated CpG sites between HPV‐positive and HPV‐negative HNSCC.

LIO2-11-e70461-s001.csv (1.1KB, csv)

Acknowledgments

Artificial intelligence–assisted tools were used for language editing and manuscript formatting support. All scientific content, analyses, interpretations, and conclusions were developed, reviewed, and approved by the authors.

Data Availability Statement

The data that support the findings of this study are available in Gene Expression Omnibus (GEO) at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE52793. These data were derived from the following resources available in the public domain: ‐ The Cancer Genome Atlas (TCGA‐HNSC), https://portal.gdc.cancer.gov/projects/TCGA‐HNSC.

References

  • 1. Barsouk A., Aluru J. S., Rawla P., Saginala K., and Barsouk A., “Epidemiology, Risk Factors, and Prevention of Head and Neck Squamous Cell Carcinoma,” Medical Science 11, no. 2 (2023): 42. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. Sousa H. D., Horita V. N., Perin M. Y., et al., “Platinum‐Based Chemoradiotherapy as Definitive Treatment in Advanced Squamous Cell Carcinoma of Head and Neck in Real‐World Setting,” Hematology, Transfusion and Cell Therapy 47 (2025): 103782. [Google Scholar]
  • 3. Wang H., Zheng Z., Zhang Y., et al., “Locally Advanced Head and Neck Squamous Cell Carcinoma Treatment Efficacy and Safety: A Systematic Review and Network Meta‐Analysis,” Frontiers in Pharmacology 14 (2023): 1269863. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Saini K. S., Somara S., Ko H. C., et al., “Biomarkers in Head and Neck Squamous Cell Carcinoma: Unraveling the Path to Precision Immunotherapy,” Frontiers in Oncology 14 (2024): 1473706. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Eze N., Lo Y. C., and Burtness B., “Biomarker Driven Treatment of Head and Neck Squamous Cell Cancer,” Cancers of the Head & Neck 2, no. 1 (2017): 6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Sun Y., Wang Z., Qiu S., and Wang R., “Therapeutic Strategies of Different HPV Status in Head and Neck Squamous Cell Carcinoma,” International Journal of Biological Sciences 17, no. 4 (2021): 1104–1118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Park J. C., Ye G. H., Merkin R., and Roberts T., “Outcomes for Recurrent or Metastatic Head and Neck Cancer by HPV Status: A Systematic Review and Meta‐Analysis,” Oncologist 30, no. 4 (2025): oyaf043. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Sharafinski M. E., Ferris R. L., Ferrone S., and Grandis J. R., “Epidermal Growth Factor Receptor Targeted Therapy of Squamous Cell Carcinoma of the Head and Neck,” Head & Neck 32, no. 10 (2010): 1412–1421. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Lahin A. E., Matthew S., Joshua S., and Chad B., “Expression of P53 and Prognosis in Patients With Head and Neck Squamous Cell Carcinoma (HNSCC),” International Journal of Cancer and Clinical Research 6 (2019): 122. [Google Scholar]
  • 10. Liu X., Liu Y., Tian X., et al., “A Novel Molecular Classification System for Head and Neck Squamous Cell Carcinoma: Predicting Treatment Response and Metastatic Potential Through Multi‐Omics Analysis,” Discover Oncology 16, no. 1 (2025): 477. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Tavares N. T., Gumauskaitė S., Lobo J., Jerónimo C., and Henrique R., “DNA Methylation Biomarkers for Prediction of Response to Platinum‐Based Chemotherapy: Where Do We Stand?,” Cancers 14, no. 12 (2022): 2918. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. Kit A. H., Nielsen H. M., and Tost J., “DNA Methylation Based Biomarkers: Practical Considerations and Applications,” Biochimie 94, no. 11 (2012): 2314–2337. [DOI] [PubMed] [Google Scholar]
  • 13. Broustas C. G. and Lieberman H. B., “DNA Damage Response Genes and the Development of Cancer Metastasis,” Radiation Research 181, no. 2 (2014): 111–130. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Yu X., Jia L., Tang Q., Zhou Q., Wang G., and Wang S., “Regulation of Cisplatin Resistance in Lung Cancer by Epigenetic Mechanisms,” Clinical Epigenetics 17, no. 1 (2025): 145. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Zeller C. and Brown R., “Therapeutic Modulation of Epigenetic Drivers of Drug Resistance in Ovarian Cancer,” Therapeutic Advances in Medical Oncology 2, no. 5 (2010): 319–329. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Pelosof L., Yerram S., Armstrong T., et al., “GPX3 Promoter Methylation Predicts Platinum Sensitivity in Colorectal Cancer,” Epigenetics 12, no. 7 (2017): 540–550. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Carrington A. M., Fieguth P. W., Qazi H., et al., “A New Concordant Partial AUC and Partial c Statistic for Imbalanced Data in the Evaluation of Machine Learning Algorithms,” BMC Medical Informatics and Decision Making 20, no. 1 (2020): 4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Carrington A. M., Manuel D. G., Fieguth P. W., et al., “Deep ROC Analysis and AUC as Balanced Average Accuracy, for Improved Classifier Selection, Audit and Explanation,” IEEE Transactions on Pattern Analysis and Machine Intelligence 45, no. 1 (2023): 329–341. [DOI] [PubMed] [Google Scholar]
  • 19. Yu L. and Li W., “Abnormal Activation of Notch 1 Signaling Causes Apoptosis Resistance in Cervical Cancer,” International Journal of Clinical and Experimental Pathology 15, no. 1 (2022): 11. [PMC free article] [PubMed] [Google Scholar]
  • 20. Liccardi G., Hartley J. A., and Hochhauser D., “EGFR Nuclear Translocation Modulates DNA Repair Following Cisplatin and Ionizing Radiation Treatment,” Cancer Research 71, no. 3 (2011): 1103–1114. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Zhang K., Zhang B., Bai Y., and Dai L., “E2F1 Promotes Cancer Cell Sensitivity to Cisplatin by Regulating the Cellular DNA Damage Response Through miR‐26b in Esophageal Squamous Cell Carcinoma,” Journal of Cancer 11, no. 2 (2020): 301–310. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22. Fan X., Song J., Fan Y., et al., “CSMD1 Mutation Related to Immunity Can Be Used as a Marker to Evaluate the Clinical Therapeutic Effect and Prognosis of Patients With Esophageal Cancer,” International Journal of General Medicine 14 (2021): 8689–8710. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. Shehata M., Mukherjee A., Deen S., Al‐Attar A., Durrant L. G., and Chan S., “Human Leukocyte Antigen Class I Expression Is an Independent Prognostic Factor in Advanced Ovarian Cancer Resistant to First‐Line Platinum Chemotherapy,” British Journal of Cancer 101, no. 8 (2009): 1321–1328. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24. Helleman J., van Staveren I. L., Dinjens W. N., et al., “Mismatch Repair and Treatment Resistance in Ovarian Cancer,” BMC Cancer 6, no. 1 (2006): 201. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25. Zhou C., Ye M., Ni S., et al., “DNA Methylation Biomarkers for Head and Neck Squamous Cell Carcinoma,” Epigenetics 13, no. 4 (2018): 398–409. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. Nakagawa T., Kurokawa T., Mima M., et al., “DNA Methylation and HPV‐Associated Head and Neck Cancer,” Microorganisms 9, no. 4 (2021): 801. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Zhang Z., Sehgal K., Shirai K., et al., “Methylation Cytometric Pretreatment Blood Immune Profiles With Tumor Mutation Burden as Prognostic Indicators for Survival Outcomes in Head and Neck Cancer Patients on Anti‐PD‐1 Therapy,” npj Precision Oncology 8, no. 1 (2024): 267. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28. Chiu T. J., Chen C. H., Chien C. Y., Li S. H., Tsai H. T., and Chen Y. J., “High ERCC1 Expression Predicts Cisplatin‐Based Chemotherapy Resistance and Poor Outcome in Unresectable Squamous Cell Carcinoma of Head and Neck in a Betel‐Chewing Area,” Journal of Translational Medicine 9 (2011): 31. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Osman A. A., Neskey D. M., Katsonis P., et al., “Evolutionary Action Score of TP53 Coding Variants Is Predictive of Platinum Response in Head and Neck Cancer Patients,” Cancer Research 75, no. 7 (2015): 1205–1215. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Ausoni S., Boscolo‐Rizzo P., Singh B., et al., “Targeting Cellular and Molecular Drivers of Head and Neck Squamous Cell Carcinoma: Current Options and Emerging Perspectives,” Cancer and Metastasis Reviews 35, no. 3 (2016): 413–426. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31. Wang Z. and Anderson K. S., “Therapeutic Targeting of FGFR Signaling in Head and Neck Cancer,” Cancer Journal 28, no. 5 (2022): 354–362. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32. Pan X., Li X., Sun J., et al., “Enhancer Methylation Dynamics Drive Core Transcriptional Regulatory Circuitry in Pan‐Cancer,” Oncogene 41, no. 26 (2022): 3474–3484. [DOI] [PubMed] [Google Scholar]
  • 33. Kreibich E., Kleinendorst R., Barzaghi G., Kaspar S., and Krebs A. R., “Single‐Molecule Footprinting Identifies Context‐Dependent Regulation of Enhancers by DNA Methylation,” Molecular Cell 83, no. 5 (2023): 787–802.e9. [DOI] [PubMed] [Google Scholar]
  • 34. Laugsand E. A., Brenne S. S., and Skorpen F., “DNA Methylation Markers Detected in Blood, Stool, Urine, and Tissue in Colorectal Cancer: A Systematic Review of Paired Samples,” International Journal of Colorectal Disease 36, no. 2 (2021): 239–251. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35. Panzarella V., Buttà M., Buttacavoli F., et al., “Human Papilloma Virus (HPV) Detection in Oral Rinse vs. Oral Sponge: A Preliminary Accuracy Report in Oral Cancer Patients,” Cancers 16, no. 19 (2024): 3256. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Li N., Song K., Chen H., and Dai M., “Advance and Challenge of DNA Methylation as Cancer Biomarkers for Risk Stratification, Screening and Early Detection,” Journal of the National Cancer Center 5, no. 2 (2025): 108–112. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Niekra P. and Adamska P., “The Role of Liquid Biopsy in the Diagnosis of Oral Squamous Cell Carcinoma: A Systematic Review,” International Journal of Molecular Sciences 27, no. 2 (2026): 677. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Poonepalle A., Yang J., Saba N. F., Liu Y., and Teng Y., “Liquid Biopsies in HNSCC: Current Landscape and Emerging Opportunities in the Era of HPV Stratification,” International Journal of Molecular Sciences 27, no. 6 (2026): 2847. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. Rendek T., Pos O., Duranova T., et al., “Current Challenges of Methylation‐Based Liquid Biopsies in Cancer Diagnostics,” Cancers 16, no. 11 (2024): 2001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40. Pharo H., Vedeld H. M., Sjurgard I. V., Pinto R., and Lind G. E., “From Concept to Clinic: A Roadmap for DNA Methylation Biomarkers in Liquid Biopsies,” Oncogene 44 (2025): 1–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41. Hernández‐Lemus E. and Ochoa S., “Methods for Multi‐Omic Data Integration in Cancer Research,” Frontiers in Genetics 15 (2024): 1425456. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42. Cai Z., Poulos R. C., Liu J., and Zhong Q., “Machine Learning for Multi‐Omics Data Integration in Cancer,” iScience 25, no. 2 (2022): 103798. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Table S1: Differentially methylated CpG sites used in constructing the methylation risk score, including logFC weights, p‐values, and genomic annotations.

LIO2-11-e70461-s002.csv (23.2KB, csv)

Table S2: Differential methylation of MRS‐associated CpG sites between HPV‐positive and HPV‐negative HNSCC.

LIO2-11-e70461-s001.csv (1.1KB, csv)

Data Availability Statement

The data that support the findings of this study are available in Gene Expression Omnibus (GEO) at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE52793. These data were derived from the following resources available in the public domain: ‐ The Cancer Genome Atlas (TCGA‐HNSC), https://portal.gdc.cancer.gov/projects/TCGA‐HNSC.

Articles from Laryngoscope Investigative Otolaryngology are provided here courtesy of Wiley

RESOURCES