Development and validation of an 11-CpG methylation signature-based nomogram for predicting prognosis in early-onset breast cancer

Introduction

Breast cancer (BC) remains a leading cause of cancer-related mortality among women worldwide. Recent epidemiological evidence indicates a concerning rise in BC incidence among younger women (1-3), underscoring the growing clinical relevance of early-onset breast cancer (EOBC), commonly defined as disease diagnosed before 50 years of age (4). Compared with late-onset BC, EOBC more frequently presents with advanced-stage disease and higher prevalence of biologically aggressive features, including higher tumor grade and triple-negative breast cancer (TNBC) (2,5). These characteristics are associated with unfavorable clinical outcomes and diminished responsiveness to standard systemic adjuvant therapies (4,6,7). Collectively, this bulk of evidence highlights the necessity of further elucidating the pathogenic mechanisms and associated prognostic risk factors underlying EOBC, thereby providing a theoretical foundation for the development of novel therapeutic strategies.

Cancer development is driven by cumulative genetic and epigenetic alterations (8-10). Among epigenetic mechanisms, DNA methylation plays a key role in modulating gene expression without changing the DNA sequence (11). It has emerged as a source of biomarkers for cancer early detection, molecular classification, and prognosis (12,13). Nevertheless, many methylation markers remain unvalidated, and numerous prognostically relevant epigenetic alterations await discovery, including those in BC (14). Early-onset cancer exhibits distinct methylation landscapes compared with late-onset disease, revealing a dynamic interplay with genetic alterations, associations with distinct clinical outcomes, and implications for novel treatment strategies (15,16). In the context of EOBC, there is substantial evidence of unique methylation profiles, including constitutional BRCA1 promoter methylation, which can promote tumorigenesis independently of germline mutations and is associated with aggressive phenotypes and adverse prognosis (17-20). These observations suggest that EOBC-specific methylation signatures may capture unique biological features with prognostic relevance.

Despite increasing interest in methylation-based prognostic models for BC, existing signatures are largely derived from unselected populations or restricted to specific subtypes such as TNBC. EOBC, as a clinically and biologically distinct entity, remains underrepresented (21,22). To address this gap, we identified differentially methylated probes (DMPs) between EOBC and normal breast tissues, performed univariable Cox regression to select prognosis-associated CpG sites (CpGs), and applied least absolute shrinkage and selection operator (LASSO) regression to construct a multi-CpG-based methylation signature for EOBC (MSEO) prognosis. We then built a nomogram incorporating this signature to predict overall survival (OS) in EOBC patients. Upon validation in independent cohorts, the model demonstrated high predictive accuracy and well-calibrated performance, supporting its utility for risk stratification and personalized clinical management in this patient population. The integration of these epigenetic analyses into clinical practice holds significant promise for advancing patient treatment. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-aw-2262/rc).

Methods

Data acquisition and cohort selection

Publicly available DNA methylation data and clinical parameters were obtained from The Cancer Genome Atlas (TCGA) and the Gene Expression Omnibus (GEO). We retrieved raw IDAT files from Illumina Infinium HumanMethylation450 arrays for EOBC samples and paired normal breast tissues, along with complete clinical annotations. Datasets GSE72245, GSE72251, and GSE75067—all generated using the HumanMethylation450 BeadChip (GPL13534)—were included (23,24). Beta values for GSE75067 were acquired from pre-normalized data published by Holm et al.

Meanwhile, clinicopathological variables—including age, tumor size, nodal status, and hormone receptor expression—were extracted. Tumor size was categorized as T1 (≤2 cm) or T2–T3 (>2 cm). hormone receptor positivity was defined as estrogen receptor- and/or progesterone receptor-positive status. For sample selection, the inclusion criteria were: (I) age ≤50 years; (II) primary invasive BC; (III) no neoadjuvant therapy; (IV) available OS data; (V) a minimum follow-up of 1 month. To enhance statistical power, samples from TCGA, GSE72245, and GSE72251 were merged as discovery set I, while GSE75067 served as an independent validation cohort. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Data preprocessing and model development

First, raw IDAT files were processed using the ChAMP pipeline (25). Probes with detection P values >0.01, those located on sex chromosomes, or those containing SNPs were excluded. Subsequently, beta-mixture quantile normalization was applied to correct for probe-type bias. Batch effects were mitigated via champ.runCombat function. Beta values (β), where β = M/(U + M + 100) (M = methylated signal intensity; U = unmethylated signal intensity), represented methylation levels. In GSE75067, k-nearest neighbors imputation was used for probes with <10% missingness. Next, DMPs between EOBC and normal tissues were identified using limma package with empirical Bayes moderated t-tests (26). Significance thresholds were false discovery rate (FDR)-adjusted P<0.05 and |Δβ| ≥0.2. Based on these DMPs, univariable Cox regression (P≤0.001) identified OS-associated CpGs in the discovery set I.

For prognostic signature construction, the discovery set I was randomly split into training set I (80%) and test set I (20%) using a range of random seeds. LASSO-Cox regression with 10-fold cross-validation was applied to OS-associated CpGs to determine the optimal λ, and those with nonzero coefficients were retained to form a methylation-based prognostic signature. Risk scores were calculated as:

Riskscore=∑i=1nCoefficient(CpGi)×Methylationlevel(CpGi)[1]

To enable risk stratification, an outcome-oriented approach using maximally selected rank statistics identified the optimal survival-related cutoff via the surv_cutpoint function (survminer R package), subsequently stratifying patients into high- and low-risk groups. Finally, survival distributions between these groups were compared using the Kaplan-Meier (KM) method, with the statistical significance evaluated by the log-rank test.

To identify clinicopathological risk factors for OS, the discovery set II was formed from samples in the discovery set I with complete clinicopathological data for univariable and multivariable Cox regression analyses. The discovery set II was subsequently subjected to a repeated random sampling procedure, allocating 80% of its samples to training set II and the remaining 20% to test set II. Significant variables from univariable Cox regression and the methylation-based risk score were included in a multivariable Cox model using training set II. This model was used to create a prognostic nomogram with the “rms” package and subsequently underwent internal and external validation. Time-dependent receiver operating characteristic (ROC) analysis and calibration curves were employed to assess predictive accuracy and the agreement between predicted and observed outcomes.

Statistical analysis

All analyses were performed using R version 4.2.3. Group comparisons were conducted using Mann-Whitney U tests for continuous variables or chi-square tests for categorical variables. Univariable and multivariable Cox proportional hazards models were used to identify independent prognostic factors. The predictive capacities of MSEO and the nomogram for OS were compared using the timeROC package (27). Statistical significance was set at two-tailed P<0.05.

Results

Patient characteristics

The study design is outlined in Figure 1. A total of 318 EOBC samples with complete survival data were included in the discovery set I, consisting of 240 samples from the TCGA cohort, 38 from GSE72245, and 40 from GSE72251. From discovery set I, the patients were randomly allocated to training set I (n=254) and test set I (n=64). Post-randomization, no significant baseline differences were observed (P>0.05), ensuring a balanced distribution for internal validation. The detailed baseline characteristics of the patients in training set I and test set I are summarized in Table S1. Additionally, 106 samples from GSE75067 were used as the validation set I.

Figure 1 Flowchart indicating study design. CpGs, CpG sites; DMA, differential methylation analysis; DMPs, differentially methylated probes; EOBC, early-onset breast cancer; LASSO, least absolute shrinkage and selection operator; MSEO, methylation signature for EOBC; OS, overall survival; TCGA, the Cancer Genome Atlas.

Selection of OS-related CpGs

As illustrated in the workflow in Figure 1, differential methylation analysis of 29 EOBC tissue pairs identified 19,343 DMPs. Univariable Cox analysis further narrowed these down to 127 DMPs significantly associated with OS. The complete list of these OS-associated CpGs is provided in Table S2. These candidate DMPs were further subjected to LASSO-Cox regression analysis in the training set I. The final prognostic signature consisted of 11 CpGs identified by the LASSO model at the optimal penalty parameter (λ =0.042) determined by the one-standard-error rule (Figure 2). The comprehensive genomic annotation of these 11 selected CpGs is detailed in Table S3. These results highlight potential DNA methylation markers for prognosis and risk stratification in EOBC.

Figure 2 Selection of CpGs for OS using LASSO Cox regression model. (A) LASSO regression with 10-fold cross-validation was employed to select the optimal lambda. The 1-SE criterion was applied to choose the most parsimonious model within one SE of the minimum. Consequently, a lambda value of 0.042 [log(lambda) =−3.161] was selected. (B) LASSO coefficient profile plot displaying 127 variables against the log(lambda) sequence. A purple vertical line was drawn at the value chosen using 10-fold cross-validation, which resulted in 11 non-zero coefficients. CpGs, CpG sites; LASSO, least absolute shrinkage and selection operator; OS, overall survival; SE, standard error.

Building a predictive methylation signature

To assess the predictive efficacy of MSEO based on individualized methylation levels of 11 CpGs for OS, a penalized Cox regression model was employed to generate weighting coefficients. This resulted in the following risk score formula: risk score = 2.1569 × methylation level of cg27106909 − 0.6385 × methylation level of cg00910015 − 0.9826 × methylation level of cg01105356 + 0.4607 × methylation level of cg03527802 − 0.5931 × methylation level of cg05084668 − 0.1980 × methylation level of cg06611426 − 1.3804 × methylation level of cg14189141 − 0.6285 × methylation level of cg14565781 − 0.6549 × methylation level of cg19614321 − 0.5838 × methylation level of cg23824801 − 0.6082 × methylation level of cg25224048.

To enhance the prognostic evaluation, the optimal cutoff value for the risk score was determined to be −1.971, categorizing patients into high-risk (risk score >−1.971) and low-risk groups based on this threshold. Figure 3A-3C illustrates the distribution of risk scores, risk stratification, survival outcomes, and methylation levels within the prognostic model for the training set I, test set I, and validation set I.

Figure 3 The methylation patterns of the 11 selected CpGs according to the distribution of risk scores in the training set I (A), test set I (B), and external validation set I (C). Upper and middle panels: risk score distribution of the 11-CpG-based signature and patient survival status. Lower panel: heatmap showing methylation levels of the 11 CpGs in the samples, with the color bar on the right representing the β-values in the range of 0–1. CpGs, CpG sites.

Subsequently, KM survival analysis was employed to compare survival outcomes between the high-risk and low-risk cohorts. The analysis revealed that patients classified within the high-risk group exhibited significantly shorter OS than those in the low-risk group, as evidenced by the log-rank test (P<0.001; Figure 4A). This pattern was consistently observed in both the internal test set I and the external validation set I, with P values of <0.001 and 0.01, respectively (Figure 4B,4C). Additionally, time-dependent ROC analysis was utilized to assess the sensitivity and specificity of the methylation signature in predicting OS. In training set I, the area under the curve (AUC) values for predicting 3-, 5-, and 8-year OS were 0.952, 0.874, and 0.903, respectively (Figure 4D). The AUC values in test set I were 0.833, 0.911, and 0.876 for 3-, 5-, and 8-year OS, respectively (Figure 4E). In the validation set I, the corresponding AUC values were 0.753, 0.760, and 0.671 (Figure 4F). Overall, all analyses highlight the robust prognostic performance of MSEO.

Figure 4 KM curves for OS prediction based on MSEO in the training set I (A), testing set I (B) and validation set I (C). The performance of time-dependent ROC curves of MSEO in predicting 3-, 5-, and 8-year OS in the training set I (D), test set I (E), and external validation set I (F). AUC, area under the curve; EOBC, early-onset breast cancer; KM, Kaplan-Meier; MSEO, methylation signature for EOBC; OS, overall survival; ROC, receiver operating characteristic.

Association between clinicopathological risk factors and prognosis in EOBC

The clinical features of the patients in discovery set II (n=299) are itemized in Table S4. In this set, the univariable Cox proportional hazards regression analysis for OS identified a hazard ratio (HR) of 4.385 for each one-unit increase in the risk score [95% confidence interval (CI): 2.914–6.600; P<0.001], signifying a markedly elevated risk of mortality. Additionally, factors such as younger age (≤35 years compared to 35–50 years: HR =2.589; 95% CI: 1.261–5.319; P=0.01), larger tumor size (>2 cm compared to ≤2 cm: HR =2.290; 95% CI: 1.172–4.473; P=0.01), nodal status (positive versus negative: HR =1.758; 95% CI: 1.029–3.002; P=0.03), and hormone receptor-negative status (negative versus positive: HR =2.792; 95% CI: 1.426–5.464; P=0.003) were significantly correlated with an increased incidence of clinical events. These significant variables were included in a multivariable Cox model, where risk score remained a robust independent predictor of OS (HR =4.447; 95% CI: 2.576–7.675; P<0.001). Detailed results are illustrated in Figure 5.

Figure 5 Forest plots of univariable Cox regression analysis and multivariable Cox regression analysis. HR >1 indicates that the variable is positively associated with event probability; the line segment parallel to the X-axis represents the confidence interval (95% CI). CI, confidence interval; EOBC, early-onset breast cancer; HR, hazard ratio; MSEO, methylation signature for EOBC.

Construction and validation of a predictive nomogram

To develop a clinically translatable and individualized predictive tool for OS, the discovery set II was randomly divided into the training set II (n=239) and the test set II (n=60). Afterward, both sets demonstrated well-balanced baseline characteristics, as elaborated in Table S5.

A multivariate Cox proportional hazards regression model was developed using the training set II, incorporating clinicopathological risk factors such as age, tumor size, nodal status, hormone receptor status, and MSEO. This model was used to create a methylation-clinicopathological nomogram for predicting 3-, 5-, and 8-year OS probabilities (Figure 6A). The application of the nomogram was described as follows. In the nomogram, individual scores for each variable are obtained by drawing a vertical line to the points axis. The sum of these scores yields a total point value, which is then used to estimate the probability of OS at 3, 5, and 8 years. Notably, within the OS nomogram, the MSEO risk score was the most significant contributor to predicting survival outcomes. For example, the predictive indicators of a representative patient with EOBC were as follows: 32 years old (assigned 1 point), a tumor size of 3 cm (8 points), presence of regional lymph node metastasis (16 points), negativity for both estrogen receptor and progesterone receptor (9 points), and a methylation-based risk score of −1.0 (80 points). The sum of these individual variable points yielded a total nomogram score of 114. Based on this score, the nomogram predicts a 3-year OS probability of approximately 43%, with conservatively estimated 5- and 8-year OS probabilities of less than 10%, collectively underscoring a markedly elevated risk of mortality.

Figure 6 The prognostic performance of MSEO-based nomogram in EOBC patients. (A) The prognostic nomogram obtained from the multivariate Cox model to predict 3-, 5-, and 8-year OS for EOBC patients. The time-dependent ROC curves of nomogram in the training set II (B), test set II (C), and external validation set II (D). The performance comparison between nomogram and MSEO in predicting 3-, 5-, and 8-year OS (E-G). AUC, area under the curve; EOBC, early-onset breast cancer; MSEO, methylation signature for EOBC; OS, overall survival; ROC, receiver operating characteristic.

Time-dependent ROC analyses were conducted using the total risk score for each patient, referred to as prognostic index. In training set II, the AUCs for predicting 3-, 5-, and 8-year OS were 0.912, 0.882, and 0.873, respectively (Figure 6B). Test set II showed AUC values of 0.971, 0.958, and 0.835 for the same time intervals (Figure 6C). In the external validation set II, the AUC values for 3-, 5-, and 8-year OS were 0.856, 0.868, and 0.756, respectively (Figure 6D).

To identify the optimal prognostic nomogram, this study compared the predictive accuracy of the nomogram with that of the MSEO across all available cases (n=384) (Figure 6E-6G). Time-dependent ROC analysis at 3- and 5-year intervals demonstrated that the nomogram achieved significantly higher AUC values compared to the MSEO alone (both adjusted P<0.05), underscoring its superior performance in predicting OS in EOBC patients. Furthermore, calibration curve analysis demonstrated strong agreement between the nomogram-predicted and observed probabilities of 5-year OS across all three datasets (Figure 7), supporting good calibration performance of the model.

Figure 7 Calibration curves for nomogram-predicted probability of 5-year OS in the training set II (A), test set II (B), and external validation set II (C). OS, overall survival.

Discussion

EOBC represents a distinct clinical entity that is becoming increasingly prevalent among young women. It is often linked to poor tumor differentiation, hormone receptor negativity, and pathogenic variants in genes like BRCA1 and BRCA2, resulting in a worse prognosis and higher treatment-related morbidity (28-30). Despite these well-documented features, prognostic stratification strategies specifically tailored to EOBC remain limited (30), underscoring the need for biomarkers that capture its unique biological context.

Epigenetic dysregulation, particularly aberrant DNA methylation, represents an early and stable molecular event in breast tumorigenesis and has emerged as a promising source of prognostic biomarkers (31,32). Accordingly, we systematically investigated methylation-based prognostic markers in EOBC. Using an integrated bioinformatics framework, we identified DMPs associated with OS, established a MSEO, and developed a nomogram incorporating 11 CpGs together with clinicopathological variables. Collectively, this model provides a novel approach for individualized outcome prediction and underscores the translational potential of DNA methylation-based risk stratification in EOBC.

Beyond their prognostic value, the CpGs comprising the MSEO may also offer mechanistic insights into EOBC biology. For instance, cg03527802 and cg05084668, linked to the MAZ and ALG1L genes, respectively, show lower methylation in EOBC samples compared to normal tissues. MAZ promotes tumor growth in BC by upregulating glycolytic genes such as CUEDC1 and directly activating the PI3K/AKT signaling pathway, thereby driving metabolic reprogramming and cell proliferation linked to poor clinical outcomes (33,34). Additionally, MAZ modulates antitumor immunity by recruiting STAT1 to chromatin and regulating interferon-γ-stimulated genes, facilitating immune evasion and exacerbating tumor progression (35). ALG1L could serve as an independent prognostic biomarker in lung squamous cell carcinoma (36).

Beyond these, the other six sites with higher methylation in EOBC are associated with the CIITA, GRASP, BTBD19, RASSF2, CBX5, and YPEL3 genes. CIITA expression correlates with favorable prognosis in BC by enhancing tumor-specific MHC-II presentation and promoting cytotoxic immune infiltration, thereby improving the response to immunotherapy and enabling anthracycline exemption in patients with high MHC-II expression (37,38). Conversely, MHC-II⁺ tumor cells in lymph nodes evade immunity via Treg expansion and impaired T cell activation, promoting metastasis—a mechanism reinforced by CIITA overexpression and rescued by MHC-II knockout (39). GRASP hypermethylation serves as a specific diagnostic and prognostic biomarker in prostate cancer (40). As for RASSF2, its hypermethylation is a frequent epigenetic event in BC and serves as a potential early biomarker for disease progression (41). Notably, detection of RASSF2 methylation in sentinel lymph nodes correlates with macrometastatic spread, highlighting its clinical relevance for nodal assessment and therapeutic decision-making (42). In TNBC, CBX5 serves as a downstream effector of the oncogenic lncRNA SNHG11, promoting tumor cell proliferation and migration via a miR-2355-5p–dependent mechanism (43). Likewise, YPEL3 acts as a tumor suppressor in BC by promoting cellular senescence and apoptosis, and its expression is directly suppressed by the oncogene YAP1, thereby linking the Hippo pathway to senescence-driven tumor suppression (44). BTBD19 promotes colorectal cancer progression by orchestrating extracellular matrix remodeling, focal adhesion assembly, and epithelial-mesenchymal transition through pathway dysregulation, while simultaneously driving immune evasion via M2 macrophage recruitment and checkpoint molecule upregulation (45). Collectively, functional and mechanistic annotation of these genes underscores their involvement in epigenetic regulation, immune evasion, and tumor progression, pointing to potential biological underpinnings of EOBC aggressiveness.

Compared to previously reported DNA methylation-based prognostic models in BC, such as the 15-CpG signature for TNBC and the 14-CpG model for general BC, our 11-CpG signature offers specific advancements tailored for EOBC—a clinically distinct and understudied population. The parsimonious nature of the signature enhances its potential cost-effectiveness and translational feasibility. Methodologically, the model was rigorously validated in an independent external cohort, where it demonstrated sustained predictive accuracy for OS, including focused assessment of long-term (8-year) outcomes, a prognostic horizon seldom addressed in prior methylation-based studies. Importantly, integrating this signature with established clinicopathological risk factors led to a statistically significant improvement in predicting mid-term survival outcomes—an enhancement not consistently evidenced in existing CpG-based prognostic models, which often report increases in AUC or C-index values without formal statistical validation.

A decline in AUC values, particularly for the 8-year OS prediction in the external validation cohort (0.756 vs. 0.873 in the training set), was observed and is consistent with patterns reported in prior prognostic modeling studies, including the methylation-based model developed by Peng et al. (22). This attenuation likely reflects inherent cohort heterogeneity—including differences in patient demographics, treatment strategies, and follow-up schedules between the independent validation set and the original training cohorts—which can lead to a modest attenuation in model performance. Furthermore, predicting long-term outcomes (e.g., 8-year OS) is inherently more challenging due to the increasing influence of competing risks and unmeasured variables over time, which may explain the more pronounced decrease at the later time point. Notably, the AUCs for 3- and 5-year OS in the external set remained high (0.856 and 0.868, respectively), underscoring the model’s robust and reliable predictive capability within a clinically relevant mid-term horizon. Together, the resulting nomogram provides a refined, statistically robust tool to assist clinicians in formulating individualized management strategies for patients with EOBC.

Nevertheless, several limitations should be acknowledged. The lack of experimental validation limits definitive confirmation of the functional relevance of the identified CpGs; however, the consistent signals observed across multiple independent cohorts provide indirect support for their robustness. Certain cohorts included relatively small sample sizes, which may restrict generalizability. Notably, the initial differential methylation analysis involved a modest sample of 29 tumor-normal pairs, potentially limiting statistical power and increasing the risk of false negatives in epigenome-wide association studies. To mitigate this, stringent statistical thresholds (FDR <0.05 and |Δβ| ≥0.2) were applied, and the prognostic signature derived from these CpGs was further validated in independent cohorts. The consistency of the methylation signature across multiple validation stages supports its biological relevance, though future studies with larger paired cohorts will help confirm these findings. Moreover, the exclusion of patients with advanced T4 stage or distant metastasis restricts the model’s current applicability, although focusing on EOBC allows for the development of a more precise and clinically actionable prognostic tool. Addressing these limitations through experimental assays and broader clinical testing will be critical for translating this prognostic model into practice.

Conclusions

Our study identifies and validates a DNA methylation–based prognostic signature that, when combined with clinical parameters, offers a powerful approach for risk stratification and personalized management of EOBC. The proposed nomogram demonstrates superior predictive accuracy and clinical utility, representing a significant advancement in the management of EOBC. Future research should prioritize the validation of these findings in larger, independent cohorts and investigate the biological implications of the identified methylation markers, thereby deepening our understanding of EOBC pathogenesis and therapeutic response.

Acknowledgments

We are grateful to all researchers who contributed to the shared GEO (GEO accession: GSE72245, GSE72251 and GSE75067) and TCGA databases.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-aw-2262/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-aw-2262/prf

Funding: This project was supported by the Chongqing Municipal Science and Health Joint Medical Research Project (grant No. 2025ZYYB038) and the Science and Technology Research Program of Chongqing Municipal Education Commission (grant No. KJQN202501152).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-aw-2262/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Seely JM, Ellison LF, Billette JM, et al. Incidence of Breast Cancer in Younger Women: A Canadian Trend Analysis. Can Assoc Radiol J 2024;75:847-54. [Crossref] [PubMed]
2. Ilic L, Haidinger G, Simon J, et al. Trends in female breast cancer incidence, mortality, and survival in Austria, with focus on age, stage, and birth cohorts (1983-2017). Sci Rep 2022;12:7048. [Crossref] [PubMed]
3. Yin M, Wang F, Zhang Y, et al. Analysis on Incidence and Mortality Trends and Age-Period-Cohort of Breast Cancer in Chinese Women from 1990 to 2019. Int J Environ Res Public Health 2023;20:826. [Crossref] [PubMed]
4. Gremke N, Wagner U, Kalder M, et al. Changes in the incidence of early-onset breast cancer in Germany between 2010 and 2022. Breast Cancer Res Treat 2023;202:167-72. [Crossref] [PubMed]
5. Akakpo PK, Imbeah EG, Edusei L, et al. Clinicopathologic characteristics of early-onset breast cancer: a comparative analysis of cases from across Ghana. BMC Womens Health 2023;23:5. [Crossref] [PubMed]
6. Francies FZ, Hull R, Khanyile R, et al. Breast cancer in low-middle income countries: abnormality in splicing and lack of targeted treatment options. Am J Cancer Res 2020;10:1568-91. [PubMed]
7. DeSantis CE, Ma J, Jemal A. Trends in stage at diagnosis for young breast cancer patients in the United States. Breast Cancer Res Treat 2019;173:743-7. [Crossref] [PubMed]
8. Zhou L, Yu CW. Epigenetic modulations in triple-negative breast cancer: Therapeutic implications for tumor microenvironment. Pharmacol Res 2024;204:107205. [Crossref] [PubMed]
9. Orsolic I, Carrier A, Esteller M. Genetic and epigenetic defects of the RNA modification machinery in cancer. Trends Genet 2023;39:74-88. [Crossref] [PubMed]
10. Karami Fath M, Azargoonjahromi A, Kiani A, et al. The role of epigenetic modifications in drug resistance and treatment of breast cancer. Cell Mol Biol Lett 2022;27:52. [Crossref] [PubMed]
11. Hao X, Luo H, Krawczyk M, et al. DNA methylation markers for diagnosis and prognosis of common cancers. Proc Natl Acad Sci U S A 2017;114:7414-9. [Crossref] [PubMed]
12. Bai Y, Xu J, Li D, et al. HepaClear, a blood-based panel combining novel methylated CpG sites and protein markers, for the detection of early-stage hepatocellular carcinoma. Clin Epigenetics 2023;15:99. [Crossref] [PubMed]
13. Hu J, Zhao FY, Huang B, et al. An Eight-CpG-based Methylation Classifier for Preoperative Discriminating Early and Advanced-Late Stage of Colorectal Cancer. Front Genet 2020;11:614160. [Crossref] [PubMed]
14. Hinoue T, Weisenberger DJ, Lange CP, et al. Genome-scale analysis of aberrant DNA methylation in colorectal cancer. Genome Res 2012;22:271-82. [Crossref] [PubMed]
15. Makabe T, Arai E, Hirano T, et al. Genome-wide DNA methylation profile of early-onset endometrial cancer: its correlation with genetic aberrations and comparison with late-onset endometrial cancer. Carcinogenesis 2019;40:611-23. [Crossref] [PubMed]
16. Ugai T, Haruki K, Harrison TA, et al. Molecular Characteristics of Early-Onset Colorectal Cancer According to Detailed Anatomical Locations: Comparison With Later-Onset Cases. Am J Gastroenterol 2023;118:712-26. [Crossref] [PubMed]
17. Muhammad N, Azeem A, Bakar MA, et al. Contribution of constitutional BRCA1 promoter methylation to early-onset and familial breast cancer patients from Pakistan. Breast Cancer Res Treat 2023;202:377-87. [Crossref] [PubMed]
18. Wong EM, Southey MC, Fox SB, et al. Constitutional methylation of the BRCA1 promoter is specifically associated with BRCA1 mutation-associated pathology in early-onset breast cancer. Cancer Prev Res (Phila) 2011;4:23-33. [Crossref] [PubMed]
19. Brianese RC, Nakamura KDM, Almeida FGDSR, et al. BRCA1 deficiency is a recurrent event in early-onset triple-negative breast cancer: a comprehensive analysis of germline mutations and somatic promoter methylation. Breast Cancer Res Treat 2018;167:803-14. [Crossref] [PubMed]
20. Scott CM, Wong EM, Joo JE, et al. Genome-wide DNA methylation assessment of 'BRCA1-like' early-onset breast cancer: Data from the Australian Breast Cancer Family Registry. Exp Mol Pathol 2018;105:404-10. [Crossref] [PubMed]
21. Tian BX, Yu ZX, Qiu X, et al. Development and validation of a 14-CpG DNA methylation signature and drug targets for prognostic prediction in breast cancer. Front Med (Lausanne) 2025;12:1548726. [Crossref] [PubMed]
22. Peng Y, Shui L, Xie J, et al. Development and validation of a novel 15-CpG-based signature for predicting prognosis in triple-negative breast cancer. J Cell Mol Med 2020;24:9378-87. [Crossref] [PubMed]
23. Jeschke J, Bizet M, Desmedt C, et al. DNA methylation-based immune response signature improves patient diagnosis in multiple cancers. J Clin Invest 2017;127:3090-102. [Crossref] [PubMed]
24. Holm K, Staaf J, Lauss M, et al. An integrated genomics analysis of epigenetic subtypes in human breast tumors links DNA methylation patterns to chromatin states in normal mammary cells. Breast Cancer Res 2016;18:27. [Crossref] [PubMed]
25. Morris TJ, Butcher LM, Feber A, et al. ChAMP: 450k Chip Analysis Methylation Pipeline. Bioinformatics 2014;30:428-30. [Crossref] [PubMed]
26. Yang X, Xiao X, Zhang L, et al. An integrative analysis of DNA methylation and transcriptome showed the dysfunction of MAPK pathway was involved in the damage of human chondrocyte induced by T-2 toxin. BMC Mol Cell Biol 2022;23:4. [Crossref] [PubMed]
27. Blanche P, Dartigues JF, Jacqmin-Gadda H. Estimating and comparing time-dependent areas under receiver operating characteristic curves for censored event times with competing risks. Stat Med 2013;32:5381-97. [Crossref] [PubMed]
28. Doğan A, İlhan N, Akdağ G, et al. Evaluating the Effectiveness of Cyclin-Dependent Kinase 4/6 Inhibitors in Early- and Very Early-Onset Metastatic Breast Cancer: A Multicenter Study. Medicina (Kaunas) 2025;61:154. [Crossref] [PubMed]
29. Stibbards-Lyle M, Malinovska J, Badawy S, et al. Status of breast cancer detection in young women and potential of liquid biopsy. Front Oncol 2024;14:1398196. [Crossref] [PubMed]
30. Zhu JW, Charkhchi P, Adekunte S, et al. What Is Known about Breast Cancer in Young Women? Cancers (Basel) 2023;15:1917. [Crossref] [PubMed]
31. Han X, Tang J, Chen T, et al. Restoration of GATA4 expression impedes breast cancer progression by transcriptional repression of ReLA and inhibition of NF-κB signaling. J Cell Biochem 2019;120:917-27. [Crossref] [PubMed]
32. Alotaibi G. A systematic review of progress toward unlocking the power of epigenetics in breast cancer: latest updates and perspectives. Front Pharmacol 2025;16:1628165. [Crossref] [PubMed]
33. Lu Z, Lei M, Chen J, et al. CUEDC1 promotes glycolytic metabolism reprogramming through the CUEDC1/CACNG4/PI3K axis to promote ER-positive breast cancer growth. Cell Mol Biol Lett 2025;30:143. [Crossref] [PubMed]
34. Hyun H, Sun B, Yazdimamaghani M, et al. Tumor-specific surface marker-independent targeting of tumors through nanotechnology and bioorthogonal glycochemistry. J Clin Invest 2025;135:e184964. [Crossref] [PubMed]
35. Li Y, Lin Y, Tang Y, et al. MAZ-mediated up-regulation of BCKDK reprograms glucose metabolism and promotes growth by regulating glucose-6-phosphate dehydrogenase stability in triple-negative breast cancer. Cell Death Dis 2024;15:516. [Crossref] [PubMed]
36. Han P, Liu Q, Xiang J. Monitoring methylation-driven genes as prognostic biomarkers in patients with lung squamous cell cancer. Oncol Lett 2020;19:707-16. [PubMed]
37. Li W, Yu X, Xia Y, et al. Combined Bioinformatics Analyses and Immunohistochemical Validation Reveal the Prognostic Relevance and Immune-Related Role of CIITA in Breast Cancer. J Cancer 2025;16:3513-24. [Crossref] [PubMed]
38. Wang Z, Wang Y, Gao Z, et al. Tumor-specific MHC-II guides anthracycline exemption and immunotherapy benefit in breast cancer. Biomark Res 2025;13:83. [Crossref] [PubMed]
39. Lei PJ, Pereira ER, Andersson P, et al. Cancer cell plasticity and MHC-II-mediated immune tolerance promote breast cancer metastasis to lymph nodes. J Exp Med 2023;220:e20221847. [Crossref] [PubMed]
40. Bjerre MT, Strand SH, Nørgaard M, et al. Aberrant DOCK2, GRASP, HIF3A and PKFP Hypermethylation has Potential as a Prognostic Biomarker for Prostate Cancer. Int J Mol Sci 2019;20:1173. [Crossref] [PubMed]
41. Cooper WN, Dickinson RE, Dallol A, et al. Epigenetic regulation of the ras effector/tumour suppressor RASSF2 in breast and lung cancer. Oncogene 2008;27:1805-11. [Crossref] [PubMed]
42. Martín-Sánchez E, Pernaut-Leza E, Mendaza S, et al. Gene promoter hypermethylation is found in sentinel lymph nodes of breast cancer patients, in samples identified as positive by one-step nucleic acid amplification of cytokeratin 19 mRNA. Virchows Arch 2016;469:51-9. [Crossref] [PubMed]
43. Yu L, Zhang W, Wang P, et al. LncRNA SNHG11 aggravates cell proliferation and migration in triple-negative breast cancer via sponging miR-2355-5p and targeting CBX5. Exp Ther Med 2021;22:892. [Crossref] [PubMed]
44. Kwon Y, Lee H, Park H, et al. YPEL3 expression induces cellular senescence via the Hippo signaling pathway in human breast cancer cells. Toxicol Res 2023;39:711-9. [Crossref] [PubMed]
45. Yang C, Geng X, Zhao Z. BTBD19 promotes colorectal cancer progression and correlates with adverse clinical outcomes. Front Oncol 2025;15:1685601. [Crossref] [PubMed]