The Lung Adenocarcinoma Challenge. Lung adenocarcinoma (LUAD) is the most common subtype of lung cancer in China and globally, accounting for the majority of lung cancer cases. Despite advances in chemotherapy, EGFR-targeted therapy, and immune checkpoint inhibitors, the 5-year survival rate remains unsatisfactory, particularly for advanced disease, and drug resistance is a persistent obstacle.
Ferroptosis as an Anti-Cancer Mechanism. Ferroptosis is an iron-dependent, non-apoptotic form of regulated cell death driven by lipid peroxidation. Unlike classical apoptosis, ferroptosis can bypass established drug resistance pathways. Tumor cells typically contain higher iron levels than normal cells, making them particularly susceptible to ferroptosis induction, providing a selective therapeutic window.
Disulfidptosis: A Newly Discovered Cell Death Mode. Disulfidptosis is a recently identified form of cell death triggered when cells with high SLC7A11 (xCT) expression experience glucose deficiency, leading to disulfide bond accumulation and actin cytoskeleton collapse. The regulatory overlap between disulfidptosis and ferroptosis -- particularly through SLC7A11's influence on intracellular iron levels -- makes studying their combined effects on LUAD biologically compelling.
Research Gap and Study Goal. While ferroptosis-related and disulfidptosis-related genes have been separately studied in LUAD, no prior research had systematically linked both pathways in a single prognostic framework. This study aimed to identify disulfidptosis-related ferroptosis genes (DFRGs), construct a prognostic risk model, and validate key biomarkers through in vitro experiments.
Gene Identification Strategy. 597 ferroptosis genes and 36 disulfidptosis genes were retrieved from the FerrDb database. Pearson correlation analysis between both sets (coefficient threshold 0.3) identified 344 disulfidptosis-related ferroptosis genes (DFRGs). These were then crossed with differentially expressed genes from TCGA-LUAD (using absolute log fold change greater than 2 and FDR less than 0.05 as filters), yielding 199 DFRGs for model construction.
TCGA Dataset and Cohort Design. RNA sequencing data, clinical information, somatic mutation data, and copy number variation for LUAD were obtained from TCGA (541 tumor samples, 59 normal samples). 472 patients with complete data were split 6:4 into training and validation sets. Three external GEO datasets (GSE30210 with 226 patients, GSE72094 with 398 patients, GSE13213 with 117 patients) served as independent external validation cohorts.
LASSO-Cox Feature Selection. Univariate Cox regression identified 59 DFRGs significantly associated with LUAD prognosis (P less than 0.05). LASSO regression with cross-validation reduced these to 13 meaningful DFRGs, which were then entered into multivariate Cox regression. This stepwise selection approach minimized overfitting and identified the final six DFRGs with independent prognostic value for model construction.
Risk Score Formula. The six-gene prognostic model was encoded as: Risk score = (0.4169 x AKT1S1) + (-0.164 x CX3CL1) + (0.2039 x DDIT4) + (0.5426 x DECR1) + (0.2968 x KIF20A) + (0.1848 x PCDH7). Five genes (AKT1S1, DDIT4, DECR1, KIF20A, PCDH7) with hazard ratios greater than 1 were risk factors, while CX3CL1 with HR less than 1 was a protective factor. Patients were stratified into high-risk and low-risk groups based on the median risk score.
Survival Stratification. Kaplan-Meier analysis confirmed significant differences in overall survival between high-risk and low-risk groups in the training set, internal validation set, total TCGA set, and all three external GEO validation cohorts (all P less than 0.001). In every cohort, low-risk patients had substantially longer survival than high-risk patients, demonstrating consistent prognostic stratification.
TCGA ROC Performance. Time-dependent ROC analysis in the training set showed AUCs of 0.836 for 1-year, 0.771 for 2-year, and 0.786 for 3-year survival prediction. Validation set AUCs were 0.693, 0.684, and 0.647 respectively. While the validation AUCs were lower, reflecting expected performance degradation in independent data, the model maintained meaningful discriminative ability.
External GEO Validation. The model generalized well across three independent datasets: GSE30210 AUCs were 0.780, 0.806, and 0.673; GSE13213 AUCs were 0.841, 0.722, and 0.733; GSE72094 AUCs were 0.695, 0.654, and 0.608 for 1-, 2-, and 3-year survival. The consistency of risk stratification across datasets from different institutions confirms the model's external validity.
Superiority Over Prior Models. Comparison with nine previously published ferroptosis-based prognostic models from the literature demonstrated that the DFRG model achieved the highest AUC values at 1, 2, and 3 years. This improvement reflects the additional prognostic information captured by incorporating disulfidptosis genes alongside ferroptosis genes, a combination not previously studied in LUAD.
Clinical Nomogram Integration. A nomogram integrating age, gender, stage, and the six-gene risk score was constructed and calibration curves confirmed excellent agreement between model-predicted and actual 1-, 2-, and 3-year survival rates. The combined model's C-index curve exceeded that of any single clinical characteristic alone, demonstrating that the molecular risk score adds substantial prognostic information beyond clinical staging.
Immune Function Differences by Risk Group. Single-sample gene set enrichment analysis (ssGSEA) revealed that immune functions were significantly more enriched in the low-risk group, including B cells, activated dendritic cells, and Type II IFN responses (all P less than 0.001). This indicates that low-risk patients have a more immunologically active tumor microenvironment that may support better antitumor immunity.
TIDE Score and Immunotherapy Prediction. TIDE (Tumor Immune Dysfunction and Exclusion) scores, which reflect the probability of immune escape and immunotherapy resistance, were significantly higher in the high-risk group (P less than 0.01). High-risk patients are more likely to evade immune surveillance and less likely to benefit from immune checkpoint inhibitors, while low-risk patients may respond more favorably to immunotherapy.
TMB and Combined Risk-TMB Stratification. Tumor Mutation Burden (TMB) was significantly higher in the high-risk group (P less than 0.001), and high TMB independently associated with survival benefit (P = 0.006). Combined stratification showed that high-TMB plus low-risk patients had the best prognosis with approximately 50% 7-year survival, while low-TMB plus high-risk patients had the worst outcomes with less than 20% 7-year survival.
DECR1 Selected for In Vitro Validation. Among the six model genes, DECR1 carried the highest importance coefficient (0.542) in the prognostic model. TCGA differential expression analysis confirmed DECR1 expression was significantly higher in LUAD tumor samples than in normal lung tissue, and Kaplan-Meier analysis showed high DECR1 expression correlated with worse OS and PFS in LUAD patients. Protein-level overexpression in LUAD versus normal tissue was confirmed in the HPA database.
DECR1 Expression in Cell Lines. qRT-PCR confirmed that DECR1 expression in lung adenocarcinoma cell lines A549 and H1975 was significantly higher than in normal lung epithelial cells BEAS-2B. Two independent siRNA sequences (si-DECR1-1 and si-DECR1-2) successfully knocked down DECR1 in both cell lines, providing the experimental basis for functional characterization.
DECR1 Knockdown Inhibits Migration and Proliferation. Wound healing assays showed significantly reduced migration speeds in both A549 and H1975 cell lines after DECR1 knockdown compared to negative controls. Transwell assays confirmed fewer migrating cells with DECR1 knockdown. CCK-8 proliferation assays demonstrated that DECR1 knockdown significantly inhibited proliferation in both cell lines over 72 hours.
DECR1 Knockdown Promotes Apoptosis. Flow cytometry Annexin V/FITC apoptosis assays showed that DECR1 knockdown significantly increased apoptosis in A549 cells. Mechanistically, DECR1 is a mitochondrial enzyme involved in beta-oxidation of unsaturated fatty acids and regulation of the saturated-to-unsaturated phospholipid ratio. DECR1 deletion in other cancers has been shown to sensitize cells to ferroptosis, suggesting a convergent mechanism linking DECR1 function to the ferroptosis regulation captured in the model.
Differential Drug Sensitivity by Risk Group. Using the oncoPredict package and GDSC database IC50 predictions, high-risk LUAD patients showed significantly higher sensitivity to eight therapeutic agents: savolitinib, foretinib, cytarabine, a PLK inhibitor (BI-2536), a selective ATR kinase inhibitor (AZD6738), 5-fluorouracil, afatinib, and crizotinib (all P less than 0.01). Lower IC50 in the high-risk group indicates higher drug responsiveness.
Translational Opportunity. The differential drug sensitivity profile provides a direct translational application of the risk model: high-risk patients identified by their molecular profile may derive greater therapeutic benefit from targeted agents including the EGFR/HER2 inhibitor afatinib and the MET/ALK inhibitor crizotinib, despite not having canonical driver mutations in the standard sense.
Biological Roles of the Model Genes. Each of the six model genes has established cancer biology: KIF20A drives mitotic chromosome transport and is linked to chemotherapy resistance; PCDH7 promotes lung cancer development and cisplatin resistance; CX3CL1 (the protective factor) increases antitumor immune responses by attracting effector immune cells; DDIT4 regulates autophagy; AKT1S1 is upregulated in hepatocellular carcinoma and promotes tumor growth.
Clinical Decision Support Potential. The combined nomogram and drug sensitivity analysis enables two distinct clinical applications: the nomogram provides individualized survival probability estimates (1-, 2-, and 3-year) from routine clinical data, while the drug sensitivity analysis guides therapeutic agent selection based on a patient's molecular risk profile. Together, they offer a framework for personalized LUAD management.
Model Contributions. This study constructed the first prognostic model for LUAD combining disulfidptosis and ferroptosis genes. The six-gene DFRG risk score demonstrated robust survival stratification across TCGA training, internal validation, and three independent external GEO cohorts. The model outperformed nine previously published ferroptosis-only models, and in vitro functional validation of DECR1 confirmed its oncogenic role in LUAD biology.
DECR1 as a Novel LUAD Biomarker. This study is the first to link DECR1 specifically to LUAD. DECR1 overexpression in tumor tissues and cell lines, combined with functional evidence that knockdown suppresses proliferation, migration, and promotes apoptosis, establishes DECR1 as a candidate therapeutic target. Its mechanistic connection to ferroptosis sensitivity may make it exploitable for combination strategies with ferroptosis-inducing agents.
Study Limitations. The core analysis relies on TCGA bulk transcriptome data, which is subject to cohort selection bias, heterogeneous clinical annotation, and platform artifacts. In vitro validation was performed only for DECR1; the biological functions of the other five model genes were not experimentally confirmed. Most of the analysis was secondary bioinformatics from public databases, limiting mechanistic depth.
Future Directions. Further work is needed to experimentally validate the biological roles of AKT1S1, DDIT4, KIF20A, PCDH7, and CX3CL1 in LUAD. In vivo tumor model studies and prospective clinical cohort validation are required before the model can be translated to clinical practice. Investigation of DECR1 as a target for ferroptosis-sensitizing combination therapy represents a promising next research direction.