A new oxidative stress-related lncRNA signature predicts the prognosis of colorectal cancer patients

Abstract

Colorectal cancer (CRC) is a common malignant tumor of the digestive system with a high incidence. Increasing evidence suggests that oxidative stress (OS) generates high levels of reactive oxygen species (ROS) and free radicals and is a significant factor in aging and disease. However, the predictive value of OS-related long non-coding RNA (lncRNA) in CRC is unclear. We constructed a survival predictor based on OS-related lncRNAs obtained from the Cancer Genome Atlas (TCGA-COAD) to predict the prognosis of patients with CRC. The feature includes four lncRNAs, LINC02474, AL513550.1, SNHG16, and AL161729.4. The OS-related lncRNA signature demonstrated superior predictive accuracy to conventional clinicopathological parameters, with an area under the receiver operating characteristic curve (AUC) of 0.768, indicating its potential as a robust prognostic biomarker. Furthermore, the nomogram based on risk scores and clinicopathological variables (age, gender, grade, stage, M stage, and N stage) showed strong predictive performance. High-risk patients were found to be more sensitive to treatment drugs ABT.263, AZD.0530, gefitinib, imatinib, PAC.1, and shikonin. The predictive model we constructed can independently predict the prognosis of patients with CRC. Further experimental validation and mechanistic studies are warranted to elucidate the precise role of OS-related lncRNAs in CRC pathogenesis.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-026-04496-1.

Introduction

The global incidence and deaths from colorectal cancer (CRC) have more than doubled over the past three decades [1]. The incidence of colorectal cancer is expected to increase further as a result of population aging, longer life expectancy, and lifestyle changes [2, 3]. Although genetic predisposition, dietary patterns, and the gut microbiome, along with their metabolites, are known to contribute to CRC development, the precise mechanisms underlying CRC pathogenesis remain incompletely understood. Emerging evidence suggests that CRC progression is driven by multiple interconnected pathways, including Wnt/β-catenin signaling, PI3K/Akt activation, and gut microbiota-induced inflammatory responses. Despite advances in targeted therapy [4] and immunotherapy [5], challenges such as acquired resistance and immune evasion continue to limit the efficacy of treatment in CRC [6]. The significant difference in survival rates (69.2% vs. 11.7%) highlights the impact of early diagnosis, genetic factors, and individualized therapeutic strategies [7]. Accordingly, numerous studies have investigated the prognostic value of various molecular markers in CRC [8–10].

However, only a few biomarkers have been implemented in clinical practice. The established clinical biomarkers for colon cancer include CEA, KRAS, BRAF, MSI, and ctDNA, which are primarily used to aid diagnosis, monitor treatment response, assess recurrence risk, and guide targeted therapy. Notably, no single biomarker can fully diagnose colon cancer. Achieving high specificity, accuracy, and diagnostic efficiency remains a key objective, often requiring a comprehensive evaluation integrating multiple test results. Moreover, predicting prognosis and treatment response remains a significant challenge due to the complex and heterogeneous nature of the disease [11].

Non-coding RNAs (ncRNAs) are a class of RNA molecules produced during genomic transcription that do not directly code for proteins. In recent years, ncRNAs have been involved in the development and progression of colon cancer (Chen & Shen, 2020; Tang et al., 2019). Long non-coding RNAs (LncRNAs) with a length longer than 200 nucleotides participate in multiple biological processes, including cell proliferation, differentiation, development, apoptosis and metastasis, often by serving as a competing endogenous RNA (ceRNA) to regulate the expression of specific miRNAs, and then target molecules downstream of these miRNAs [12]. We are optimistic that using new, powerful sequencing technologies will elucidate the role of lncRNAs involved in colon tumorigenesis and ultimately accelerate the clinical application of lncRNAs in diagnosis, treatment, and prognosis assessment.

Oxidative stress (OS) [13] results from an imbalance between reactive oxygen species (ROS) production and antioxidant defenses, leading to cellular damage. In CRC, OS contributes to genomic instability, inflammation, and therapy resistance. ROS are highly reactive molecules that can cause damage to cellular structures, including DNA, proteins, and lipids. Strategies to enhance antioxidant defenses or induce oxidative stress selectively in cancer cells are being developed as potential therapeutic approaches.

Accumulating evidence has demonstrated that high levels of ROS can contribute to tumor progression by promoting cell proliferation, migration, survival, and therapeutic resistance [14]. To date, it has been reported that three critical signaling pathways mediate oncogene-induced ROS production, including the PI3K/Akt/mTOR signaling pathway [15], the MAPK signaling pathway [16], and the NF-κB signaling pathway [15]. Targeting oxidative stress in cancer treatment is an active area of research, with several potential therapeutic approaches being investigated [16]. Therefore, unraveling the molecular mechanisms underlying colorectal cancer is crucial for identifying reliable biomarkers to facilitate early detection and monitor disease progression, ultimately improving patient outcomes. To address this gap, this study aimed to develop a prognostic model based on OS-related lncRNAs and evaluate its ability to predict CRC patient outcomes. We identified a novel four-lncRNA signature associated with oxidative stress and demonstrated that this signature serves as a robust independent prognostic factor. Notably, our OS-related lncRNA signature demonstrated improved prognostic accuracy compared to traditional clinicopathological variables, providing insights into the interplay between oxidative stress, tumor immunity, and therapy response in CRC.

Materials and methods

Patients and datasets

We obtained standardized RNA sequencing (RNA-seq) data in fragments per kilobase of transcript per million mapped reads (FPKM) format, along with corresponding clinical and prognostic information for colorectal cancer (CRC) patients, from The Cancer Genome Atlas-Colon Adenocarcinoma (TCGA-COAD) database via the Genomic Data Commons (GDC) portal (https://portal.gdc.cancer.gov/); Data were obtained for 547 patients with lncRNA expression levels and survival times. We obtained disease-free survival (DFS) data for CRC patients from the cBioPortal database (https://www.cbioportal.org/); patients with incomplete clinical data were excluded. Thus, 417 CRC patients were ultimately included in the study. 1093 genes related to oxidative stress were retrieved from GeneCards (https://www.genecards.org/), and the limit values were set to a relevance value > 7 [17].

Differentially expressed oxidative stress-related lncRNAs

With |log2FC|> 2 and false discovery rate (FDR) < 0.005, we used the R package “limma” to identify the oxidative stress-related differentially expressed genes (DEGs), We performed Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses in the “ggplot2” package.

Construction of the oxidative stress-related lncRNA predictive signature

We used the “limma” package to calculate the correlation between oxidative stress-related genes and lncRNAs. A correlation coefficient (R) > 0.3 and a p-value of 0.001 were considered significant. We first used univariate Cox regression analysis to obtain oxidative stress-related lncRNAs associated with the prognosis of CRC patients. Then, we performed multivariate Cox regression analysis to identify oxidative stress-related lncRNAs for the Construction of a predictive signature for lncRNAs related to oxidative stress. Detailed preprocessing methods (normalization, filtering criteria) have been included. Variables in Cox regression were selected based on univariate Cox regression (p < 0.05). To address multiple comparisons, we applied false discovery rate (FDR) correction (Benjamini–Hochberg method). The analytical scripts will be made available upon request. The computational formula used for this analysis was as follows:

X represents the expression value of selected OS-related lncRNAs, and coef represents the coefficient value. This formula was used to calculate the risk score of each CRC patient. Patients were divided into two groups based on their median risk score: a low-risk group and a high-risk group. The log-rank test was used to compare differences in survival rates between groups. Based on Pearson correlation, the DElncRNA-mRNA coexpression network associated with oxidative stress was visualized using Cytoscape. The Sankey diagram of prognostic DElncRNAs related to oxidative stress was visualized using the “ggalluvial” R package.

Survival, ROC and analysis

Kaplan–Meier (K-M) survival curve analysis was performed for OS rates in the entire data set using the R packages “survival” and “survminer.” Subsequently, the “timeROC” package was also used for analyzing the receiver operating characteristic (ROC) curve. The area under the curve (AUC) value was calculated to evaluate the predicted accuracy of the prognostic signature of DElncRNA related to oxidative stress.

Nomogram and calibration

Using the R packages “survival,” “regplot,” and “RMS,” we combined the risk score with the clinicopathological characteristics of age, gender, stage, tumor size (T), and lymph node metastasis (N) to create a nomogram that can predict the 1-, 3- and 5-year survival of CRC patients. Multivariate Cox regression analysis was performed with backward stepwise selection, adjusting for potential confounders, and p-values were corrected using the Benjamini–Hochberg method. We used a calibration curve to test whether the predicted survival rate matched the actual survival rate.

Immune infiltration analysis of the lncRNA signature

Colorectal cancer patients were divided into high- and low-risk groups based on their median risk score. Increased and decreased lncRNAs were enriched separately using GSEA v4.1.0 (http://www.broad.mit.edu/gsea), with p < 0.05 and FDR < 0.25 as thresholds for statistical significance. The infiltration levels of 16 immune cells and the activities of 13 immune-related functions were calculated by single-sample gene set enrichment analysis (ssGSEA) using the R package “GSVA.”

The predictive role of our signature related to clinical therapeutics

To assess the predictive role of the signature on clinical response to treatment, we calculated the IC50 of commonly used chemotherapeutic agents for the clinical treatment of CRC. The Wilcoxon signed-rank test was used to compare the IC50 values between the high-risk and low-risk groups.

Patient samples and qPCR validation

To validate the expression trends of the four lncRNAs in clinical settings, tumor and adjacent normal tissues were collected from 16 Chinese patients with CRC who were treated at our hospital. Total RNA was extracted using the FastPure^® Cell/Tissue Total RNA Isolation Kit V2 (NoviZan, China) according to the manufacturer’s instructions. Briefly, approximately 10–20 mg of fresh tissue was homogenized in 500 μL of Buffer RL. Following sequential ethanol precipitation and column-based purification, RNA was eluted using RNase-free water. cDNA was synthesized and qPCR was performed to assess lncRNA expression levels (Fig. S1). All experimental procedures were conducted under RNase-free conditions to ensure the integrity of the RNA.

TNM stage stratification and expression trend analysis

For clinical validation, we further extracted TNM staging information (pT, N, M) from the same patient cohort. A TNM score was calculated by assigning weights (T = actual value; N = actual value; M1 = 5, M0 = 0), enabling numeric trend modeling. We then plotted the average expression of each lncRNA across TNM score groups and applied non-linear regression (spline interpolation with Savitzky-Golay smoothing) to fit the trends.

Statistical analysis

All computational and statistical analyses were performed using R (v4.2.3) software. The Wilcoxon test was used to analyze the expression levels of oxidative stress-related DEGs in cancer tissues and normal tissues. Principal component analysis (PCA) was used to determine the distribution of patients with varying risk scores. Statistical tests were bilateral, with p < 0.05 being significant.

Results

Screening and enrichment of differential genes

This study identified 104 OS-related differentially expressed genes (DEGs) associated with CRC, comprising 49 upregulated genes and 55 downregulated genes (Fig. 1A, B). Kyoto Encyclopaedia of Genes and Genomes (KEGG) pathway analysis showed that these DEGs were primarily enriched in pathways related to Neuroactive ligand-receptor interaction, IL-17 signaling pathway, Chemical carcinogenesis − receptor activation, Drug metabolism-cytochrome P450, Salivary secretion (Fig. 1C). Gene Ontology (GO) analysis demonstrated that OS-related DEGs were significantly associated with biological processes related to response to xenobiotic stimulus, signal release, response to oxidative stress, learning or memory; (b) cellular composition associated with the neuronal cell body, presynapse, distal axon, membrane raft; and (c) molecular functions related to the receptor-ligand activity, signaling receptor activator activity, cytokine activity, heme binding, tetrapyrrole binding (Fig. 1D).

Fig. 1.

GO and KEGG analyses of OS-related DEGs in cancer and adjacent tissues. A Volcano plot of 104 OS-related genes in CRC. Red and blue dots represent up-regulated and down-regulated genes, respectively. B Heat map of OS-related DEGs. C KEGG analysis of OS-related DEGs. D GO analysis of OS-related DEGs. GO, Gene Ontology; KEGG, Kyoto Encyclopaedia of Genes and Genomes; DEGs, differentially expressed genes; FC, fold change; FDR, false discovery rate; BP, biological process; CC, cellular components; MF, molecular function

Construction of a predictive signature for OS-related lncRNAs

A total of 814 OS-related lncRNAs were identified based on correlation analysis with OS-related genes. Through univariate Cox regression analysis, 30 lncRNAs were initially identified as being associated with patient prognosis. Multifactorial Cox regression analysis revealed that 4 OS-associated lncRNAs, including LINC02474, AL513550.1, SNHG16, and AL161729.4, were used to construct a predictive signature. The expression of these four lncRNAs is shown in Fig. 2A.

Fig. 2.

Expression and lncRNA–mRNA network of four OS-related lncRNAs in the predictive signature. A Expression of four OS-related lncRNAs in CRC and normal tissues. B The co-expression network of prognostic OS-related lncRNAs. C Sankey diagram of prognostic OS-based lncRNAs. lncRNA, long non-coding RNAs; N, normal; T, tumor

We further visualized lncRNAs using Cytoscape and the ggalluvial R package. AL513550.1 was co-expressed with 5 OS-related genes (BRCA2, NDUFV2, SLC7A11, CXCL8, NOX4), SNHG16 was co-expressed with 2 OS-related genes (BRCA2, NDUFV2), AL161729.4 was co-expressed with 2 OS-related genes (BRCA2, NDUFV2), LINC02474 was co-expressed with NDUFV2, Fig. 2B, C.

The risk score was calculated as follows: Risk Score = (0.724 × LINC02474 expression) + (0.87 × AL513550.1 expression) + (-1.217 × SNHG16 expression) + (0.734 × AL161729.4 expression).

Correlation between the predictive signature and the prognosis of CRC patients

The risk score for each patient was calculated using the formula, and patients were divided into two groups: high-risk (n = 209) and low-risk (n = 205) based on the median risk score. The results showed that the overall survival rate of the high-risk group was significantly lower than that of the low-risk group (Fig. 3A, p < 0.001). The 5-year survival rates in the high-risk and low-risk groups were 17.7% and 28.1%, respectively. At the same time, the risk scores and clinical status differed significantly between the high-risk and low-risk groups (Fig. 3B, C).

Fig. 3.

Correlation between the predictive signature and prognosis of patients with CRC. A Kaplan–Meier analysis of the overall survival rate of patients with CRC in the high-risk and low-risk groups. B The distribution of risk score among patients with CRC. C The number of dead and alive patients with different risk scores. Blue and yellow represent the numbers of survivors and deaths, respectively. D Forest plot for univariate Cox regression analysis. E Forest plot for multivariate Cox regression analysis. F The ROC curve of the risk score and clinicopathological variables. G ROC curve and AUCs at 13- and 5-year survival for the predictive signature. CRC, colorectal carcinoma; ROC, receiver operating characteristic; AUC, area under the curve; T, tumor; N, lymph node

Univariate Cox regression analysis showed that the grade, overall stage, T stage, N stage, and M stage, and risk scores of CRC patients were significantly correlated with overall survival (Fig. 3D). Multivariate Cox regression analysis showed that age, T stage, and risk score were independent predictors of overall survival (Fig. 3E). The area under the curve (AUC) value of the risk score was 0.768, which was better than all clinicopathological variables in predicting patient outcomes (Fig. 3F). The AUC values for 3-year survival and 5-year survival were 0.754 and 0.742, respectively, indicating good predictive performance (Fig. 3G).

To further predict the prognosis of patients with CRC, we constructed a nomogram containing clinicopathological variables and risk scores to predict the prognosis of patients with CRC at 1, 2, 3, and 5 years (Fig. 4A). The calibration curve showed excellent agreement between the actual overall survival rates and the predicted survival rates for 1, 2, 3, and 5 years (Fig. 4B–E).

Fig. 4.

Construction and verification of the nomogram. A Nomogram combining clinicopathological variables and risk scores to predict overall survival at 1, 2, 3, and 5 years of patients with CRC. B–E The calibration curves test the agreement of actual overall survival rates with predicted survival at 1, 2, 3, and 5 years

Relationship between predictive features and prognosis of CRC patients under different clinicopathological variables

To investigate the association between predictive characteristics and prognosis in CRC patients classified according to various clinicopathological variables, patients were divided into several groups based on age, sex, grade, T stage, and N stage. For each different classification, patients in the high-risk group had significantly shorter OS than those in the low-risk group (Fig. 5). These results suggest that predictive functions can predict outcomes in patients with colorectal cancer independent of clinicopathological variables.

Fig. 5.

Kaplan–Meier survival curves of high-risk and low-risk groups were classified according to different clinicopathological variables

Internal validation of predictive features

To confirm the reliability of the prediction features, we randomly divided 507 features into two groups: 104 patients for the training set and 104 patients for the test set. In both studies, patients in the high-risk group had a lower overall survival rate than those in the low-risk group (Fig. 6A and B, p < 0.05), and the receiver operating characteristic (ROC) curves in both groups demonstrated good predictive performance. The AUC values for 1-year, 3-year, and 5-year survival were 0.81, 0.785, and 0.825 in the training set and 0.748, 0.734, and 0.759 in the testing set (Fig. 6C, D).

Fig. 6.

Internal validation of total survival prediction features based on the entire TCGA dataset. A Kaplan–Meier survival curve of the training cohort. B Kaplan–Meier survival curve of the test cohort. C ROC curves and AUC values for the training cohort’s 1-year, 3-year, and 5-year survival. D ROC curves and AUC values for the test cohort’s 1 -, 3 -, and 5-year survival. TCGA, Cancer Genome Atlas; ROC, receiver operating characteristics; AUC, area under the curve

Analysis of immune cell infiltration

We used principal component analysis maps to visualize the distribution of patients based on the whole genome, OS-associated genome, OS-associated lncRNAs, and the four lncRNAs identified as the prognostic signatures of OS. Patients with high-risk and low-risk scores in the lncRNA signature group were distributed in different quadrants, indicating that this approach is the most effective way to differentiate among patients (Fig. 7A–D). The results showed DCs, plasmacytoid dendritic cells (pDCs), and T helper type 1 (Th1) cells were significantly varied between the groups discussed (Fig. 7E). Co-inhibition of antigen-presenting cell, cytokine–cytokine receptor interaction (CCR), Checkpoint, HLA, para-inflammation, co-inhibition and co-stimulation of T cells were lower in the high-risk group than those in the low-risk group (Fig. 7F). Immune checkpoint analysis revealed differences in CD244, CD48, BTNL2, TNFRSF25, CD44, CD276, and CD274 in the high- and low-risk groups (Fig. 7G).

Fig. 7.

Patients in the high-risk and low-risk groups were divided into high-risk and low-risk groups with different immune statuses, immune infiltrating cells, and immune-related function scores. A Genome-wide distribution of patients. B OS-related gene set. C OS-related lncRNAs. D 4-lncRNA prognostic characteristics of OS. E Infiltration levels of 16 immune cells in high-risk and low-risk groups using the ssGSEA algorithm. F Correlation of predicted features with 13 immune-related functions. G Expression of immune checkpoints among high and low-risk groups. OS, oxidative stress; lncRNA, long non-coding RNA; ssGSEA, single sample gene set enrichment analysis; aDCs, activated dendritic cells; iDCs, immature dendritic cells; NK, natural killer; pDCs, plasmacytoid dendritic cells; Tfh, T follicular helper cells; Th1, type 1 helper T cells; TIL, tumor-infiltrating lymphocytes; Treg, regulatory T cells; APC, antigen-presenting cell; CCR, cytokine—cytokine receptor interaction; HLA, Human leukocyte antigen; MHC, major histocompatibility complex; IFN, interferon. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant

Association between predictive features and BC therapy

In addition to immunotherapy, we also analyzed the association between the predictive signature and the efficacy of general chemotherapy for BC. We note that ABT-263, AZD-0530, Gefitinib, Imatinib, PAC-1, and Shikonin are candidates for the treatment of high-risk patients. In contrast, BMS.509744, CEP.701, and Sorafenib may not be suitable for high-risk patients (Fig. 8).

Fig. 8.

Comparison of sensitivity to treating drugs across high- and low-risk groups. A ABT.263. B AZD.0530. C BMS.509744. D CEP.701. E IGefitinib. F Imatinib. G PAC.1. H ISorafenib. I Shikonin

Expression trend of candidate lncRNAs across TNM stages

Using patient samples from our hospital, we observed consistent upregulation of all four lncRNAs (lnc02474, AL513550.1, SNHG16, AL161729.4) in tumor tissues compared to adjacent normal tissues, as validated by qPCR (Fig. 9). Furthermore, TNM-based staging scores were used to model disease progression. Non-linear regression curves demonstrated a positive trend between TNM scores and lncRNA expression, with all four showing increased levels in more advanced stages (Fig. 9). This clinical validation reinforces the prognostic significance of these lncRNAs and supports their potential involvement in CRC progression.

Fig. 9.

Expression analysis and clinical relevance of four selected lncRNAs in tumor tissues. A Box plots showing the expression levels (log₂ TPM + 1) of four lncRNAs (AL161729.4, AL513550.1, LINC02474, and SNHG16) in tumor (red) versus normal (blue) tissues based on TCGA datasets. Expression differences were analyzed using a two-tailed t-test. B qRT-PCR validation of lncRNA expression in paired tumor (T, red) and adjacent normal (N, blue) tissues. Data are presented as mean ± SD. Statistical significance is indicated as: ***P < 0.001; **P < 0.01; ****P < 0.0001. C Line graph showing the correlation between tumor size (mm) and the relative expression levels of the four lncRNAs. Expression levels increase with tumor size, particularly after 40 mm. D Spline regression curves showing the relationship between lncRNA expression and TNM stage. The expression patterns of the lncRNAs vary across TNM stages, indicating potential involvement in tumor progression

Discussion

As the third most common malignancy, colorectal cancer has become a severe economic burden and a significant global health challenge. Despite advancements in early detection and treatment, the high risk of recurrence and metastasis remains a primary concern, leading to poor clinical outcomes [18]. Therefore, accurate prognostic assessments are essential for improving treatment strategies. OS plays a critical role in cancer progression; however, few long non-coding RNA (lncRNA) biomarkers have been explored to elucidate its impact on CRC pathophysiology [19]. Further research on OS-related lncRNAs is expected to provide guidelines for clinical decision-making.

In this study, we identified 104 differentially expressed OS-related DEGs. KEGG enrichment analysis revealed that these genes were primarily associated with Neuroactive ligand-receptor interaction, IL-17 signaling pathway, Chemical carcinogenesis − receptor activation, and Drug metabolism-cytochrome P450. Notably, the neuroactive ligand-receptor interaction pathway was overactivated in tumors with high homologous recombination deficiency and was associated with immunosuppression in colon cancer with high homologous recombination deficiency. Furthermore, the signaling pathway has been associated with prognosis and response to immunotherapy in colorectal cancer [20]. The inhibition of this pathway has been shown to enhance anti-tumor immunity and improve the efficacy of colon cancer, suggesting that neuroactive ligand-receptor signaling contributes to immune evasion in CRC. Recent studies have shown that IL-17A-induced mTOR activation regulates HIF1α gene transcription and protein levels, with HIF1α and glycolysis contributing to tumor progression and metastasis. The IL-17 signaling pathway, driven by the pro-inflammatory cytokine IL-17, thus promotes a chronic inflammatory milieu in CRC. For example, IL-17A activation of the mTOR/HIF-1α axis enhances glycolytic metabolism in tumor cells, thereby accelerating tumor progression. This IL-17A–HIF1α axis exemplifies how inflammatory signaling and oxidative stress may intersect in CRC, offering potential therapeutic targets within the IL-17 pathway [21].

We identified four differentially expressed lncRNAs, including LINC02474, AL513550.1, SNHG16, and AL161729.4. Previously, LINC02474 has been validated as an oncogenic lncRNA in CRC [22], where it promotes metastasis while inhibiting apoptosis through the downregulation of GZMB expression [23]. AL513550.1 has been identified as a prognostic biomarker in oral squamous epithelial carcinoma [24] and lung adenocarcinoma [10], suggesting its potential role in multiple malignancies. SNHG16S and miR-124-3p were dysregulated in human colorectal tumors or cells; SNHG16 promotes colorectal cancer cell proliferation, migration, and epithelial-mesenchymal transition through miR-124-3p/MCP-1 [25]. The expression of AL161729.4 can be regulated by ferroptosis, a biological process that has been revealed to suppress CRC progression. However, the underlying mechanism of oxidative stress in colon cancer is unclear [26].

Kaplan–Meier survival analysis and ROC curve evaluations demonstrated that the OS-related lncRNA signature effectively stratifies CRC patients into high- and low-risk groups, with high-risk patients exhibiting significantly worse overall survival. Notably, this signature showed superior predictive accuracy compared to traditional clinicopathological variables, further confirming its robustness as an independent prognostic factor. The incorporation of this lncRNA signature into clinical practice could enhance prognostic assessment and aid in individualized treatment planning.

Notably, this signature showed superior predictive accuracy compared to traditional clinicopathological variables, further confirming its robustness as an independent prognostic factor. The incorporation of this lncRNA signature into clinical practice could enhance prognostic assessment and aid in individualized treatment planning. Importantly, our OS-related lncRNA model exhibits several distinct advantages over classical protein biomarkers such as CEA. For instance, our four-lncRNA signature achieved a higher prognostic accuracy (AUC = 0.768) than typically reported for CEA, indicating an improved ability to predict patient outcomes. Additionally, because the signature is derived from tumor-specific lncRNA expression profiles, it offers greater molecular specificity, being restricted mainly to cancerous tissue. In contrast, protein markers like CEA can be elevated in benign conditions and thus lack tumor exclusivity. Moreover, our model provides predictive insights into immunotherapy and targeted therapy responses (as evidenced by the distinct immune checkpoint expression and drug sensitivity patterns in high-risk patients), capabilities that conventional markers such as CEA do not possess. Finally, this lncRNA-based risk score can be integrated into clinical workflows using routine molecular assays (e.g., RT-PCR on tumor samples) and potentially even non-invasive liquid biopsy techniques to detect circulating tumor-derived lncRNAs, making it a clinically applicable tool for risk stratification and personalized treatment decisions. Immune cell infiltration plays a pivotal role in CRC progression, influencing tumor growth, metastasis, and patient prognosis [27]. Our analysis revealed a distinct immune landscape between high- and low-risk CRC patients. Specifically, the low-risk group exhibited higher infiltration of dendritic cells (DCs), plasmacytoid dendritic cells (pDCs), neutrophils, and Th1 cells, all of which contribute to anti-tumor immunity. Conversely, the high-risk group had significantly increased levels of CD244, CD48, CD44, CD276, and CD274 compared to the low-risk group. The elevated expression of multiple immune checkpoint molecules in the high-risk group suggests a more immunosuppressive tumor environment, potentially leading to a diminished response to immune checkpoint inhibitors (ICIs) in this subgroup. These findings underscore the potential value of combining OS-targeted therapeutic strategies with immunotherapy to improve outcomes for high-risk CRC patients.

Our drug sensitivity analysis revealed that high-risk patients demonstrated increased sensitivity to several targeted therapies, including ABT-263, AZD-0530, gefitinib, imatinib, PAC-1, and shikonin. These agents target key oncogenic pathways, suggesting that OS-related lncRNAs may play a role in modulating the response to drugs. Conversely, high-risk patients exhibited resistance to BMS.509744, CEP.701, and sorafenib highlighting the need for alternative therapeutic approaches in this subgroup. These findings provide a rationale for tailoring treatment strategies based on OS-related molecular profiles, potentially improving therapeutic efficacy through personalized medicine.

Study limitations

Although our results have potential clinical significance, they have some limitations. Further cell biology experiments and clinical studies are required to verify the results; however, due to the constraints in experimental conditions and funding, we were unable to perform additional in vitro or in vivo experiments to validate these findings at this time. We sincerely acknowledge this shortcoming and plan to conduct such validations in future studies to strengthen our results. In addition, the possible mechanisms, molecular interactions, and clinical applications of the prognostic genes in CRC also need to be investigated.

Conclusions

This study establishes a novel OS-related lncRNA signature as a reliable prognostic tool for CRC, offering valuable insights into the interplay between oxidative stress, tumor biology, and therapeutic response. The identified signature holds promises for enhancing risk stratification, informing personalized therapy, and advancing our understanding of OS-driven CRC pathogenesis. Future experimental validation and clinical trials are crucial for translating these findings into clinical practice, ultimately paving the way for more effective treatment strategies tailored to oxidative stress-related tumor vulnerabilities.

Author contributions

P.C., F.L., and J.M. contributed equally to this work. P.C. and F.L. designed the study and conducted the data analysis. J.M. performed the literature review and statistical validation. Y.W. assisted with data curation and manuscript editing. J.S. and L.L. contributed to data interpretation and manuscript revision. M.N. and L.L. supervised the project and provided critical revisions. All authors reviewed and approved the final manuscript.

Funding

This work was supported by grants from the National Science Foundation of Jiangsu Province (BK20220722), Foundation of National Natural Science (82200731), Foundation of China Postdoctoral Science (2022TQ0131 and 2022M711409).

Data availability

The datasets analysed during the current study are publicly available. RNA sequencing data of colorectal cancer patients were obtained from The Cancer Genome Atlas Colon Adenocarcinoma (TCGA-COAD) project and are accessible via the Genomic Data Commons portal (https://portal.gdc.cancer.gov/). Disease-free survival (DFS) data were retrieved from the cBioPortal for Cancer Genomics repository (https://www.cbioportal.org/).

Declarations

Ethics approval and consent to participate

Sample collection and study procedures were carried out in accordance with the Declaration of Helsinki. This study was approved by the Ethics Committee of the First Affiliated Hospital of Nanjing Medical University (Jiangsu Province Hospital) (Approval number: 2021-SRFA-424, approved on May27, 2021). Patients and/or their legal guardians signed informed consent after receiving oral and written information.

Consent for publication

Not applicable. The manuscript does not contain any individual person’s data in any form.

Competing interests

The authors declare no competing interests.