Development and application of a predictive model for survival and drug therapy based on COVID-19-related lncRNAs in non-small cell lung cancer

Abstract

Numerous studies have substantiated the pivotal role of long non-coding RNAs (lncRNAs) in the progression of non-small cell lung cancer (NSCLC) and the prognosis of afflicted patients. Notably, individuals with NSCLC may exhibit heightened vulnerability to the novel coronavirus disease (COVID-19), resulting in a more unfavorable prognosis subsequent to infection. Nevertheless, the impact of COVID-19-related lncRNAs on NSCLC remains unexplored. The aim of our study was to develop an innovative model that leverages COVID-19-related lncRNAs to optimize the prognosis of NSCLC patients. Pertinent genes and patient data were procured from reputable databases, including TCGA, Finngen, and RGD. Through co-expression analysis, we identified lncRNAs associated with COVID-19. Subsequently, we employed univariate, LASSO, and multivariate COX regression techniques to construct a risk model based on these COVID-19-related lncRNAs. The validity of the risk model was assessed using KM analysis, PCA, and ROC. Furthermore, functional enrichment analysis was conducted to elucidate the functional pathways linked to the identified lncRNAs. Lastly, we performed TME analysis and predicted the drug sensitivity of the model. Based on risk scores, patients were categorized into high- and low-risk subgroups, revealing distinct clinicopathological factors, immune pathways, and chemotherapy sensitivity between the subgroups. Four COVID-19-related lncRNAs (AL161431.1, AC079949.1, AC123595.1, and AC108136.1) were identified as potential candidates for constructing prognostic prediction models for NSCLC. We also observed a positive correlation between risk score and MDSC, exclusion, and CAF. Additionally, two immune pathways associated with high-risk and low-risk subgroups were identified. Our findings further support the association between COVID-19 infection and neuroactive ligand-receptor interaction, as well as steroid metabolism in NSCLC. Moreover, we identified several highly sensitive chemotherapy drugs for NSCLC treatment. The developed model holds significant value in predicting the prognosis of NSCLC patients and guiding treatment decisions.

1. Introduction

Lung carcinoma remains a prevalent malignancy worldwide, constituting 11.4% of all cancer cases and emerging as the foremost cause of cancer-related mortality, accounting for approximately 18.0% of such deaths.^[1] Among the diverse subtypes, non-small cell lung cancer (NSCLC) predominates, encompassing approximately 85% of all lung cancer incidences.^[2] NSCLC, an acronym for non-small cell lung cancer, represents a cohort of profoundly malignant neuroendocrine tumors. These tumors exhibit an alarming propensity for rapid metastasis, a remarkably short tumor doubling time, and a notable vulnerability to drug resistance. Regrettably, a mere 26% of individuals diagnosed with NSCLC manage to endure beyond a 5-year threshold.^[3] Given the inherent constraints of conventional therapeutic approaches and the burgeoning domains of immunotherapy and targeted therapy, an urgent imperative emerges to investigate more refined biomarkers capable of augmenting the overall survival time (OS) of patients afflicted with NSCLC.

Long non-coding RNA (lncRNA), an RNA molecule exceeding 200 nucleotides in length, plays a pivotal and extensive role in various biological processes.^[4] Extensive research has substantiated the close association between dysregulated long non-coding RNAs (lncRNAs) and the pathophysiological mechanisms underlying NSCLC.^[5,6] For instance, the lncRNA POU6F2-AS2 plays a pivotal role in promoting the aggressiveness of nonsmall-cell lung cancer by orchestrating the upregulation of E2F3 through microRNA-125b-5p-mediated mechanisms.^[7] The lncRNA JPX has been found to promote an increase in lung cancer cell population and expedite the growth of tumor cells.^[8] Additionally, a splice isoform of the lncRNA PD-L1 has been observed to facilitate the development of LUAD by directly enhancing the activity of c-Myc.^[9] Therefore, it is worth studying the many unknown characteristics between lncRNA and NSCLC.

In recent years, the rapid global spread of the novel coronavirus disease (COVID-19) has raised significant concerns. This acute respiratory syndrome is caused by severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2).^[10] Notably, patients with cancer have been reported to exhibit heightened susceptibility to COVID-19 and a greater likelihood of experiencing severe manifestations of the disease.^[10] However, the prognosis of patients with NSCLC infected with COVID-19 is still unclear. As the exploration of the intricate interplay between lncRNAs and the immune system advances, the recognition of COVID-19-related lncRNAs (CRlncRNAs) as potential biomarkers is gradually unfolding. Nevertheless, the precise impact of COVID-19 infection on the dysregulation of lncRNAs in the development of NSCLC remains enigmatic. Literature review found that up to now, no scholars have constructed a model based on CRLncRNAs to link the prognosis and immunotherapy sensitivity of patients with NSCLC. Hence, we have decided to contribute in this direction. We aim to develop a prognostic model of non-small cell lung cancer based on COVID-19-related lncRNAs.

In this study, we meticulously scrutinized a comprehensive dataset encompassing the expression profiles of lncRNAs in NSCLC obtained from The Cancer Genome Atlas (TCGA), meticulously sieving through to identify CRlncRNAs with significant prognostic implications. Subsequently, we meticulously evaluated the prognostic value of these CRlncRNAs in NSCLC, while also exploring their potential impact on the tumor microenvironment and their therapeutic implications in the realm of drug therapy.

2. Materials and methods

2.1. Collection and screening of NSCLC patient data

We procured RNA sequencing data (RNA-seq data) (FPKM normalized), mutation data, and corresponding clinical data for 503 NSCLC samples from The Cancer Genome Atlas (TCGA) database - NSCLC cohort (https://portal.gdc.cancer.gov/). Nine patients with incomplete clinicopathological information were excluded from our study. A total of 494 patients with comprehensive clinical information (age, gender, race, tumor stage, survival, and death status) were randomly divided into a training cohort (n = 330) and a testing cohort (n = 164). The training cohort was utilized for the construction of predictive models, while the testing cohort was employed for validation purposes. All RNA sequence data, as well as microarray data, underwent normalization and log2 transformation using the R package. The retrospective nature of the study and the anonymized analysis of patient data led to the waiver of informed consent.

2.2. Identification of COVID-19-related LncRNA

We gathered 670 genes associated with COVID-19 from the COVID-19 Disease Portal (Human Phenotype) of the Rat Genome Database (RGD)^[11] (https://rgd.mcw.edu/rgdweb/portal/home.jsp?p=14). Subsequently, we obtained 932 distinct genes (count ≥ 30) from the COVID-19 Drug and Gene Set Library in Ma’ayan Laboratory (https://maayanlab.cloud/covid19/). Additionally, we downloaded pertinent genomic data from the Finngen database, applying a filtering criterion of P < 1e-6 (https://www.finngen.fi/en/covid-19). By subtracting duplicate counting genes, noncoding genes, and pseudogenes, we identified a total of 1450 COVID-19-related genes (CRGs). The patient datasets employed in this study were publicly accessible and derived from previous relevant investigations.

By performing co-expression analysis of genes related to COVID-19 with lncRNAs and visualizing the results using Sankey plots, we identified significant COVID-19-related lncRNAs. R software (version 4.1.3; https://www.r-project.org/) was used to filter lncRNAs from TCGA transcriptome data. To extract COVID-19-related mRNA expression in NSCLC and identify co-expressed CRLncRNAs with gene sets, Pearson correlation analysis (| cor | ≥0.4 and P < 10^(-70)) was performed using the “limma” package in R. Subsequently, we conducted univariate Cox regression analysis on the selected lncRNAs to identify CRLncRNAs highly correlated with prognosis.

2.3. Construction and validation of relevant prognostic features

The training cohort consisted of 330 patients, while the testing cohort included 164 patients. Univariate Cox regression analysis was performed on the training cohort to identify lncRNAs associated with prognosis (screening P value < .05). To obtain more refined features, we employed the least absolute shrinkage and selection operator regression analysis, which effectively handles data with complex covariance. This approach allowed us to construct penalty functions for feature selection.^[12–15] Finally, through multivariate Cox regression analysis, we identified CRLncRNAs associated with OS and established a predictive model for NSCLC prognosis (significance filtering criteria P value < .05). The following equations were used to calculate the risk scores (RS) of NSCLC patients:

$${\displaystyle \begin{array}{l}{\displaystyle Riskscore(RS)={\displaystyle \sum _{i=1}^N}Coef\left(i\right)\times x\left(i\right)}\end{array}}$$

Where N represents the number of CRLncRNAs, x(i) denotes the expression value of each CRLncRNA, and Coef(i) is the estimated regression coefficient of the respective lncRNA obtained from the multivariate Cox regression analysis.

We utilized the median RS as a criterion to categorize patients in the training cohort (n = 330) into low-risk subgroups (RS < median) and high-risk subgroups (RS > median). Similarly, patients in the training cohort (n = 164) and the entire cohort (n = 494) were classified into high- and low-risk subgroups. To assess the impact of CRLncSig on the prognosis of NSCLC patients, we employed the Kaplan–Meier (KM) method to generate survival curves. The “survival” package in R was utilized to validate grouped clinical data, including age, sex, race, stage, survival status (fustat), AJCC stage, primary tumor (T), distant metastasis (M), and regional lymph nodes (N). Subsequently, univariate COX regression analysis and multivariate COX regression analysis were conducted to evaluate the predictive value of RS based on seven clinicopathological factors (age, gender, race, etc.), thereby verifying the independent predictive significance of CRLncSig. Receiver operating characteristic (ROC) curves were constructed for 1, 3, and 5 years using the “timeROC” package of R software to assess the accuracy of CRLncSig, with the area under the ROC (AUC) serving as the evaluation metric.^[16–18] Additionally, the concordance index (C-index) curve, calculated by the survConcordance function of the “survival” package, was employed for revalidation. Principal component analysis (PCA) was employed to explore the distribution of genes among different risk groups.

2.4. Construction of the predictive nomogram

We devised a Nomogram employing multivariate Cox regression analysis and the “rms” R package to prognosticate patient clinical outcomes. The Nomogram allocated scores to the seven clinicopathological factors (age, gender, ethnicity, stage, and tumor-node-metastasis staging) as well as the RS of the model, facilitating the computation of the 1-, 3-, and 5-year OS rates for NSCLC patients. Calibration plots, encompassing predictive calibration curves and standard curves, were employed to assess the predictive accuracy of the Nomogram.

2.5. Functional and mechanistic enrichment analysis of related LncRNAs

Drawing upon the aforementioned screened lncRNAs, we delved into the genes linked to noteworthy disparities between risk subgroups utilizing the “limma” package of R software. Gene Set Enrichment Analysis algorithm served as the foundation for conducting Gene Ontology (GO, available at http://geneontology.org/) and Kyoto Encyclopedia of Genes and Genomes (KEGG, accessible via http://www.genome.jp/kegg/) analyses. To accomplish enrichment analysis and visualization of GO and KEGG elements, we employed the “lusterProfiler” package. For gene ID conversion, we relied on the “org.Hs.e.g..db” package, a comprehensive human gene annotation resource. Furthermore, the “clusterProfiler” package facilitated the retrieval of corresponding pathway annotation information from the KEGG API. Various visualization techniques were implemented using the “enrichplot” package. Our GO/KEGG analysis was grounded in the widely employed DAVID database, renowned for gene enrichment and functional annotation analysis (https://david.ncifcrf.gov/). A significance threshold of P = .05 was utilized for the analysis, with P < .05 indicating substantial enrichment of functional annotations. We used bar plots to show differentiated lncRNAs associated with molecular function (MF), cellular component (CC), and biological process (BP).

Additionally, employing the “gsva” package of R software, we computed the immune score and activity of thirteen immune-related pathways for each non-small cell lung cancer (NSCLC) patient through single-sample gene set enrichment analysis (ssGSEA).

2.6. Analysis of the tumor microenvironment

In order to elucidate the relationship between risk scores and the immune milieu, we conducted a comprehensive analysis encompassing Tumor Mutational Burden (TMB) differentials, Tumor Immune Escape and Immunotherapy Analysis (TIDE), as well as an examination of immune checkpoint-associated genes in patients with NSCLC.^[19,20] TMB, a measure of the relative number of mutations within a tumor sample, has been demonstrated to correlate with the efficacy of immunotherapy.^[21] We scrutinized the disparities in TMB and survival rates between high-risk and low-risk subgroups based on TMB data obtained from TCGA-NSCLC patients. Subsequently, we acquired the tumor TIDE score file from the TIDE website (http://tide.dfci.harvard.edu/). TIDE is a computational framework that evaluates the potential of tumor samples to evade immune responses, utilizing gene expression profiles as a means to predict patient OS.^[22] Additionally, we explored the expression levels of well-known immune checkpoint genes and immune cell populations within the high-risk and low-risk groups. These encompassed IFGN, microsatellite instability (MSI) score, Merck18, CD274, CD8, cancer-associated fibroblasts (CAF), tumor-associated macrophage M2 (TAMM2), and myeloid-derived suppressor cells (MDSC).^[23–27]

2.7. Assessment of chemotherapy drug sensitivity

The IC50, representing the concentration required for 50% inhibition, serves as a measure of the chemotherapeutic agent’s affinity for its target. A lower value indicates a stronger affinity. To predict the response to chemotherapy drugs, we utilized the “pRRophetic” package, which utilizes the expression level of the oncogene. Leveraging the Genomics of Drug Sensitivity in Cancer database (https://www.cancerrxgene.org/), we employed the “pRRophetic” R package along with its dependent packages (“genefilter” and “ridge preprocessCore,” among others) to estimate the IC50 of the drugs. For visual representation, we employed the “ggplot2” R package to create box plots. This approach enabled us to identify potential drugs for treating NSCLC patients.

2.8. Calculation of the stemness indices

The Stemness Indices, which describe the similarity between cancer cells and stem cells, were used to predict tumor recurrence risk and guide treatment decisions.^[28] Recognizing that the stemness index of tumors is a crucial characteristic, we conducted gene expression analysis focused on stem cells in the TCGA-NSCLC patient cohort. Human stem cell data were sourced from the Progenitor Cell Biology Consortium (https://www.synapse.org). Employing the OCLR algorithm (One Class Linear Regression), we quantified the stemness of tissue samples and calculated the corresponding mRNAsi, which reflects the gene expression profile of stem cells. By calculating the mRNAsi of patients, we delved into its intricate correlation with cancer progression, prognosis, as well as four clinicopathological factors, namely gender, AJCC stage, T stage, and M stage.

3. Results

3.1. Identification of COVID-19-related lncRNA

The methodology employed in this investigation is depicted in Figure S1, Supplemental Digital Content, http://links.lww.com/MD/O81. By means of Pearson correlation analysis of 1450 differentially expressed CRGs and lncRNAs within samples obtained from the TCGA-NSCLC cohort, we successfully identified CRlncRNAs (with corFilter > 0.4 as the screening criteria). Furthermore, we utilized the “ggplot2” package within the R software to construct Sankey plots, visually representing the correlations between these CRlncRNAs (Figure S2, Supplemental Digital Content, http://links.lww.com/MD/O81). Our findings revealed that LINC01537, LINC00707, AC108136.1, LINC02320, and AC025419.1 exhibited the most significant associations with patient prognosis. The lncRNAs exhibiting high correlation, with P values < .0005, are presented in Table S1, Supplemental Digital Content, http://links.lww.com/MD/O79.

3.2. Development and validation of prognostic models

No statistically significant differences were observed in the clinical characteristics between the training cohort (n = 330) and the testing cohort (n = 164) (P > .05), implying promising grouping results. Initially, by combining gene set related lncRNAs with patient survival data, we conducted univariate Cox analysis on gene set related lncRNAs and identified 82 lncRNAs associated with cancer prognosis. The graph in Table S2, Supplemental Digital Content, http://links.lww.com/MD/O79 showcases the 41 most significant CRlncRNAs. Subsequently, least absolute shrinkage and selection operator regression analysis was employed to mitigate the issue of overfitting, resulting in the identification of 6 promising CRlncRNAs (AP000695.2, AL161431.1, AC079949.1, AC123595.1, AC108136.1, and LINC00707) (Figure S3, Supplemental Digital Content, http://links.lww.com/MD/O81). Finally, multivariate COX regression analysis was conducted to further identify 4 most correlated CRlncRNAs (AL161431.1, AC079949.1, AC123595.1, and AC108136.1), with AC108136.1 having the highest regression coefficient and contributing the most to the multiple Cox model (Table S3, Supplemental Digital Content, http://links.lww.com/MD/O79). We calculated risk scores for patients within the training cohort and divided them equally into 2 distinct risk subgroups using the median risk score (RS = 0.8916) as the threshold. Employing RS = 0.8916 as the threshold, patients within the test cohort were categorized into high-risk (n = 82) and low-risk (n = 82) subgroups. The entire cohort was also divided into high-risk (n = 247) and low-risk (n = 247) subgroups. Figure 1 depicted the prognostic model’s predictive performance, specifically the 4-CRlncRNA prognostic model. The Kaplan–Meier (KM) survival curves revealed that the high-risk subgroup had a lower OS compared to the low-risk subgroup (P < .01) (Fig. 1A). Patients were ranked in ascending order of risk score (Fig. 1B). Patients were ranked in ascending order of risk score (Fig. 1C). The survival analysis results from the testing cohort and the entire cohort validated the findings of the training cohort (Fig. 1D–I). Figure 1J to L exhibited the differential expression of the 4 CRLncSig among various risk subgroups. In the heat map, red represents high expression and blue represents low expression. The high risk of NSCLC was associated with AL161431.1 (HR = 1.202, 95% CI: 1.074–1.346, P < .01), AC079949.1 (HR = 1.194, 95% CI: 1.002–1.423, P < .05) and AC108136.1 (HR = 1.855, 95% CI: 1.419–2.426, P < .05), while AC123595.1 (HR = 0.692, 95% CI: 0.516–0.928, P < .05) was related to the low risk of NSCLC.

Figure 1.

Construction and validation of CRLncSig. Kaplan–Meier survival curves, distribution of risk scores and survival status in the training cohort (A–C). Kaplan–Meier survival curves, distribution of risk scores and survival status in the training cohort and the entire cohort (D–I). Heat maps of 4 CRLncSig in the training cohort (J), the testing cohort (K) and the entire cohort (L). CRLncSig = COVID-19-related lncRNA signature.

3.3. Correlation between risk scores and clinicopathological factors

The findings indicated that risk scores exhibited an upward trend with higher AJCC stage, N stage, and M stage (Figure S4A–D, Supplemental Digital Content, http://links.lww.com/MD/O81). However, there was no significant correlation observed with age (P = .086), race (P > .25), gender (P = .23), and T stage (P ≥ .05) (Figure S4E–H, Supplemental Digital Content, http://links.lww.com/MD/O81). Furthermore, our findings suggest that patients with higher risk scores have an increased likelihood of mortality.

3.4. Identification of factors associated with prognosis in patients with NSCLC

To validate the independent prognostic ability of the prognostic model in NSCLC patients, we conducted univariate and multivariate COX regression analyses on clinicopathological factors and RS. Univariate analysis revealed significant associations between patient overall survival (OS) and AJCC stage (P value < .01), T stage (P value = .01), and N stage (P value < .01) (Figure S5A, Supplemental Digital Content, http://links.lww.com/MD/O81). However, multifactorial analysis demonstrated that none of these factors were significantly linked to patient OS (Figure S5B, Supplemental Digital Content, http://links.lww.com/MD/O81). The prognostic model’s performance was evaluated using ROC curves, demonstrating its favorable predictive efficacy for patient OS at 1 year, 3 years, and 5 years (Figure S5C, Supplemental Digital Content, http://links.lww.com/MD/O81). The AUCs were 0.691, 0.693, and 0.703 for 1-year, 3-year, and 5-year OS, respectively. The ROC curves for 1-year RS and clinicopathological factors are depicted in Figure S5D, Supplemental Digital Content, http://links.lww.com/MD/O81 showing the AUCs for AJCC staging (0.698), RS (0.691), N staging (0.652), T staging (0.633), age (0.470), gender (0.554), race (0.497), and M staging (0.497). The ROC curves for 3-year RS, 5-year RS, and their corresponding clinicopathological factors are illustrated in Figure S5E, F, Supplemental Digital Content, http://links.lww.com/MD/O81. These findings suggest that AJCC staging holds promise as a reliable predictor of patient prognosis, while age, race, and M staging may not be suitable for prognostic prediction.

3.5. Validation of the predictive independence of the prognostic model

Subsequently, we amalgamated clinicopathological variables and risk scores to construct a nomogram (Fig. 2A). The C-index curve once again affirms the commendable predictive capability of AJCC staging (index > 0.5) (Fig. 2B). The nomogram was employed to visually and accurately prognosticate the 1-year, 3-year, and 5-year OS for patients. We noted that when the cumulative score reached 359, the 1-year OS stood at 90.9%, the 3-year OS at 63.3%, and the 5-year OS at 34.4%. The 1-year, 3-year, and 5-year calibration curves exhibited a favorable correspondence between the nomogram-predicted survival and actual survival (Fig. 2C). This attested to the commendable predictive performance of the nomogram.

Figure 2.

Prognostic model constructed from risk scores and clinicopathological factors. (A) We constructed a nomogram model to predict the probability of survival of NSCLC patients at 1, 3, and 5 years. Nomogram calculates the total score for each patient by assigning a score to the level at which each variable is taken. The predicted total score on the bottom scale represents the probability of survival at 1, 3, and 5 years. (B) C-index curves showing clinical features with significant prognostic impact. (C) Calibration curves of the nomogram showing the agreement between predicted and observed 1-, 3-, and 5-year OS. NSCLC = on-small cell lung cancer, OS = overall survival.

3.6. Clinical subgroup analysis of prognostic models

To ascertain the prognostic model’s capacity to forecast the OS of patients based on distinct clinicopathological variables, we further conducted a clinical subgroup survival analysis. These clinical subgroups were stratified by age, gender, race, AJCC stage, T stage, N stage, and M stage. The findings revealed that high-risk scores were significantly associated with male patients (P < .01, Figure S6A, Supplemental Digital Content, http://links.lww.com/MD/O81), female patients (P < .01, Figure S6B, Supplemental Digital Content, http://links.lww.com/MD/O81), patients aged 65 years or younger (P = .009, Figure S6C, Supplemental Digital Content, http://links.lww.com/MD/O81), patients older than 65 years (P = .008, Figure S6D, Supplemental Digital Content, http://links.lww.com/MD/O81), black or African American patients (P = 0006, Figure S6E, Supplemental Digital Content, http://links.lww.com/MD/O81), white patients (P = .002, Figure S6F, Supplemental Digital Content, http://links.lww.com/MD/O81), patients with AJCC stage II (P = .02, Figure S6G, Supplemental Digital Content, http://links.lww.com/MD/O81), patients with T2 (P = .032, Figure S6H, Supplemental Digital Content, http://links.lww.com/MD/O81), patients with T3 (P = .001, Figure S6I, Supplemental Digital Content, http://links.lww.com/MD/O81), patients with N0 (P = .01, Figure S6J, Supplemental Digital Content, http://links.lww.com/MD/O81), and patients with M0 (P < .001, Figure S6K, Supplemental Digital Content, http://links.lww.com/MD/O81).

Conversely, patients with AJCC stage I, AJCC stage III, AJCC stage IV, T1, T4, N1, N2, and M1 did not exhibit any significant association with risk scores (Figure S6L–S, Supplemental Digital Content, http://links.lww.com/MD/O81). We excluded the discussion of the results pertaining to the sample size of Asian patients, as it was too small to be included (Figure S6T, Supplemental Digital Content, http://links.lww.com/MD/O81).

3.7. Results of PCA

Differences in the differentiation of genes in the TCGA-NSCLC cohort, the set of COVID-19-related genes in the NSCLC cohort, co-expressed associated lncRNAs, and the four risk-associated lncRNAs were visualized by PCA. The results showed a relatively scattered distribution of these genes in high and low-risk subgroups, indicating good differentiation (Figure S7A–D, Supplemental Digital Content, http://links.lww.com/MD/O81).

3.8. Gene set enrichment analysis

In order to delve into the biological functions and pathways of the relevant lncRNAs, we carefully selected 417 genes displaying significant disparities between the high- and low-risk subgroups for comprehensive GO and KEGG pathway analysis. The GO analysis revealed a total of 498 GO terms associated with the relevant lncRNAs, with the first 30 terms elegantly depicted in Figure 3A. These encompass a wide range of functions, including the inhibition of enzymes (GO:0004857, molecular functions), the presence of collagen-containing extracellular matrix (GO:0062023, cellular components), the negative regulation of proteolysis (GO:0045861, biological processes), and the regulation of hormone levels (GO:0010817, biological processes). The meaningful lncRNAs and their corresponding GO terms are visually represented in Figure 3B. Furthermore, the related lncRNAs are implicated in 44 major functional pathways, with the 30 most significant pathways showcased in Figure 3C. Notably, the KEGG term is predominantly intertwined with several pathways, such as Neuroactive ligand-receptor interaction (hsa04080), Steroid hormone biosynthesis (hsa00140), and bile secretion (hsa04976). It is plausible that these lncRNAs may exert an influence on the progression of NSCLC through the aforementioned signaling pathways.

Figure 3.

GO/KEGG pathways of all significantly different lncRNAs in the high-risk and low-risk subgroups. Results of GO pathway analysis (A, B). Results of KEGG pathway analysis (C). GO = gene ontology, KEGG = Kyoto Encyclopedia of Genes and Genomes, lncRNA = long non-coding RNA.

3.9. Immune pathway enrichment analysis

To unravel the intricate relationship between immune function and RS, we employed ssGSEA to meticulously calculate the infiltration score of immune pathways. Remarkably, the immune-related lncRNA in the TCGA-NSCLC cohort exhibited a significant enrichment for 13 GSVA terms, which are vividly presented in Figure 4. The findings unequivocally demonstrate that the HLA and Cytolytic activity signaling pathways are intimately associated with low risk, while the MHC class I and parainflammation signaling pathways are closely linked to high risk (P < .05). Conversely, other immune-related pathways did not exhibit any discernible association with RS.

Figure 4.

Analysis of immune function pathways in high-risk and low-risk subgroups. Results of immune pathway analysis in the training group (A). Results of the analysis in the test group and entire cohort (B, C).

3.10. Analysis of TMB

We further explored the association between RS and the tumor microenvironment (TME). There were no notable variations in TMB among the risk subgroups in the training cohort (P = .83), the testing cohort (P = .12), and the entire cohort (P = .29) (Figure S8A–C, Supplemental Digital Content, http://links.lww.com/MD/O81). Survival difference analysis revealed that cases with high TMB had a higher probability of survival, although this was only statistically significant in the training group cohort (P = .033) (Figure S8D–F, Supplemental Digital Content, http://links.lww.com/MD/O81). We depicted the survival curves incorporating TMB and RS in Figure S8H and I, Supplemental Digital Content, http://links.lww.com/MD/O81. The findings revealed that patients with high TMB and low RS (blue line) had the longest OS in the training cohort and the entire cohort, while patients with low TMB and high RS (purple line) had the shortest OS. Conversely, in the testing cohort, patients with low TMB and low RS (green line) exhibited the longest OS, whereas those with high TMB and high RS (red line) had the shortest OS. In conclusion, TMB can serve as a prognostic marker, although its predictive efficacy requires further investigation through subsequent experiments.

3.11. Analysis of TIDE and immunotherapy

TIDE was employed to forecast the probability of immune evasion in NSCLC tumor samples. No statistically significant disparity in TIDE scores was observed among different risk subgroups (Fig. 5A–C). Regrettably, MSI also exhibited no significance in both high- and low-risk subgroups (Fig. 5D–F). Examination of IFNG revealed that IFNG expression was higher in the low-risk subgroup compared to the high-risk subgroup in the training cohort (P < .05) and the entire cohort (Fig. 5G–I). While MSDC exhibited a positive association with high risk in the training cohort (P < .01) and the testing cohort (P < .05), no significant distinction was observed among the various risk subgroups in the entire cohort (P < .001) (Fig. 5J–L). The findings for Merck18 (Fig. 6A–C), exclusion (Fig. 6D–F), dysfunction (Fig. 6G–I), CD274 (Fig. 6J–L), CD8 (Fig. 6M–O), TAMM2 (Figs. 6P–R), and CAF (Fig. 6S–U) are presented.

Figure 5.

TME results based on CRLncSigs. Results of difference analysis between high- and low-risk subgroups for TIDE (A–C), MSI (D–F), IFNG (G–I), and MDSC (J–L). “ *” stands for P < .05, “**” stands for P < .01, “***” means P < .001. “ns” means the difference is not significant. CRLncSig = COVID-19-related lncRNA signature, MDSC = myeloid-derived suppressor cells, MSI = microsatellite instability, TIDE = Tumor Immune Dysfunction and Exclusion.

Figure 6.

TME results based on CRLncSigs. Results of difference analysis between high- and low-risk subgroups for Merck18 (A–C), exclusion (D–F), dysfunction (G–I), CD274 (J–L), CD8 (N–O), TAMM2 (P–R), CAF (S–U). “*” stands for P < .05, “**” stands for P < .01, “***” means P < .001. “ns” means the difference is not significant. CRLncSig = COVID-19-related lncRNA signature.

3.12. The mRNAsi analysis results

The mRNAsi of tumor tissues surpassed that of normal tissues (Figure S9A, Supplemental Digital Content, http://links.lww.com/MD/O81) (P = 4.508e-30), and elevated mRNAsi levels were associated with tumor progression and metastasis. Intriguingly, no significant correlation was observed between higher mRNAsi and survival time in NSCLC patients (Figure S9B, Supplemental Digital Content, http://links.lww.com/MD/O81) (P = .182). Figure S9C–F, Supplemental Digital Content, http://links.lww.com/MD/O81 emonstrated that mRNAsi levels were higher in male patients compared to female patients and predominantly increased with cancer progression (advanced patients > early patients).

3.13. Screening of potential chemotherapeutic agents

We investigated the response of patients in the TCGA-NSCLC cohort with varying risk scores to conventional chemotherapeutic agents. Utilizing the “pRRophetic” R package, we estimated the IC50 values for 76 chemotherapeutic agents. The disparities in drug sensitivity between different risk subgroups were found to be statistically significant. The IC50 estimates for commonly employed chemotherapeutic agents in lung cancer are depicted in Figure S10A–F, Supplemental Digital Content, http://links.lww.com/MD/O81. For instance, Cisplatin (P = 5e-06), Docetaxel (P = 3.4e-11), Erlotinib (P = 4e-05), Methotrexate (P = .0029), Paclitaxel (P = 1.3e-13), Vinorelbine (P = 1.7e-05). Analysis of other drugs revealed that the low-risk subgroup exhibited higher IC50 estimates and lower sensitivity towards most chemotherapeutic agents, including Bortezomib (P = 2.6e-15), BI.2536 (P = 2.9e-12), Dasatinib (P = .0002), Elesclomol (P = 2.4e-06), Epothilone.B (P = 5.7e-06), GW843682X (P = 9.8e-12), NVP.BEZ235 (P = .0062), Thapsigargin (P = 1.9e-09), TW.37 (P = .0026), Vinblastine (P = 4.5e-05) (Figure S10G–P, Supplemental Digital Content, http://links.lww.com/MD/O81). These highly responsive chemotherapeutic agents hold promise as novel therapeutic options for NSCLC and warrant further investigation in our subsequent studies.

4. Discussion

NSCLC is a prevalent and highly fatal respiratory malignancy. Despite the existence of numerous effective treatment modalities, such as surgery, chemotherapy, radiotherapy, targeted therapy, and immunotherapy, the prognosis for patients with advanced NSCLC remains dismal.

LncRNA has been established as a biomarker for diagnosing, treating, and prognosticating cancer patients. A total of 166 potential lncRNAs have been identified in pan cancer research, which can enhance molecular characteristics, clinical variables, and possess clinical predictive value.^[29] Therefore, it is crucial to develop a prognostic risk model for NSCLC based on COVID-19-related lncRNA and apply it to immune microenvironment and immunotherapy analysis to improve prognosis and aid clinicians and researchers in determining suitable immunotherapy approaches.

Our investigation reveals that the lncRNA prognostic model associated with COVID-19 serves as an autonomous prognostic factor for patients with lung cancer. Through meticulous statistical analysis, we have identified four lncRNAs that exhibit significant associations with the prognosis of NSCLC (AL161431.1, AC079949.1, AC123595.1, and AC108136.1). Among these, AL161431.1, AC079949.1, and AC108136.1 were associated with a favorable prognosis, as their high expression levels predicted a good outcome for the patients. Conversely, AC123595.1 was linked to a higher risk, and its expression level showed a negative correlation with OS. In the context of endometrial carcinoma, AL161431.1 functions as an oncogenic factor, stimulating cellular proliferation and migration via the MAPK signaling cascade.^[30] Previous studies have showcased the pivotal role of AL161431.1 in the cuprotosis model, autophagy model, and hybrid model, all based on NSCLC samples.^[31,32] In addition to our research, we firmly believe that AL161431.1 plays an exceedingly crucial role in NSCLC, influencing the occurrence and progression of tumors from various perspectives, levels, and mechanisms. Consequently, AL161431.1 may serve as a vital biomarker and a potential therapeutic target in chemotherapy and immunotherapy. Guo et al identified AC123595.1 as a risk factor for lung adenocarcinoma (LUAD), as its expression promotes metastasis and invasion of tumor cells in LUAD.^[33] The functional roles of AC079949.1 and AC108136.1 have not been reported in the literature and require further investigation.

The constructed lncRNA risk prognosis model effectively stratifies patients into high-risk and low-risk groups, with the former exhibiting a lower overall survival rate. Univariate COX regression analysis has demonstrated significant correlations between indicators such as AJCC stage, T stage, and N stage with OS. Furthermore, the model’s exceptional predictive ability has been validated through C-index values, calibration curve validation, and ROC curve analysis, which have been further substantiated by Kaplan–Meier analysis and principal component analysis. Utilizing a Nomogram, we developed an intuitive tool to assess risk and predict OS based on the clinicopathological characteristics of individual patients. This approach aims to guide and enhance clinical decision-making. The predictive quality of the Nomogram was validated by 1-year, 3-year, and 5-year calibration curves. Hence, the “CRLncSig” risk model we have developed stands as a proficient tool for prognosticating NSCLC patients and, when combined with a Nomogram, can guide personalized treatment strategies to enhance patient prognosis.

GO/KEGG enrichment analysis revealed a close correlation between COVID-19 and the intricate interplay of neuroactive ligands and receptors (hsa04080). The research findings unequivocally confirm that indicative manifestations of COVID-19, such as muscular pain, cephalalgia, and abdominal discomfort, can be regulated through the intricate interplay of calcium signaling pathways and neuroactive ligands and receptors.^[10] Furthermore, we identified a potential link between COVID-19 and steroid metabolism, specifically in the biosynthesis of steroid hormones (hsa00140) and bile secretion (hsa04976). The expression and functionality of the steroid hormone receptor family exert a pivotal influence on multiple facets that impact the repercussions of COVID-19 within the pulmonary system. Specifically, they play a critical role in pulmonary morphogenesis, physiological function, immune reactivity, as well as the expression patterns of TMPRSS2 and ACE2.^[34] Notably, a study demonstrated that estrogen signaling exerts a protective role in individuals infected with COVID-19, potentially enhancing the immune response against the virus.^[35] Additionally, the insufficiency of steroid hormones emerged as a pivotal factor contributing to the morbidity and mortality of elderly COVID-19 patients.^[36] In conclusion, the aforementioned signal pathways are all intricately linked to the progression of COVID-19 within the body, and these pathways may emerge as potential novel targets for the treatment, alleviation of symptoms, and improvement of prognosis associated with COVID-19. However, the specific mechanisms underlying the impact of these pathways on the occurrence, development, and progression of non-small cell lung cancer tumors remain elusive, necessitating further exploration in the future.

The tumor microenvironment consists primarily of tumor cells and their surroundings, which can be categorized into non-immune microenvironments dominated by fibroblasts and immune microenvironments dominated by immune cells. GSVA outcomes demonstrated that HLA and cytolytic activity signaling pathways are low-risk pathways, implying that their heightened activation portends a favorable prognosis for NSCLC patients. Prior investigations have unequivocally demonstrated the pivotal role of HLA-DR in impeding the uncontrolled proliferation of malignant neoplasms. HLA-DR adeptly presents tumor-specific antigens, effectively recognized by CD4 + T lymphocytes, thereby instigating the production of an array of immunomodulatory cytokines, encompassing interleukins and interferon-γ, orchestrating a concerted effort to curtail the expansion of tumorous entities.^[37,38] The immunoregulatory factor TIM3 potentially governs the proliferation and invasion of pulmonary adenocarcinoma cells through intricate signaling pathways, including cellular lytic activity, while exerting profound effects on the immune microenvironment of pulmonary adenocarcinoma. It is widely acknowledged that the human immune response to tumor tissue hinges on the interaction between MHC and T-cell receptors. Diminished MHC-1 expression promotes immune evasion by tumor tissue, fostering tumor invasion and metastasis.^[39] Interestingly, our investigation into MHC class I yielded contrasting outcomes. It was revealed that diminished activation of MHC class I and parainflammation signaling pathways were indicative of a favorable prognosis.

In regard to the tumor mutational burden (TMB) analysis, no statistically significant differences were observed between high- and low-risk subgroups. However, our examination identified MDSC, exclusion, and CAF as factors associated with heightened risk. As per scientific reports, MDSCs have emerged as the primary orchestrators of immune responses and inflammation induced by tumors. Their influence is formidable, as they possess a robust suppressive capacity over anti-tumor immune responses mediated by CD4 + T cells, CD8 + T cells, and NK cells, thereby fostering the proliferation of tumors.^[40,41] The indispensable role of CAFs in orchestrating intricate immune evasion mechanisms in NSCLC has been unequivocally elucidated. This phenomenon is predominantly mediated through the robust suppression of CD4 + and CD8 + T lymphocytes, achieved by finely modulating the signaling pathways of COX2 and PD-L1.^[42,43] This suggests that targeted regulation of CAFs could potentially serve as a novel approach to cancer treatment. Our investigation has unveiled a remarkable correlation between TAMM2 and risk reduction. However, an extensive scientific inquiry has revealed the pivotal involvement of M2 TAMs in tumor development. These TAMs release a substantial number of cytokines, particularly vascular endothelial growth factor (VEGF), to induce angiogenesis, thereby facilitating tumor proliferation and dissemination.^[44,45] The mRNA Stemness Index (mRNAsi) manifests a notable discrepancy between lung carcinoma tissues and healthy tissues, emphasizing its potential in clinical categorization. Furthermore, mRNAsi showcases a strong correlation with diverse clinical attributes. Subsequently, we screened numerous conventional chemotherapeutic agents and identified 16 highly promising drugs with remarkable sensitivity. Cisplatin, Docetaxel, Erlotinib, Methotrexate, Paclitaxel, Vinorelbine, Bortezomib, BI.2536, Dasatinib, Elesclomol, and NVP.BEZ235 have been previously validated for their significant role in NSCLC chemotherapy.^[46] Further investigations are warranted to elucidate the precise mechanisms underlying the actions of Epothilone.B, GW843682X, Thapsigargin, and TW.37 in NSCLC patients.

As we embark on this groundbreaking endeavor, we aim to delve into the intricate relationship between CRlncRNA and NSCLC, constructing a prognostic model of unparalleled accuracy. It is important to acknowledge the limitations of our study. Firstly, our risk prognostic model, centered around COVID-19, was exclusively based on the TCGA-NSCLC database, without external validation from other sources. Secondly, it is plausible that we may have inadvertently overlooked certain crucial genes of relevance. Lastly, the enigmatic mechanisms underlying the role of CRLncSig in NSCLC necessitate further exploration.

In conclusion, we have established an immune prognosis model based on a signature of COVID-19-related lncRNAs, which plays a pivotal role in predicting patients’ overall survival. The corroborative GO/KEGG analysis has substantiated the association between COVID-19 infection and neuroactive ligand-receptor interaction, as well as steroid metabolism. Moreover, within the tumor immune microenvironment, HLA and cell cytolytic activity signaling pathways have been identified as protective pathways, whereas MHC class I and paramagnetization signaling pathways have been identified as high-risk pathways. Notably, TAMM2 has been linked to a reduced risk, whereas MDSC, exclusion, and CAF have been associated with an elevated risk. Lastly, we have meticulously screened chemotherapeutic agents that exhibit remarkable sensitivity in treating NSCLC. This model holds significant clinical value as it can enhance patient prognosis, evaluate lung cancer staging, and facilitate the screening of lung cancer chemotherapy drugs.

Author contributions

Conceptualization: Xiao Zhu.

Data curation: Ziyuan Huang.

Investigation: Kairong Zheng, Yidan Zhang, Yanjun Liang.

Methodology: Xiao Zhu.

Software: Zenglei Han, Xiao Zhu.

Supervision: Xiao Zhu.

Validation: Xiao Zhu.

Visualization: Kairong Zheng, Yidan Zhang, Yanjun Liang.

Writing – original draft: Ziyuan Huang.

Writing – review & editing: Ziyuan Huang, Zenglei Han.