Clinical data for individuals with LUAD were acquired from the TCGA database (https://portal.gdc.cancer.gov/), including RNA-seq data, patient demographics, and somatic mutation data, covering 522 tumor samples. Because different analyses required different types of available and matched data, the final sample size varied across analyses. Specifically, 507 samples with complete survival information and matched expression data were included in the survival analysis, 496 samples with available somatic mutation data were included in the mutation analysis, and 448 samples with complete subtype-related information were included in the subtype analysis. We excluded the samples missing complete prognostic data or healthy tissue specimens. Gene selection for palmitoylation was based on a recent study [23, 24, 25, 26, 27]. The flowchart and the specific model construction process are illustrated in Fig. 1 and Supplementary Fig. 1.

Fig. 1.

The flowchart of this paper. HRG, high-risk group; LRG, low-risk group; *p 0.05, **p 0.01, ***p 0.001, ****p 0.0001.

After an exhaustive review of the literature, 30 genes implicated in palmitoylation were identified, including ZDHHC1-9, ZDHHC11-24, PPT1, PPT2, ABHD17A, ABHD17B, ABHD17C, LYPLA1, and LYPLA2. To examine the link between these palmitoylation-associated genes and lncRNAs, a Pearson correlation analysis was performed. This analysis used a Pearson correlation coefficient threshold of $>$0.4 and a significance level of p 0.001 to identify lncRNAs associated with palmitoylation [28]. The ggalluvial, dplyr, and ggplot2 R packages (R version 4.2.2) were utilized to create a Sankey diagram, illustrating the interactions between palmitoylation-associated genes and lncRNAs.

Separate training and test sets of LUAD tumor samples were created. To identify more representative genes, the training cohort underwent LASSO regression analysis. Then, genes that could be prognostic were filtered using univariate Cox regression [29, 30]. Genes were categorized as potential prognostic indicators if they showed statistical significance in the Cox regression analysis (p-value 0.05). The training cohort was classified into two groups, low-risk (LRG) and high-risk (HRG), with the median risk score serving as the threshold. Subsequently, the model’s accuracy was confirmed by validating it on test samples drawn from the training set. This allowed us to detect the risk value for every sample. To further substantiate the model’s reliability, all samples from the TCGA database were used as a comprehensive dataset to verify the model’s precision.

By performing a correlation analysis between palmitoylation-correlated genes and the modeled lncRNAs, we identified connections between these lncRNAs and palmitoylation genes, aiming to clarify their relationship and deepen our understanding of the pathways involved.

We established a correlation between the modeled lncRNAs and genes associated with palmitoylation using a correlation analysis. This analysis illustrated the interactions between numerous lncRNAs and palmitoylation-associated genes, thereby improving our understanding of how these lncRNAs are interconnected with pathways relevant to palmitoylation.

To evaluate survival disparities between HRG and LRG, we conducted a receiver operating characteristic (ROC) analysis, which provided insights into the prognostic capabilities of the genetic characteristics. Additionally, to corroborate the predictive accuracy of our model, we evaluated its performance across distinct subgroups by classifying the samples and executing separate survival analyses based on gender and disease stage.

By analyzing immune-associated functions, distinct immune-related functions that differed between HRG and LRG were identified, providing a valuable reference for future research.

By integrating the risk values for each sample with the gene expression matrix, we conducted a functional enrichment analysis of the differentially expressed genes identified in HRG and LRG. Using gene ontology (GO), gene set enrichment analysis (GSEA), and Kyoto encyclopedia of genes and genomes (KEGG), we established filtering criteria based on pathways showing differential expression between the two groups. Subsequently, we selected the pathways that showed significant expression differences for further investigation [31, 32, 33].

The ESTIMATE algorithm assessed stromal and immune scores between the two patient groups, and we also examined the relationship between TMB and the two risk groups.

We analyzed variations in TMB and TIDE using the TCGA database mutation load data and correlated them with sample risk values. This helped us compare mutation loads and variations in modeled genes between HRG and LRG, revealing tumor mutation mechanisms. Additionally, using http://tide.dfci.harvard.edu, we assessed the potential for immunotherapy evasion in LUAD samples, comparing HRG and LRG [34].

To explore the relationship between the prognostic risk score developed in this study and chemotherapy drug sensitivity, the oncoPredict package was used to predict drug response. The training dataset was sourced from the publicly available Genomics of Drug Sensitivity in Cancer 2 (GDSC2) database. First, the gene expression matrix from LUAD tumor samples obtained from the TCGA database was preprocessed. Duplicate genes were averaged using the avereps function, and low-expression genes (rowMeans $>$0.5) were filtered out. To improve the prediction model’s accuracy, the empirical Bayesian (ComBat) algorithm (batchCorrect = ‘eb’) was used within the calcPhenotype function to correct for batch effects between TCGA cohort and GDSC2 cell lines. Using a ridge regression model, the oncoPredict package predicted drug sensitivity for multiple agents in the GDSC2 database for each tumor sample in the TCGA cohort. Sensitivity scores were expressed as half-maximal inhibitory concentrations (IC₅₀), where lower IC₅₀ values indicate greater sensitivity of the tumor sample to the drug. After obtaining the predicted IC₅₀ values for each drug across all samples, patients were stratified into HRG and LRG groups based on median risk scores. The Wilcoxon signed-rank test was used to compare IC₅₀ values for the same drug across these groups. To control for multiple testing-induced false positives, the Benjamini-Hochberg (BH) method was used to correct for false discovery rate (FDR) across all p-values. In this study, only drugs with FDR-corrected p-values below 0.05 were considered statistically significant and included in subsequent analyses.

The ATCC provided the Beas-2B, A549, PC9, H1299, and HCC827 cell lines. After rapid thawing in a 37 °C water bath, the frozen cell stocks were transferred to sterile 15 mL centrifuge tubes containing 10 mL of DMEM/F12 medium and centrifuged to remove the freezing medium. The cell pellets were then resuspended in fresh complete medium and seeded into culture flasks. All cell lines were validated by STR profiling and tested for mycoplasma contamination. The tubes were then maintained in a humidified incubator at 37 °C with 5% CO₂. For RNA extraction, total RNA was isolated via TRIzol reagent (Invitrogen, Carlsbad, CA, USA). This total RNA was then utilized alongside a PrimeScript RT reagent kit (Takara) to synthesize complementary DNA (cDNA). qRT-PCR was performed using a CFX-96 apparatus (Bio-Rad Laboratories, Inc., Hercules, CA, USA) with Takara SYBR Green assays. Data normalization was performed using the 2^{-Δ⁢Δ⁢Ct} method, with GAPDH as the reference gene. Table 1 displays the primer sequences employed in this investigation’s qRT-PCR.

All statistical analyses in this study were conducted usingR software version 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria) and GraphPad Prism version 9.0 (GraphPad Software, San Diego, CA, USA).. LASSO-Cox regression was used to construct prognostic models, with the median risk score serving as the cut-off to separate samples into high- and low-risk groups. Survival comparisons were performed using Kaplan-Meier curves and log-rank tests. In Cox regression analyses, all variables satisfied the proportional hazards assumption. Comparisons of clinical and molecular features between subtypes were conducted using chi-square tests. Multiple group comparisons were analyzed using analysis of variance (ANOVA), followed by Bonferroni and Dunnett multiple-comparison tests. Two-group comparisons were assessed using t-tests. A p-value 0.05 was considered statistically significant. When FDR correction was required, a corrected p-value 0.05 was used as the threshold for statistical significance.

Through the application of co-expression analysis, we determined 1662 lncRNAs related to palmitoylation. The interconnections between palmitoylation-related genes and lncRNAs were illustrated via a Sankey diagram (Fig. 2A). We categorized the 522 LUAD samples retrieved from the TCGA database into training and validating groups. Subsequently, a univariate Cox regression analysis was conducted on the training cohort to determine potential prognostic genes. Fig. 2B depicts the outcomes.

Fig. 2.

Correlation analysis outcomes. (A) The correlation analysis indicates a relationship between palmitoylation and lncRNAs, revealing connections among various modules, with distinct palmitoylation-associated genes denoted by different colors. (B) The outcomes of Cox regression analysis are shown; red dots indicate an HRG classification, and green dots indicate an LRG. (C,D) LASSO regression analysis results demonstrate that a model incorporating 13 genes offers increased accuracy and reliability. (E) A correlation analysis between lncRNAs included in the model and the palmitoylation-associated genes is presented, with positive (red) and negative (blue) correlations. HRG, high-risk group; LRG, low-risk group; LASSO, least absolute shrinkage and selection operator.

Fig. 2C,D display the results of LASSO regression analysis, which identified genes more representative of the population. By combining 13 lncRNAs related with palmitoylation (AC007384.1, AC026355.2, AL157895.1, AL355472.3, AL590666.4, AP000942.5, ATXN2-AS, CRNDE, LINC00862, LINC01128, SALRNA1, SEPSECS-AS1, TRG-AS1), a predictive model was subsequently produced. The prognostic model was used to determine the risk value for each sample. Individuals were then categorized into two groups, LRG and HRG, with the median risk score serving as the cut-off point. A correlation analysis was performed between palmitoylation-correlated genes and the lncRNAs used to build the model to detect the link between lncRNAs and palmitoylation genes. The outcomes illustrated that the strongest correlation was with 13 lncRNAs associated with palmitoylation, which belonged to the Zinc Finger DHHC-Type Containing protein family (ZDHHC). This helped us understand the relationship between various lncRNAs and palmitoylation genes. Fig. 2E displays the results. The previous studies of 13 lncRNAs are also provided (Table 2, Ref. [35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45]).

We represented each sample using risk curves, illustrating that, over time, mortality rates among individuals in the HRG exceeded those of the LRG. Fig. 3A–I depict these results. Survival analysis revealed that, across all samples and within both training and validation groups, the survival rates for the LRG exceeded those of the HRG over time (Fig. 3J–L). In conclusion, the predictive model based on risk scores effectively distinguishes survival outcomes between HRG and LRG, with patients classified as LRG generally displaying higher survival rates at all evaluated time points. This underscores the potential utility of risk scores as a predictive tool. The risk score was measured as follows:

Fig. 3.

Prognostic assessment of the risk model in different cohorts. (A–C) Risk score distribution in training, testing, and full sets. (D–F) OS and survival status. (G–I) Heatmap of 13 lncRNA expression. (J–L) Kaplan-Meier survival curves. T, tumor; N, node; M, metastasis. The risk score = LINC00862 $\times$ (0.647357039426962) + AP000942.5 $\times$ (0.432933775017398) + AC007384.1 $\times$ (–0.459024413104407) + AL157895.1 $\times$ (–0.679729904104978) + SALRNA1 $\times$ (–0.71354527981346) + AL590666.4 $\times$ (0.792768471143142) + CRNDE $\times$ (–0.295394660360208) + AC026355.2 $\times$ (–0.308431731440862) + ‘ATXN2–AS’ $\times$ (–0.745454499520386) + AL355472.3 $\times$ (0.886000515848374) + LINC01128 $\times$ (–0.721839634875733) + ‘SEPSECS–AS1’ $\times$ (–0.693383458750904) + ‘TRG–AS1’ $\times$ (–0.711741132752532). HRG, high-risk group; LRG, low-risk group.

To assess the prognostic significance of palmitoylation-related lncRNAs in individuals with LUAD, univariate and multivariate Cox regression analyses were performed. The findings demonstrated a significant correlation between the risk score derived from palmitoylation-related lncRNAs and overall survival (OS) in LUAD patients, establishing it as an independent prognostic factor (Fig. 4A,B). Moreover, the area under the ROC curve (AUC) for 1-, 3-, and 5-year survival rates was 0.743, 0.724, and 0.748, respectively (Fig. 4C). Furthermore, the predictive capacity of this model’s risk score was superior to that of other clinical parameters (Fig. 4D). Similarly, the concordance index (C-index) for the risk model surpassed that of other clinical features evaluated (Fig. 4E). These analyses show that the risk score model constructed from palmitoylation-associated lncRNAs exhibits high independence and accuracy in prognostic prediction for LUAD patients. Moreover, its predictive ability surpasses that of other clinical features, demonstrating its effectiveness as an independent prognostic factor and as a predictive tool.

Fig. 4.

Evaluation of the prognostic value of the palmitoylation-related lncRNA risk model in LUAD patients. (A,B) Univariate and multivariate analyses demonstrating the model’s independent prognostic significance. (C) AUC values for forecasting 1-, 3-, and 5-year OS. (D) Comparison of the model’s predictive accuracy with clinical and pathological factors (age, gender, and stage). (E) Concordance index curve of the risk model. AUC, Area under the curve; OS, overall survival.

To assess the 1-, 3-, and 5-year OS rates in LUAD patients, a nomogram integrating clinical features and risk scores was developed (Fig. 5A). The calibration curve illustrated that OS rates predicted by the nomogram were in strong agreement with the actual observed rates (Fig. 5B). PCA was performed to examine gene expression differences between HRG and LRG. The results illustrated significant variations between HRG and LRG in the expression profiles of total genes, palmitoylation-associated genes, palmitoylation-related lncRNAs, and the 13 palmitoylation-associated lncRNAs (Fig. 5C–F). Among these, the expression profiles of the 13 palmitoylation-associated lncRNAs showed the most significant variations between HRG and LRG, illustrating that these 13 palmitoylation-associated lncRNAs have the best prognostic capability in distinguishing HRG and LRG for LUAD (Fig. 5F). In summary, these analyses suggest that identifying and analyzing specific palmitoylation-related lncRNAs can provide more accurate risk stratification and prognostic prediction for LUAD patients.

Fig. 5.

Nomogram development and PCA analysis. (A) Nomogram predicting 1-, 3-, and 5-year OS for LUAD patients. (B) Calibration curves assessing consistency between predicted and actual OS. (C) PCA according to all genes. (D) PCA according to palmitoylation-related genes. (E) PCA based on palmitoylation-associated lncRNAs. (F) PCA based on lncRNAs included in the risk model. ***p 0.001. PCA, principal component analysis.

The clinical prognostic analysis for HRG and LRG was conducted. OS rates of HRG individuals with clinical features (age, sex, and stage) were significantly lower than those of LRG individuals (Fig. 6A–F). Fig. 6G illustrates the percentage of diverse clinical data in HRG and LRG of LUAD. Given the limited number of patients, this comparison of clinical data is for reference purposes only. Overall, using risk stratification to predict the prognosis of LUAD patients has potential clinical value, but may be limited by small sample sizes.

Fig. 6.

Clinical prognostic analysis. (A–F) Survival curves for HRG and LRG under various grouping circumstances. (G) The proportion of clinical data in HRG and LRG.

Functional enrichment analysis was performed to identify the significantly enriched GO terms in the HRG. For biological processes (BP), the enriched terms included microtubule-based movement, cilium movement, humoral immune response, antimicrobial humoral response, cilium-dependent cell motility, antibacterial humoral response, cilium or flagellum-dependent cell motility, cilium movement participating in cell motility, microtubule bundle formation, and intermediate filament organization. For cellular components (CC), the enriched terms included collagen-containing extracellular matrix, motile cilium, cytoplasmic region, plasma membrane-bounded cell projection cytoplasm, ciliary plasm, axoneme, cornified envelope, dynein complex, axonemal dynein complex, and outer dynein arm. For molecular functions (MF), the enriched terms included serine hydrolase activity, cytidine deaminase activity, serine-type peptidase activity, serine-type endopeptidase activity, heparin binding, growth factor activity, structural constituent of skin epidermis, gap junction channel activity, alkali metal ion binding, and minus-end-directed microtubule motor activity. The results are shown in Fig. 7A,B. KEGG enrichment analysis was performed to identify significantly enriched pathways in the HRG. The enriched pathways included neuroactive ligand-receptor interaction, complement and coagulation cascades, Wnt signaling pathway, pancreatic secretion, IL-17 signaling pathway, Staphylococcus aureus infection, folate biosynthesis, amoebiasis, renin-angiotensin system, graft-versus-host disease, and glycerolipid metabolism. Fig. 7C,D depict the results. Furthermore, to further examine the most significantly enriched functional terms between HRG and LRG patients, GSEA was conducted. Keratinocyte differentiation, regulation of mitotic nuclear division, intermediate filament cytoskeleton, Alzheimer’s disease, cell cycle, Huntington’s disease, cornified envelope, Parkinson’s disease, and oxidative phosphorylation were enriched in HRG. However, spliceosomal snRNP assembly, ciliary plasm, spliceosomal tri snRNP complex assembly, Cajal body, T cell receptor complex, asthma, intestinal immune network for IgA production, and systemic lupus erythematosus were enriched in LRG patients (Fig. 7E–H). These results suggest that the functional characteristics of HRG patients are primarily associated with cellular metabolism, the cell cycle, and neurodegenerative diseases, whereas LRG patients are more closely linked to immune responses and gene splicing. These mechanisms suggest that LUAD patients, especially those in HRG, may face a broader range of health issues, including an increased risk of metabolic disorders and neurodegenerative diseases. In contrast, LRG may be more stable in immune defense and in the regulation of cellular function. However, the specific risks need to be further analyzed in conjunction with the patient’s clinical characteristics and treatment process.

Fig. 7.

Immune-related and functional enrichment analysis. (A,B) GO enrichment results: outer to inner circles represent GO terms, gene counts, significance (deeper red = higher significance), co-expressed genes, and gene heat ratios. (C,D) KEGG enrichment results: bar color indicates p-value (darker = more significant), and point size reflects the number of enriched genes. (E–H) GSEA identified key pathways with p 0.05. GO, gene ontology; KEGG, Kyoto encyclopedia of genes and genomes; GSEA, gene set enrichment analysis.

LUAD samples in HRG showed a significant reduction in StromalScore, ImmuneScore, and ESTIMATEScore compared to samples in LRG (Fig. 8A). To evaluate the percentage of ICs within LUAD patient samples, we utilized methodologies (XCELL, EPIC, TIMER, QUANTISEQ, MCPCOUNTER, CIBERSORT-ABS, and CIBERSORT), utilizing marker genes and deconvolution algorithms. A Pearson correlation test was performed to examine the relationship between the risk coefficient model and tumor-infiltrating ICs, applying a screening criterion of p 0.05. Data visualization was conducted using R language software (Fig. 8B). The HRG samples were negatively related to more ICs in LUAD. Based on the subtype-specific immune model, prognostic risk genes in HRG and LRG were analyzed. The results revealed significant differences in the distribution of PRG low-expression and high-expression groups across different TCGA patient subtypes (p = 0.001) (Fig. 8C). Furthermore, a CRNDE-centered lncRNA-miRNA-mRNA regulatory network was constructed, suggesting that CRNDE may interact with multiple mRNAs through miRNA regulation, thereby influencing relevant biological processes (Fig. 8D).

Fig. 8.

Comparison of TME and ICs between risk groups. (A) Violin plots of three TME scores in HRG and LRG. (B) Correlation of IC infiltration in LUAD. (C) Distribution of PRG high and low expression groups across different subtypes. (D) CRNDE-centered lncRNA-miRNA-mRNA network diagram. **p 0.01, ***p 0.001. IC, immune cell.

Using tumor mutation load data from the TCGA database, in combination with the risk values for the samples, we employed a waterfall plot to visualize tumor mutation load. In HRG, the genes with the highest mutation frequencies were TP53 (49%), TTN (47%), and CSMD3 (41%), while in LRG, the most regularly mutated genes were TP53 (42%), TTN (40%), and MUC16 (39%). Fig. 9A,B depict these findings. This analysis indicates that tumors within HRG may harbor a greater number of genetic mutations, potentially correlating with increased tumor aggressiveness and poorer prognosis. A comparative analysis of mutation loads was conducted between HRG and LRG, and the mutation discrepancies in modeled genes across these groups were examined. The results in Fig. 9C demonstrate a higher mutation rate in the HRG than in the LRG.

Fig. 9.

The outcome of the tumor mutation load. (A,B) Waterfall plots of tumor mutations in LRG (blue) and HRG (red). (C) Box plot showing differential mutation analysis between groups. (D) Kaplan-Meier curves comparing high- vs. low-TMB groups. (E) Kaplan-Meier survival curves across four combined subgroups. (F) Violin plot of tumor immune dysfunction and exclusion in LRG (blue) and HRG (red). ***p 0.001. TMB, tumor mutational burden.

Additionally, LUAD patients were stratified into high- and low-TMB groups based on the median TMB, followed by Kaplan-Meier survival analysis. The findings indicated that individuals with LUAD who had a high TMB demonstrated improved prognostic outcomes, suggesting that TMB is a potential prognostic biomarker for this patient cohort (Fig. 9D). To further explain the synergistic effects of risk scores and TMB on OS, LUAD patients were stratified into four distinct subgroups for survival analysis. The outcomes revealed that those classified as low TMB and HRG had the most unfavorable prognosis, thereby validating our predictive model and delineating the most favorable prognostic subgroup for clinical application (Fig. 9E). In addition, Fig. 9F shows that the TIDE score is lower in the high-risk group compared to the low-risk group.

Each sample’s drug sensitivity was evaluated using database-derived data alongside our own findings. We compared HRG and LRG sensitivities to different treatment drugs by combining the risk scores for each sample. Through rigorous screening and exclusion of outcomes with no significant differences, we identified disparities in drug-sensitivity outcomes across 16 drugs between HRG and LRG (Fig. 10). HRG exhibits increased sensitivity to drugs including 5-fluorouracil, cisplatin, erlotinib, foretinib, savolitinib, trametinib, ERK_6604, AZD6738, BI-2536, and VX-11e, whereas LRG demonstrates heightened sensitivity to axitinib, doramapimod, ribociclib, sinularin, BMS-754807, and ZM447439 [46, 47, 48, 49, 50, 51]. These findings suggest a significant association between risk scores and drug sensitivity in lung adenocarcinoma patients.

Fig. 10.

Drug sensitivity analysis. (A–P) Comparison of the variations in sensitivity to chemotherapy medications (5-fluorouracil, axitinib, cisplatin, doramapimod, erlotinib, foretinib, ribociclib, savolitinib, sinularin, trametinib, BMS-754807, ERK_6604, AZD6738, BI-2536, VX-11e, ZM447439) between LRG and HRG.

Additionally, the relationship between high- and low-expression groups of AL157895.1 in the prognostic model and drug sensitivity was further evaluated. The results show that the high-expression group exhibits greater sensitivity to axitinib, cisplatin, doramapimod, ribociclib, BMS-754807, and ZM447439, while the low-expression group is more sensitive to 5-fluorouracil, erlotinib, foretinib, sorafenib, TAF1-5469, talazoparib, ERK_6604, AZD6738, BI-2536, and VX-11e (Supplementary Fig. 2).

In the previous section of this study, we identified 13 effective palmitoylation-related lncRNA genes in the model. Through further differential expression analysis in LUAD ($|$Log₂FC$|$ $>$1, adjusted p-value 0.05), we identified seven differentially expressed palmitoylation-associated lncRNAs in LUAD (including: AL157895.1, AL355472.3, SALRNA1, AP000942.5, AL590666.4, AC026355.2, LINC00862). A gene heatmap was generated in R (Fig. 11A), showing that AL157895.1 was significantly downregulated in LUAD tissues, whereas the other six genes were significantly upregulated in LUAD tissues compared to healthy tissues. Initially, AL157895.1 levels were compared between LUAD and healthy lung tissues using data from the TCGA database. The results indicated a notable downregulation of AL157895.1 in LUAD tissues (Fig. 11B). To further examine the clinical significance of AL157895.1 in LUAD, individuals were classified into high- and low-expression groups based on the median AL157895.1 expression level. The Kaplan-Meier analysis showed that LUAD patients with elevated AL157895.1 levels had significantly improved OS (Fig. 11C). To further assess the independent prognostic significance of AL157895.1, univariate and multivariate Cox regression analyses were conducted. The univariate Cox regression analysis revealed that stage and AL157895.1 expression were significantly associated with OS in individuals with LUAD (Fig. 11D). Then, multivariate Cox regression analysis demonstrated that AL157895.1 was an independent predictor of OS in LUAD patients (Fig. 11E). Patients were stratified based on age, gender, staging, and AL157895.1 expression, and their survival probabilities at 1, 3, and 5 years were predicted using a combined scoring system, with calibration curves applied for correction; Fig. 11F,G display the results. There is a paucity of studies investigating the relationship between AL157895.1 and LUAD. ROC analysis indicated that AL157895.1 possesses a significant capacity to differentiate between LUAD and normal individuals, yielding an AUC of 0.861 (Fig. 11H), suggesting its potential utility in identifying individuals with LUAD.

Fig. 11.

Evaluation of the predictive ability of the palmitoylation-associated lncRNA risk model in LUAD patients. (A) A heatmap showcasing the expression of seven palmitoylation-associated lncRNA genes in LUAD compared to normal tissues, specifically highlighting AL157895.1, AL355472.3, SALRNA1, AP000942.5, AL590666.4, AC026355.2, and LINC00862. (B) The expression of the lncRNA AL157895.1 in LUAD was analyzed using the TCGA-LUAD dataset. (C) The link between the AL157895.1 levels and OS in the TCGA database was also examined. Additionally, forest plots were generated for (D) univariate and (E) multivariate Cox regression analyses. (F) A column line graph was utilized to forecast patient survival, estimating the 1-, 3-, and 5-year survival rates through a scoring system, supplemented by the correction curve depicted in (G). (H) ROC analysis of AL157895.1 expression revealed a robust capacity to differentiate between tumor and non-tumor samples. *p 0.05, and ***p 0.001.

Seven DEGs were identified when comparing gene expression between individuals in HRG and LRG groups, allowing detection of variations in biological functions. GO analysis showed that lncRNAs related to palmitoylation were mostly enriched in BP, including pathways for platelet formation, mast cell degranulation regulation, and myeloid cell differentiation. Regarding the CC, the voltage-gated sodium channel complex, the sodium channel complex, and the glial cell projection were the primary foci of interest. The MF was primarily concerned with MAP kinase activity, C2H2 zinc finger domain binding, and the minor groove of adenine-thymine-rich DNA binding (Fig. 12A). DEGs were primarily involved in several signaling pathways, as determined by KEGG enrichment analysis. These pathways include Fc epsilon RI, sphingolipid, fluid shear stress and atherosclerosis, and MAPK (Fig. 12B). Further investigation into the functional terms most significantly enriched in HRG and LRG patients was conducted using GSEA. RNA biology and gene expression regulation were enriched in the HRG (e.g., gene silencing by RNA, Sm-like protein family complex, spliceosomal snRNP complex, RNA-binding involved in post-transcriptional genes, u6 snRNA binding) (Fig. 12C).

Fig. 12.

Functional enrichment and IC analysis of palmitoylation-related lncRNAs in LUAD. (A) GO analysis. (B) KEGG analysis showing 11 enriched pathways. (C–E) GSEA-identified representative pathways (p 0.05). (F) Correlation between risk score and ICs via CIBERSORT.

However, regulating BP, metabolism, and hormones was enriched in LRG patients (e.g., DNA geometric change, embryonic skeletal system morphogens, regulation of mitotic nuclear division, RNA 3 end processing, hormone activity, citrate cycle TCA cycle, DNA replication, maturity onset diabetes of the young, progesterone-mediated oocyte maturation, proteasome) (Fig. 12D,E). The above findings suggest that RNA biology and gene expression regulation play a significant role in HRG, possibly indicating a more active or disrupted transcriptional environment in these patients. On the other hand, the enrichment of BP related to metabolism, hormones, and cell division in LRG may reflect a more stable or controlled physiological state. This contrast could offer insights into the underlying molecular mechanisms contributing to the different risk profiles, highlighting the importance of targeted interventions based on these biological pathways for better patient management and prognosis. The content of each sample was detected. We determined the link between ICs and risk score via the risk values from each sample in our developed prognostic model (Fig. 12F). ICs that were significantly positively related include: Mast cells resting, dendritic cells activated, dendritic cells resting, T cells CD4 memory resting, macrophages M2, and monocytes. ICs that were significantly negatively correlated include: T cells CD8, T cells CD4 memory activated, plasma cells, macrophages M0, and NK cells resting. The analysis results indicate that RNA biology and gene expression regulation play a significant role in HRG, suggesting that transcriptional activity in these patients is more complex or abnormal, affecting immune response and disease progression. In contrast, LRG may be in a more stable physiological state, in which metabolism and hormone regulation could help maintain a stable disease state.

Utilizing the CIBERSORT algorithm, the link between lncRNA AL157895.1 and palmitoylation-related genes was evaluated. The scatter diagrams illustrate that there was a positive relationship between lncRNA AL157895.1 and palmitoylation-related genes (ABCA6, ANKRD44, C1QTNF7, CPA3, ERVFRD-1, FAT4, GATA1, GCSAML, HDC, HPGDS, MAPK10, MEF2C, MS4A2, RAB44, RGS13, RHEX, SCN7A, and SIGLEC6) (Supplementary Fig. 3). These analyses may suggest that the levels of these palmitoylation-associated genes show a synergistic relationship with AL157895.1 in certain BP, or that lncRNA AL157895.1 may play a vital role in regulating the expression or function of these genes. We further analyzed the expression of Palmitoylation-related lncRNA AL157895.1 in HRG and LRG of 18 palmitoylation-related genes. The palmitoylation-related lncRNA AL157895.1 is correlated with these palmitoylation-associated genes, and there are significant differences in its expression between HRG and LRG (Supplementary Fig. 4). This analysis may suggest that AL157895.1 plays a crucial role in the risk assessment of tumors or diseases.

The qRT-PCR method was utilized to examine AL157895.1, AL355472.3, SALRNA1, AL590666.4, AC026355.2, and LINC00862 expression levels in LUAD A549, PC9, H1299, and HCC827 cell lines and Beas-2B control cell line (the primer sequence of AP000942.5 is missing). The results revealed that AL355472.3, SALRNA1, AL590666.4, AC026355.2 [52, 53], and LINC00862 [41] were overexpressed in LUAD A549, PC9, H1299, and HCC827 cell lines, while AL157895.1 was suppressed in these cell lines, which was consistent with the TCGA cohort (Fig. 13) [38, 52]. These outcomes suggest that these six genes may play a vital role in tumor progression and could serve as potential biomarkers for the diagnosis, prognosis, or treatment response in LUAD.

Fig. 13.

The degree of six palmitoylation-related lncRNAs prognostic signature mRNA expression. The mRNA levels of (A) AL157895.1; (B) AL355472.3; (C) SALRNA1; (D) AL590666.4; (E) AC026355.2; (F) LINC00862 in Beas-2B, A549, PC9, H1299, and HCC827 cell lines. *p 0.05, **p 0.01, ***p 0.001 and ****p 0.0001.