Unveiling the impact of cell death-related genes and immune dynamics on drug resistance in lung adenocarcinoma: a risk score model and functional insights

Abstract

Lung adenocarcinoma (LUAD), characterized by its heterogeneity and complex pathogenesis, is the focus of this study which investigates the association between cell death-related genes and LUAD. Through machine learning, a risk score model was developed using the Coxboost rsf algorithm, demonstrating strong prognostic accuracy in both validation (GSE30219, GSE31210, GSE72094) and training (TCGA-LUAD) datasets with C-indices of 0.93, 0.67, 0.68, and 0.64, respectively. The study reveals that the expression of Keratin 18 (KRT18), a key cytoskeletal protein, varies across LUAD cell lines (DV-90, PC-9, A549) compared to normal bronchial epithelial cells (BEAS-2B), suggesting its potential role in LUAD's pathogenesis. Kaplan–Meier survival curves further validate the model, indicating longer survival in the low-risk group. A comprehensive analysis of gene expression, functional differences, immune infiltration, and mutations underscores significant variations between risk groups, highlighting the high-risk group's immunological dysfunction. This points to a more intricate tumor immune environment and the possibility of alternative therapeutic strategies. The study also delves into drug sensitivity, showing distinct responses between risk groups, underscoring the importance of risk stratification in treatment decisions for LUAD patients. Additionally, it explores KRT18's epigenetic regulation and its correlation with immune cell infiltration and immune regulatory molecules, suggesting KRT18's significant role in the tumor immune landscape. This research not only offers a valuable prognostic tool for LUAD but also illuminates the complex interplay between cell death-related genes, drug sensitivity, and immune infiltration, positioning KRT18 as a potential therapeutic or prognostic target to improve patient outcomes by personalizing LUAD treatment strategies.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-024-01336-y.

Background

Approximately 85% of cases of lung cancer are non-small cell lung cancer (NSCLC), which is presently the most common cause of cancer-related death worldwide [1–3]. Within non-small cell lung cancer (NSCLC), lung adenocarcinoma (LUAD) is the most common histologic subtype [4, 5]. The overall survival rate is still dishearteningly poor despite progress in treatment modalities and our understanding of the pathophysiology of NSCLC [6, 7]. Late-stage diagnosis, high recurrence rates, and the development of treatment resistance are among the chief obstacles in effectively managing this devastating disease [8, 9]. Therefore, in order to improve patient outcomes, new diagnostic and prognostic markers, as well as tailored therapy approaches, are desperately needed.

Over the past few years, the emerging role of cell death-related genes in cancer biology has attracted much attention [10, 11]. Cell death is an important process that keeps cells in equilibrium and protects them from various pathologies [12, 13]. Several disorders, including cancer, have abnormalities in cell death mechanisms linked to their pathogenesis [14, 15]. Necrosis, autophagy, apoptosis, and other types of cell death are essential for developing and propagating cancer, including NSCLC. [16, 17]. These pathways of cell death affect not just the tumor cells but also the surrounding tissue of the tumor, greatly impacting the prognosis of the illness [18]. Therefore, cell death-related genes have been increasingly considered promising therapeutic targets and prognostic markers in cancer.

However, there is still much to learn about the involvement of genes linked to cell death in LUAD. Previous studies investigating cell death-related genes have often focused on individual genes, and their collective role in LUAD remains uncharted territory [19]. It is well-established that cancer is not a consequence of a single gene mutation but rather an amalgamation of alterations in multiple genes [20, 21]. Therefore, it is imperative to have a thorough understanding of the network of genes linked to cell death and how they collectively affect the prognosis of LUAD.

High-throughput genomic technologies have made it possible to comprehensively profile gene expression and genetic changes in various diseases, including LUAD, in recent years [22, 23]. These advancements have facilitated the identification of molecular signatures associated with prognosis, treatment response, and tumor microenvironment characteristics [24, 25]. Integrating multi-omics data and applying machine learning algorithms have further enhanced our ability to develop robust prognostic models and uncover novel biomarkers with clinical relevance.

A thorough study was undertaken to analyze the functionality of genes implicated in cellular demise in LUAD. Considering these variables, a risk score model was constructed utilizing several machine learning algorithms. The primary objectives were identifying gene signatures associated with prognosis, exploring the functional implications of differentially expressed genes, assessing the immune microenvironment, and investigating drug sensitivity patterns. Our goal was to thoroughly comprehend the intricate molecular landscape of LUAD and provide insights into prospective therapeutic options by combining multiple data sets, including genomic, transcriptomic, and clinical information.

Methods

Data collection and preprocessing

Transcriptome data for LUAD and corresponding clinical information were obtained from the GEO (https://www.ncbi.nlm.nih.gov/geo) and The Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov/). The TCGA-LUAD dataset was used as the training cohort, encompassing all LUAD samples. For validation, the GSE68465 dataset, which includes 19 normal and 443 LUAD tumor samples, was employed. To ensure consistency and comparability between the transcriptome data from the GEO and TCGA datasets, batch effects were minimized using the "ComBat" function from the sva package. Samples with missing clinical data—such as gender, age, stage, grade, or prognostic details—or survival times of zero or unspecified duration were excluded from further analysis.

Construction and validation of a prognostic model

Each patient's risk score was determined using the model formula: Risk score = Σi Coefficient (mRNAi) × Expression (mRNAi). To identify the optimal cut-off point, the "surv_cutpoint" function from the "survminer" R package was employed, maximizing rank statistics through repeated simulations. This approach allowed the identification of two groups—high-risk and low-risk—with the most significant difference in survival outcomes. Kaplan–Meier (K-M) analysis was then used to compare overall survival (OS) between the two groups. The prognostic model's predictive performance and accuracy were further assessed using time-dependent receiver operating characteristic (ROC) curves. Additionally, the findings were validated using the GSE cohort.

Gene set enrichment analyses

To gain insights into the biological processes and pathways associated with the risk score, Gene Set Enrichment Analysis (GSEA) was performed using the R packages "clusterProfiler," "enrichplot," and "ggplot2." The gene sets “c5.go.v7.4.symbols.gmt,” “c2.cp.kegg.v7.4.symbols.gmt,” and “h.all.v7.4.symbols.gmt” were selected as reference sets for the analysis [26, 27]. Pathway enrichment was deemed significant when the normalized enrichment score (NES) was greater than 1, the FDR-adjusted q-value was less than 0.25, and the nominal p-value was below 0.05.

Immune infiltration and biomarkers analysis

The TIMER2.0 database (http://timer.comp-genomics.org) was used to explore the relationship between immune infiltration and the CDI signature. Immune infiltration was assessed by integrating six advanced algorithms—xCell, quanTIseq, CIBERSORT, MCP-counter, TIMER, and EPIC—through the R package "immunedeconv." Each algorithm has distinct strengths, with MCP-counter providing absolute abundance scores for ten immune and stromal cell types based on a normalized FPKM expression matrix (log2-transformed). CIBERSORT, another key algorithm, identified the relative proportions of 22 immune cell types from bulk expression data.

To further investigate differences in stromal and immune cell infiltration, the "estimate" R package and the ESTIMATE method were used to calculate stromal, immune, and ESTIMATE scores. The Tumor Immune Dysfunction and Exclusion (TIDE) algorithm was applied to predict two key mechanisms of tumor immune evasion. Using the TIDE database (http://tide.dfci.harvard.edu), LUAD patients’processed expression profiles were analyzed to calculate TIDE scores and predict treatment response.

Association between somatic mutations and risk score

Single-nucleotide polymorphism (SNV) data in MAF format were obtained from the TCGA database and processed using the "map tools" function in R. The R package "maftools" was employed to visualize the most frequently mutated genes [28]. Co-mutation analysis was conducted on the top 20 most commonly mutated genes, and their correlations were calculated to generate a corresponding plot.

Drug sensitivity analysis

The half-maximal inhibitory concentration (IC50), which represents the concentration required to inhibit 50% of drug activity, was calculated for 138 drugs using the "Prophet" R package. Drug sensitivity was compared between high- and low-risk groups, with statistical tests used to determine the differential response to low-dose treatments across these groups.

Statistical analysis

All statistical analyses were performed using R version 4.1.3. Kaplan–Meier survival curves were generated, and p-values were calculated using the log-rank test. Differences between groups were assessed using the Wilcoxon and Student’s t-tests, while Spearman correlation analysis was applied to compute correlation coefficients. A p-value < 0.05 was considered statistically significant for all tests.

Results

Establishment and validation of the risk score model

Using various machine learning techniques, we developed a risk score model based on 12 genes associated with cell death. Figure 1 displays the C-index of several risk models associated with various algorithms in the validation sets GSE30219, and the training sets TCGA-LUAD, GSE31210, and GSE72094. Figure 1A shows that the model constructed with the coxboost rsf approach performed well predictively in the sets used for training and validation, with C-index values of 0.93, 0.67, 0.68, and 0.64, respectively. The CoxBoost RSF algorithm was selected for our risk score model due to its strengths in survival analysis. CoxBoost enhances the Cox model by addressing high-dimensional data, a common feature in genomic studies, through boosting techniques that improve prediction accuracy. The RSF component adds ensemble learning, capturing complex gene interactions and enhancing model generalizability. This combination is ideal for analyzing the intricate data from 12 cell death-associated genes, offering a robust method for patient stratification based on genetic profiles, as evidenced by strong predictive performance across our datasets [29]. Kaplan–Meier curves were constructed to analyze the prognostic differences between low-risk and high-risk groups for the training and validation sets (Fig. 1B). Overall training and validation sets, the prognosis for the low-risk group was significantly better than that of the high-risk group (p < 0.05). At 1, 3, and 5 years, robust AUC values were also observed in the risk score model for the training and validation sets (Fig. 1C). This was especially clear in the training set, which showed excellent performance prediction at 1, 3, and 5 years, respectively, with AUC values of 0.955, 0.980, and 0.962. Finally, based on the box boost of the method, 19 genes were included in the risk score model's creation (Figure S1A). Functional enrichment analysis shows the main signaling pathways the genes above enrich (Figure S1B). Figure S1C displays the inter-gene correlations, among which KRT8 and KRT18 exhibit a significant positive correlation.

Fig. 1.

Performance of the 12-gene cell death-associated risk score model. A Comparison of the C-index across multiple risk score models based on various machine learning algorithms in the training set (TCGA-LUAD) and validation sets (GSE30219, GSE31210, and GSE72094). B Kaplan–Meier survival curves illustrating the significant prognostic differences between the high-risk and low-risk groups in both training and validation sets (p < 0.05). C Time-dependent ROC curves assessing the risk score model's accuracy at 1, 3, and 5 years in the training and validation sets, demonstrating strong AUC values

Functional difference analysis between high and low-risk groups

The volcano graphic in Fig. 2A shows how differentially expressed genes, or DEGs, manifest in groups of high and low risk. Genes including SLC6A4, MYOC, RS1, and AGER were significantly downregulated, whereas KRT81, PADI1, KRT16, and so on were highly upregulated. The PPI network in Fig. 2B illustrates the main functions of proteins related to DEGs. It can be observed that DEGs are highly gathered in the functions of intrinsic components of the presynaptic membrane, spliceosomal snRNP complex, and myofibril, suggesting that DEGs might play key roles in these functions (Fig. 2D). A second KEGG analysis identified DEGs as being substantially abundant in the following processes: protein digestion and absorption, thiamine metabolism, malaria, and ECM-receptor interaction (Fig. 2C). Figure 2D of the GSVA analysis demonstrated that DEGs in the high-risk group were downregulated in the pathways of BILE ACID METABOLISM and FATTY ACID METABOLISM but raised in GLYCOLYSIS, E2F TARGETS, and G2M CHECKPOINT. Signaling pathway enrichment differences between the low- and high-risk groups, as indicated by the GO analysis results (Figure S2A). In high-risk individuals, there may be a strong association between increased tumor growth and poor prognosis, as shown in (Figures S2B-C). The results suggest that the high-risk group demonstrates significantly elevated scores compared to the low-risk group regarding homologous recombination deficiency (HRD), RNA stemness, DNAss, and DMP stemness (p < 0.05) (Fig. 2E).

Fig. 2.

Differentially expressed genes (DEGs) and pathway enrichment analysis between high-risk and low-risk groups. A Volcano plot showing the DEGs between high- and low-risk groups. Genes such as SLC6A4, MYOC, RS1, and AGER were significantly downregulated, while KRT81, PADI1, and KRT16 were upregulated in the high-risk group. B Protein–protein interaction (PPI) network highlighting the key biological functions of DEGs, including components of the presynaptic membrane, spliceosomal snRNP complex, and myofibrils. C KEGG pathway enrichment analysis of DEGs revealing significant enrichment in pathways such as protein digestion and absorption, thiamine metabolism, malaria, and ECM-receptor interaction. D Gene Set Variation Analysis (GSVA) showing that pathways such as BILE ACID METABOLISM and FATTY ACID METABOLISM were downregulated, while pathways like GLYCOLYSIS, E2F TARGETS, and G2M CHECKPOINT were upregulated in the high-risk group. E Boxplots showing significantly higher homologous recombination deficiency (HRD), RNA stemness, DNAss, and DMP stemness scores in the high-risk group compared to the low-risk group (p < 0.05)

Immune infiltration analysis between high and low-risk groups

In addition, we compared immune cell infiltration between low-risk and high-risk groups to identify the cause of prognostic differences between these groups. Figure 3A displays a heatmap showing the differences in expression levels between high-risk and low-risk groups regarding immune cell infiltration, stromal cell infiltration, and MeTIL score. Contrasting the groups at high and low risk, Fig. 3B and Figure S4A show that the former had significantly higher CAF infiltration and exclusion levels (p < 0.05), while the latter had much lower dysfunction levels (p < 0.05). Similarly, Fig. 3C illustrates how the high-risk group responded to the release of cancer cell antigen by having notably larger levels of Th1 cell, CD8 T cell, macrophage, eosinophil, NK, and basophil recruitment than the low-risk group did. The TIMER, CIBERSORT, QUANTISEQ, MCP-counter, x-cell, and EPIC algorithms are used in Figures S3A and S4A to show how immune cell infiltration differs in high- and low-risk groups. A heatmap in Figure S3B shows the expression differences of immune response regulatory components across high and low-risk groups, including chemokines, chemokine receptors, MHC, immune inhibitors, and immunostimulators. Additionally, as shown in Figure S4C, we examined the relationship between immune-related processes and risk score. In light of these findings, the efficacy of immunological therapy varied between high-risk and low-risk populations. The high-risk group's TIDE score was significantly higher than the low-risk group's (Figure S4B) score, suggesting a higher chance of immunotherapy benefit for the latter. The study also showed a substantial association between the TIDE score and the effectiveness of immunotherapy. As seen in Fig. 3D, the high-risk group's immunotherapy response rate (27%) was noticeably lower than that of the low-risk group (46%).

Fig. 3.

Immune cell infiltration and immunotherapy response analysis in high-risk and low-risk groups. A Heatmap depicting differences in immune cell and stromal infiltration as well as the MeTIL score between high-risk and low-risk groups. B Boxplot comparing cancer-associated fibroblast (CAF) infiltration, exclusion levels, and dysfunction levels between the two groups, revealing significantly higher CAF infiltration and exclusion in the high-risk group and lower dysfunction levels in the low-risk group (p < 0.05). C Bar graph illustrating enhanced recruitment of immune cells such as Th1, CD8 T cells, macrophages, eosinophils, NK cells, and basophils in the high-risk group compared to the low-risk group. D Immunotherapy response rates in the high- and low-risk groups, showing a higher response rate (46%) in the low-risk group compared to the high-risk group (27%)

Mutation analysis between high and low-risk groups

The top 20 genes for both Low-risk and high-risk groups are shown in Fig. 4A, B, together with the sorts of mutations that occur in each gene. TP53, TTN, and MUC16 showed the highest incidence of mutations in the low-risk group, according to an analysis of the mutation conditions across the groups. Three genes: TP53, TTN, and CSMD3. Missense mutations made up the majority of the mutations in the high-risk group. Next, we examined exclusivity relationships and co-mutations for the top 20 genes exhibiting the highest mutation frequencies in the high- and low-risk groups (Fig. 4C, D). Results revealed that most genes displayed co-mutation interactions, including MUC16 and NPAP1, TP53 and PCLO, etc., while TP53 and STK11, TP53 and KEAP1, TP53 and KRAS, and KRAS and PCLO established an exclusivity link within the low-risk group. The high-risk group had a strong exclusive relationship between TP53 and KRAS, while most genes showed significant co-mutation relationships, such as TTN and SI, CSMD3 and CDH10, USH2A and TNR, etc. The FNBP4, PPM1B, and PHF2 mutant genes in the low-risk group showed higher mutation frequencies than those in the high-risk group, according to a comparison of the mutated genes in the two groups. This finding suggests that these genes may function as tumor suppressors, preventing the growth and spread of tumors. (Fig. 4E).

Fig. 4.

Mutation analysis of high-risk and low-risk groups. A, B Bar plots depicting the top 20 mutated genes in low-risk and high-risk groups, respectively. In the low-risk group. C, D Co-occurrence and exclusivity analysis of the top 20 genes revealed strong co-mutation interactions in both risk groups. E Comparison of the mutation frequencies for FNBP4, PPM1B, and PHF2 genes, showing higher mutation rates in the low-risk group, suggesting their potential tumor-suppressor role

Drug sensitivity analysis between high and low-risk groups

Figure 5A's box plots show how the IC50 values of several drugs differ for those at high risk compared to individuals at low risk. The findings indicated that under the backgrounds of CCT007093, Nutlin 3a, EHT 1864, MK2206, GW 441756, Erlotinib, VX702, PD 0332991, Roscovitine, AS601245, due to a lower IC50 value in the high-risk category, these drugs may have a greater impact on those at low risk. However, under the backgrounds of RO3306, BI2526, JNK inhibitor, GW 843682X, A 443654, Vinblastine, BID 1870, VX680, SL0101, etc., The IC50 value of the high-risk group was lower than that of the low-risk group, suggesting that they might be more susceptible to these drugs.

Fig. 5.

Drug Sensitivity Analysis of High-risk and Low-risk Groups. Box plot showing the IC50 values of different drugs in high-risk and low-risk groups

Expression and function analysis of KRT18 in LUAD

We selected KRT18, the gene that, while the risk score model was being created, had the highest weight coefficient for the subsequent study. In GSE19188, GSE31210, GSE68571, and TCGA-LUAD, KRT18 expression in the tumor was considerably higher (p < 0.05) than in the normal tissue, as shown in Fig. 6A. Similarly, KRT18 expression levels in tumor tissues with varying T-stages differed significantly (p < 0.05) in GSE13213 and TCGA-LUAD. (Fig. 6B). Next, the link between KRT18 and the COX regression analysis was examined and the prognosis of lung cancer further. Figure 6C's forest plot illustrates the strong correlation between high KRT18 expression and the shortened OS in GSE31210, GSE68571, GSE30219, GSE26939, and TCGA-LUAD; high KRT18 expression and shortened DFS in GSE14814, GSE30219; high KRT18 expression and shortened RFS in GSE31210; high KRT18 expression and shortened DSS and PFS in TCGA-LUAD. These findings point to KRT18 as a significant determinant of lung cancer prognosis. The GSEA functional analysis shows that KRT18 is highly enriched in pathways such as DNA unwinding involved in DNA replication, mitochondrial translation, and mitochondrial gene expression, as well as in hallmarks such as E2F targets, MYc targets, and G2m checkpoint (Figs. 6D, E). Next, we analyzed the correlation between KRT18 and mutations and copy numbers of other genes. The heatmap displayed in Fig. 6F shows that the group with high and low KRT18 expression had significantly different (p < 0.05) mutation levels of MUC16, USH2A, KRAS, and ANK. The groups with high levels of KRT18 expression are compared to those with low levels in Fig. 6F, where copy number amplification or deletions are also displayed. In order to study the effects of KRT18 on LUAD drug resistance, researchers calculated the correlation between KRT18 and IC50 values of various drugs in LUAD across different cohorts, as shown in the heatmap in Fig. 6G.

Fig. 6.

Expression and functional analysis of KRT18 in lung adenocarcinoma (LUAD). A Boxplots showing KRT18 expression levels in tumor and normal tissues across multiple datasets (GSE19188, GSE31210, GSE68571, TCGA-LUAD), demonstrating significantly higher expression in tumor tissues (p < 0.05). B KRT18 expression levels in different T-stages in GSE13213 and TCGA-LUAD, showing significant differences (p < 0.05). C Forest plot illustrating the relationship between high KRT18 expression and overall survival (OS), disease-free survival (DFS), recurrence-free survival (RFS), and other prognostic indicators in various datasets, revealing its association with poor prognosis. D, E GSEA functional analysis indicating KRT18 enrichment in pathways related to DNA replication, mitochondrial translation, and key hallmark processes such as E2F targets, MYC targets, and G2M checkpoint. F Heatmap displaying significant differences in mutation levels of MUC16, USH2A, KRAS, and ANK between high and low KRT18 expression groups. Copy number amplification and deletions are also shown. G Heatmap showing the correlation between KRT18 expression and drug sensitivity (IC50 values) in LUAD across various cohorts

KRT18 expression and functional analysis across various Cancer types

We then compared KRT18 expression levels in tumors and healthy tissue and in various cancer types. As shown in Fig. 7A, KRT18 is significantly overexpressed in GBM, UCEC, BRCA, CESC, and OV tumor tissues, indicating KRT18's potential to promote the development of these cancers. In contrast, KRT18 is significantly under-expressed in GBMLGG, LGG, WT, SKCM, and ALL tumor tissues, suggesting that KRT18 may inhibit the progression of these cancers (p < 0.05). Epigenetic modifications can generally be categorized into three main aspects of gene regulation: writers, readers, and erasers. The writer-reader-eraser system enables precise cell regulation. Using the m1A, m5C, and m6A changes as a reference, we examined the connection between KRT18 expression and the genes associated with writers, readers, and erasers (Fig. 7B).

Fig. 7.

KRT18 expression and functional analysis across multiple cancer types. A Boxplots comparing KRT18 expression between tumor and normal tissues in various cancer types, showing significant overexpression in GBM, UCEC, BRCA, CESC, and OV, and underexpression in GBMLGG, LGG, WT, SKCM, and ALL (p < 0.05). B Heatmap illustrating the correlation between KRT18 expression and epigenetic regulators (writers, readers, erasers) associated with m1A, m5C, and m6A modifications. C Heatmap showing the relationship between KRT18 expression and immune cell infiltration across various cancer types, highlighting its impact on immune recruitment in different tumors. D Heatmap analyzing the correlation between KRT18 expression and immune stimulatory and inhibitory molecules across multiple cancer types

In addition, we looked at the connection between KRT18 and the level of immune cell infiltration in various cancer types (Fig. 7C). KRT18 inhibits LUAD, PPRAD, CHOL, TGCT, CD8 T cells, CD4 T cells, DCs, and macrophages from infiltrating different areas. However, UCS, PCPG, LIHC, LGG, GBM, and KRT18 recruit B cells, macrophages, CD4 T cells, CD8 T cells, neutrophils, and DCs. We examined the relationship between immune stimulatory and inhibitory chemicals and KRT18 expression levels in various cancer types (Fig. 7D).

Discussion

We aimed to offer a comprehensive investigation of the involvement of genes linked to cell death in the prognosis of LUAD in this work. Our results demonstrate how crucial these genes are for developing LUAD and provide important information about their possible applications as prognostic and therapeutic targets. We created a risk score model with sophisticated bioinformatics tools that can reliably predict the prognosis for LUAD. These findings bring forth a promising strategy for the personalized management of LUAD patients, which can significantly improve their outcomes.

Our risk score model showed remarkable prediction accuracy in the training and validation sets, as seen by its high C-indices. These findings align with earlier research that used machine learning methods to create prognostic models for LUAD [30]. Incorporating multiple machine learning algorithms in our model construction enhances its robustness and generalizability. Treatment planning and patient risk assessment may benefit from the application of our risk score model in clinical decision-making. KRT18, a key gene in our model, is involved in epithelial-mesenchymal transition (EMT), a process integral to cancer metastasis and therapy resistance [31]. Its role in EMT highlights its potential as a therapeutic target in LUAD, suggesting that interventions disrupting KRT18-related pathways could hinder cancer progression and enhance treatment efficacy [31].

Within this study, we further examined KRT18's expression patterns and functional consequences in LUAD. We assessed KRT18 expression changes between neighboring normal tissue samples and LUAD tumor samples by analyzing large transcriptome datasets. In addition, an examination was conducted to determine the correlation between KRT18 expression and clinical outcomes among LUAD patients. Disease-free survival (DFS), recurrence-free survival (RFS), overall survival (OS), and progression-free survival (PFS) were among these results. Furthermore, functional enrichment analysis shed light on potential molecular pathways and processes related to KRT18 dysregulation in LUAD. Understanding the role of KRT18 in LUAD could provide valuable insights into its underlying molecular mechanisms and potential as a therapeutic target or prognostic biomarker. Furthermore, researching the functional effects of KRT18 dysregulation may help develop novel LUAD patient treatment methods.

In our research, the analysis of immune cell infiltration revealed pronounced differences between low-risk and high-risk LUAD groups, highlighting the complexity of the tumor immune microenvironment. This complexity encompasses a diverse array of immune cells and functional pathways that interact within the tumor ecosystem, affecting tumor progression, metastasis, and therapy response [32, 33]. The observed variations in immune infiltration, particularly the more immunosuppressive environment suggested in high-risk groups, underscore the aggressive nature of their tumors and the potential for immune evasion. These insights not only deepen our understanding of LUAD's biological underpinnings but also open up avenues for targeted immunotherapeutic interventions. By pinpointing specific immune cells and pathways prevalent in each risk group, we uncover novel opportunities to modulate the immune landscape—potentially enhancing antitumor immunity in high-risk patients and sustaining immune surveillance in low-risk cases—thereby paving the way for more tailored and effective treatment strategies in LUAD management.

Significantly, our study revealed that the high-risk group had more immunological dysfunction. This implies that the tumor's immune microenvironment significantly influences a patient's prognosis and responsiveness to treatment. Other research has demonstrated the effect of immune infiltration on the LUAD prognosis and the possibility of immunotherapy therapies, which aligns with our findings [34, 35]. Further investigation into the mechanisms underlying immune dysfunction in the high-risk group could pave the way for developing immunotherapeutic strategies tailored to LUAD patients.

In our study, mutation analysis highlighted distinct patterns between low-risk and high-risk groups, with TP53, TTN, and MUC16 being the most mutated genes in the low-risk group, suggesting their potential role in tumor suppression. The predominance of missense mutations in the high-risk group, particularly in genes like TP53, TTN, and CSMD3, underscores the functional impact of these alterations on protein structure and function, potentially driving more aggressive tumor behavior. The observed exclusivity and co-mutation relationships, especially the exclusive TP53 and KRAS mutations in the high-risk group, indicate complex gene interaction networks that could influence cancer progression and patient prognosis. Notably, genes like FNBP4, PPM1B, and PHF2 exhibited higher mutation frequencies in the low-risk group, hinting at their protective role against tumor advancement. This nuanced understanding of mutation patterns and their functional implications enriches our interpretation of the risk score model's predictive power and opens avenues for targeted therapeutic strategies in lung adenocarcinoma.

Moreover, our study demonstrated differences in drug sensitivity between the high- and low-risk groups. Previous studies have also identified associations between molecular subtypes, gene expression profiles, and drug sensitivity in LUAD [36, 37]. Drugs such as CCT007093, Nutlin 3a, and Erlotinib, exhibiting lower IC50 values in the high-risk group, suggest enhanced efficacy in targeting the aggressive tumor characteristics associated with this group. This indicates a potential for these drugs to be more effective in patients classified as high risk, who may benefit from aggressive therapeutic strategies. Conversely, the lower IC50 values for drugs like RO3306 and BI2526 in the low-risk group hint at a different sensitivity profile, suggesting these treatments could be more beneficial for patients with less aggressive disease.

Although our analysis offers insightful information about the function of genes linked to cell death in LUAD, there are certain limitations to consider. Firstly, the retrospective nature of our study design may introduce selection biases. Additional confirmation of the prognostic efficacy of our risk score model requires prospective validation studies with larger cohorts. Additionally, the functional implications of specific genes and pathways identified in our analysis require further validation by experiment.

In conclusion, our study elucidates the role of cell death-related genes in LUAD and presents a robust risk score model for predicting patient prognosis. In addition to providing prospective targets for therapeutic intervention, the discovery of differently expressed genes, functional pathways, immune infiltration patterns, and drug sensitivity profiles advances our knowledge of LUAD heterogeneity. Future research should concentrate on applying our findings to clinical settings and look into the molecular mechanisms behind immunological dysfunction and LUAD medication response.

Author contributions

DJ designed the study. YX, SH, BX, and WZ performed data analysis. YX drafted the manuscript. DJ revised the manuscript. All authors read and approved the final manuscript.

Funding

None.

Data availability

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.