1 Introduction

Lung cancer is still the leading cause of cancer death worldwide [1], and the second most common cancer worldwide, with the highest mortality rate, due to environmental exposure factors such as smoking and genetic factors [2]. Lung adenocarcinoma (LUAD) is the main form of lung cancer, and is known for its significant intratumor and intratumor heterogeneity, rapid tumour progression, high drug resistance, and the fact that patients are often diagnosed with advanced cancer, resulting in a high mortality rate [3, 4]. Although imaging examinations can effectively reduce lung cancer mortality, due to uneven economic development, the popularity of these measures is still very limited [1], Therefore, there is still an urgent need for new biological markers to help with the early screening of LUAD.

Programmed cell death (PCD) is a genetically controlled process of autonomous cell death that is vital for maintaining body homeostasis and regulating the number of cells in physiological and pathological states [5, 6]. The PCD signal is dependent on precise molecular mechanisms and can be regulated by drugs or genetic interventions to delay or accelerate cell death [7]. In cancer treatment, PCD can clear dysfunctional, infected or potentially tumourous cells, thus playing a key role in maintaining homeostasis, host defence against pathogens and cancer control in a variety of pathological scenarios [6, 8]. The study included 12 PCD [5, 9, 10] (Apoptosis, Pyroptosis, Ferroptosis, Autophagy, Necroptosis, Cuproptosis, Parthanatos, Entotic cell death, Netotic cell death, Lysosome-dependent cell death, Alkaliptosis, Oxeiptosis) discovered in recent years, even though different PCD have different biological processes and flexibility. It is well known that cancer results from the uncontrolled growth of rapidly dividing cells, and therefore, malfunctioning of the apoptosis-inhibiting mechanism is one of the major causes of the progression of most cancers [11]. In the last few decades, how to influence the PCD pathway including apoptosis has been a major goal in the clinical elimination of cancer cells [12]. However, PCD interactions in LUAD are always complex, and autophagy, ferroptosis, etc. and necroptosis often play a double-edged role in the oncogenic process of LUAD [13,14,15]. However, many PCD-related genes (PCDRGs) are shared by multiple pathways, which together form a large, coordinated, highly interconnected network of cell death [6, 16], targeted therapy against these genes can provide support for cancer treatment.

Exploring the PCDRGs and itheirregulatory network may be crucial to finding therapeutic and diagnostic strategies for LUAD, as it is known that regulating PCDRGs can inhibit tumour development [17, 18]. However, its specific role and regulatory mechanism in LUAD have not yet been fully elucidated. The analysis of the expression pattern and functional changes of PCDRGs in LUAD has become a cutting-edge research direction. This study used data from TCGA and GEO databases to construct a feature set containing multiple PCDRGs to explore their role in lung adenocarcinoma and their correlation with patient prognosis. Through these analyses, we aim to develop a prognostic assessment model (PCDI) based on PCDRGs to provide scientific evidence and accurate biomarkers for the personalised treatment of lung adenocarcinoma.

2 Methods and materials

2.1 Data acquisition and screening of differential expression genes (DEGs)

The LUAD training set used in this study was obtained from TCGA, including 502 tumour samples with complete clinical information and 59 corresponding normal tissue samples. The validation set was obtained from the GSE3141, GSE29013 and GSE31210 datasets in the GEO database. After removing samples with missing clinical information, a total of 393 samples were included in the analysis. 1254 PCDRGs from previous studies were cited [9], see Supplementary Table 1 for details. Differential analysis of the filtered (lowly expressed molecules removed) expression profiles was performed using the R packages edgeR [19], DESeq2 [20], and limma [21], and the results were normalised using logCPM. The screening conditions were foldchange > 2 and padj < 0.05. Volcano plots and Venn diagrams were created using the ggplot2 [22] and VennDiagram [23] packages to screen for differentially expressed PCDRGs in LUAD. In addition, Genotype-Tissue Expression (GTEx) data were used to compare gene expression levels in normal tissue as a baseline for identifying differentially expressed genes in lung adenocarcinoma.

2.2 Enrichment analysis

The clusterProfiler [24] was used for enrichment analysis (GO and KEGG) and Gene Set Enrichment Analysis (GSEA) [25]. The enrichment analysis results were visualised using the ggplot2 [22], igraph [26], and ggraph [27] packages. The GSVA [28] was used to convert the matrix, and the limma package [21] was used to perform differential analysis of the GSVA results pathways to compare the pathways with activity differences between different groups.

2.3 Development of PCD-related indices (PCDI)

Proportional hazard assumption testing and Cox regression analysis were performed using the survival package [29] to screen for PCDRGs that are associated with prognosis. Based on 101 algorithm combinations constructed using 10 different machine learning algorithms, the C index and risk score corresponding to different samples were calculated in different datasets, and the High risk group and Low risk group in different datasets were determined using the survival cutpoint function. In addition to this, the rms package [30] was used to construct and visualise nomogram related models. The use of calibration curves was used to assess the 1-, 3- and 5-year accuracy of the prediction models. In addition by plotting risk ratios calculated using the Cox proportional risk model. We used Cox regression analysis to assess the impact of changes in risk scores over time on patient prognosis. Finally, decision curve analysis (DCA) was used to assess the net benefit of the model at 1, 3 and 5 years. In addition, to assess the importance of genes in prognosis, we used univariate Cox regression analysis to calculate the Hazard Ratio (HR) and its 95% confidence interval for each gene. The HR value reflects the correlation between the gene expression level and the survival prognosis of patients.

2.4 Tumour immune microenvironment analysis

The immune cell scores of the samples were calculated using the TIMER, CIBERSOFT, EPIC, XCELL, MCPCOUNTER, and QUANTISEQ algorithms in the R package IOBR [31]. Heatmaps were generated using the R package ComplexHeatmap [32], and the Wilcoxon method was used to assess the differences in immune cell infiltration scores between groups. The R package Estimate [33] was used to score the tumour microenvironment. Based on the ssGSEA algorithm in GSVA [28], the immune infiltration of the corresponding matrix was calculated using the markers of 24 immune cells provided by previous studies [34], and the correlation between the corresponding feature expression and immune infiltration was analyzed using Spearman analysis, providing insights into the relationship between gene expression and immune response. The proportion of eight immune cells in the sample was calculated using the R package EPIC [35], and the differences between groups were compared to assess the variations in immune response.

2.5 Drug sensitivity analysis

The R packages oncoPredict [36] and pRRophetic [37] were used to evaluate the half-maximal inhibitory concentration (IC50) of the samples for the drugs, calculate and compare the drugs with differences in sensitivity between different groups, and use ggplot2 to visualise the results.

2.6 Analysis of spatial transcriptome data

Based on spatial transcriptomic data from the lung adenocarcinoma dataset (GSE179572-GSM5420754) [38], the R package AUCell [39] was used to quantify the expression distribution of the corresponding PCD features in different micro-regions in each partition. By analysing the gene expression patterns in different regions of tumour tissue, especially in relation to PCDRGs, the heterogeneity of the tumour microenvironment can be revealed, key microregions with high expression of PCD-related genes can be identified, and microregions that may play an important role in tumour growth, invasiveness and treatment response can be identified.

The expression of gene sets was compared across different tissue types, and Spearman correlation analysis was used to calculate the expression of gene sets and the expression of different cells. The R package linkET [40] was used for visualization.

2.7 Immunohistochemistry (IHC)

Download the stained images of key genes in IHC of LUAD tumour tissue based on the human protein atlas (HPA) to compare the staining and intensity of key molecules.

3 Results

3.1 Identify PCDRGs and its potential function in LUAD

Based on three different algorithms, a large number of genes with differential expression in LUAD tumours and normal tissues were screened (Fig. 1A–C), and 113 PCDRGs were identified from these genes (Fig. 1D, Supplementary Table 2). These 113 genes were enriched in the terms of apoptosis-related biological functions (Fig. 1E) and were associated with pathways such as ferroptosis, non-small cell lung cancer, and IL-17 signaling pathway (Fig. 1F).

Fig. 1
figure 1

Identification and enrichment analysis of differentially expressed genes (DEGs) associated with programmed cell death (PCD) in lung adenocarcinoma. The volcano plot of differential analysis in lung adenocarcinoma was calculated using A Deseq2; B edgeR; C limma. C The Venn diagram shows the PCD that was differentially expressed in the three different methods; D the heat map of the top 3 IDs in the GO enrichment analysis of Biological Process (BP), Cellular Component (CC), and Molecular Function (MF); F the network diagram shows the top 5 results of the KEGG enrichment analysis

Full size image

3.2 Calculation and model construction of PCDI

First, 24 genes associated with the prognosis of LUAD patients were screened from 113 PCD-related DEGs by univariate Cox analysis (Fig. 2A). To further develop a highly accurate signature, 10 machine learning algorithms were combined, as shown in Fig. 2B. CoxBoost had the highest average C-index of 0.613 among all the algorithms in the four datasets, and a total of 12 PCDRGs were screened. According to the Riskscore, the High risk group in the four datasets was grouped, and the K-M analysis showed that the High risk group in the four datasets had a poorer prognosis than the Low risk group (Figs. 2C–D). The ROC curves of the four datasets based on PCDI were plotted, and it was found that the AUC was mostly > 0.6 in the test set and the three validation sets, indicating that PCDI has a strong predictive ability for patient prognosis (Fig. 2G–J). In addition, we compared the PCDI with models constructed from seven previous publications in Pubmed. The results showed that the PCDI ranked first not only in the TCGA dataset (Fig. 3A) but also in the other three datasets (Fig. 3B–C), indicating that the PCDI has excellent accuracy.

Fig. 2
figure 2

Construction and validation of PCD-related index (PCDI). A Forest plot showing the PCD-related genes (PCDRGs) associated with prognosis by univariate Cox analysis; B heatmap of C-index of the prognostic model constructed by 101 machine learning algorithms; C TCGA; D GSE29013; E GSE31 210; F GSE3141 data sets to observe the difference in prognosis between High risk and Low risk groups of patients; use ROC curves to observe the predictive performance of G TCGA; H GSE29013; I GSE31210; J GSE3141 data sets

Full size image
Fig. 3
figure 3

PCDI in four datasets compared to other models. Compared with the characteristics of other studies, the C-index of PCDI in the A TCGA; B GSE29013; C GSE31210; and D GSE3141 cohorts is in a dominant position

Full size image

3.3 PCDI evaluates the prognosis of LUAD patients

Univariate Cox regression and multivariate Cox regression analysis were used to compare the different clinical characteristics and Riskscore of LUAD patients, the results collectively show that PCDI is not only a risk factor for LUAD patients (Fig. 4A), but even after being mixed with other variables, PCDI remains an independent prognostic factor for LUAD, as illustrated in Fig. 4B, demonstrating its significance in the prognosis of lung adenocarcinoma. On this basis, a survival curve was constructed to estimate the overall survival of patients at 1, 3 and 5 years (Fig. 4C). The proportional hazards assumption was met for PCDI (Fig. 4E, P > 0.05), and the models with these variables and the Riskscore had higher discriminative ability than other predictors at different time points (Fig. 4F–H).

Fig. 4
figure 4

Univariate and multivariate Cox regression analysis and nomogram to evaluate the impact of PCDI and clinicopathological characteristics on patient prognosis. A Univariate and B multivariate Cox regression analyses to compare the impact of PCDI and clinical and pathological characteristics on patient prognosis; C construction of a nomogram to predict patient prognosis; D calibration curve to assess the 1-, 3- and 5-year overall survival rate of patients; E risk ratio graph showing the changes in riskscore over time; DCA showing the benefit of PCDI for patients at F 1 year; G 3 years; H 5 years

Full size image

The results of the dichotomisation of the clinical characteristics of the patients showed that, regardless of the patient's gender, age, tumour size, lymph node metastasis or different clinical stages, the survival rate of the High risk group was lower than that of the Low risk group at the same time point (Fig. 5A–G, I–M). There was no significant difference in the survival rate between the two groups, only in patients with an M1 stage (Fig. 5H). However, it should be related to the extremely small number of M1-stage patients, and further verification by expanding the patient cohort is needed.

Fig. 5
figure 5

Clinical and pathological characteristics of different patients. A Age > 60 years and B age ≤ 60 years; C T1 + T2 stage and D T3 + T4 stage; E N0 stage and F N1 + N2 + N3 stages; G M0 stages and H M1 stages; I clinical stages I + II and J clinical stages III + IV; L the effect of female and male LUAD patients on survival rates

Full size image

3.4 The correlation between PCDI and immune cells

The PCDI correlations with immune cells, including B cells, CD4 T cells, CD8 T cells, and macrophages, are significantly different in a relatively High risk condition compared to a relatively Low risk condition (Fig. 6A). Additionally, both the ImmuneScore and ESTIMATEScore of the High risk group were significantly lower than those of the Low risk group, indicating a distinct difference in immune activity between the groups (Fig. 6B). The expression of 12 PCDRGs was closely related to 24 immune cells (Fig. 6C). Additionally, the proportion of B cells, CD4 T cells, and CD8 T cells in the High risk group was significantly lower compared to the Low risk group (Fig. 6D).

Fig. 6
figure 6

Correlation between PCDI and immune cells. A Immune cell correlation heatmap calculated by multiple algorithms in different risk groups; B Comparison of the differences in Stromal score, Immune score and Estimate score between different groups using the Estimate algorithm; C Correlation heatmap between 12 PCD expressions and 24 immune cells; D Proportional difference calculation of 8 immune cells between two groups using the EPIC algorithm

Full size image

3.5 The relationship between PCDI and immune cells

GSVA analysis showed that the G2M checkpoint, Myc targets V2, PI3K AKT Mtor signaling and other signaling pathways that regulate the cell cycle and maintain proliferation and survival signals were more active in the High risk group. These pathways may promote tumor progression and affect patient prognosis (Fig. 7A). Enrichment analysis of differentially expressed genes revealed that the High risk group compared with the Low risk group was mainly involved in biological processes associated with immune response, cytoskeleton organization, and epidermal composition (Fig. 7B). The chord plot reveals significant differentially expressed genes in key pathways including asthma, Staphylococcus aureus infection and the intestinal immune network IgA production network, such as HLA-DRB1 and HLA-DRA. The expression patterns of these genes in different risk groups show significant differences, reflecting the differences in biological mechanisms between patients with different risk levels (Fig. 7C). In addition, pathways related to the cell cycle and mitosis were enriched in the High risk group, indicating a higher level of cellular proliferation and division activities in these patients (Fig. 7D–E). Indicating that this may be a potential regulatory mechanism associated with risk classification. Conversely, pathways such as surface active substance metabolism were negatively enriched, suggesting that the High risk group was downregulated in these pathways compared to the Low risk group (Fig. 7F).

Fig. 7
figure 7

Biological function analysis of different risk groups in PCDI. A Using GSVA to quantify the differences in the activity of pathways in different risk groups; B bubble chart showing the results of GO enrichment analysis of genes with differences between the two groups; C chord diagram showing the results of KEGG enrichment analysis; D GSEA analysis of the cell cycle, E mitotic cell cycle and F surface active substance metabolism and the genes that contribute most to the pathway

Full size image

3.6 Drug sensitivity results

Our analysis using oncoPredict and pRRophetic evaluated the IC50 values for various drugs in different risk groups. The results demonstrated that patients in the High risk group showed increased sensitivity to Doramapimod_1042, BMS.754807_2171, PRIMA.1MET_1131, JQ1_2172, GSK269962A_1192, and MK.2206_1053, suggesting these drugs could be more effective for them. Conversely, the Low risk group exhibited greater sensitivity to drugs such as BI.2536, GW843682X, A.443654, and Paclitaxel. This indicates differential drug responses between the two groups, highlighting the potential for tailored therapeutic strategies based on the risk classification (Fig. 8).

Fig. 8
figure 8

Comparison of drug sensitivity between different risk groups. Box plot of the sensitivity comparison of A Doramapimod_1042; B BMS.754807_2171; C)PRIMA.1MET_1131; D JQ1_2172; E GSK269962A_1192; F MK.2206 _1053; pRRophetic was used to compare the differences in drug sensitivity between the High risk group and the Low risk group, G BI.2536; H GW843682X; I A.443654; J MK.2206; K Paclitaxel; L CCT007093

Full size image

3.7 Spatial transcriptome analysis reveals the active distribution of 12 PCDRGs in the micro-regions of lung adenocarcinoma tissue

By calculating the AUC scores for each micro-region, it was found that 12 PCDRGs gene sets were enriched in the tumour region (Fig. 9A) and were more active in the expression of malignant and mixed malignant tissues, especially in the mixed malignant region, which may reflect the biological complexity of lung adenocarcinoma brain metastases (Fig. 9B). Exploring the relationship between the AUC scores of the 12 PCDRGs genes and specific cell components in the microenvironment of lung adenocarcinoma, such as tumour cells, dendritic cells and neutrophils, showed a significant positive correlation, suggesting that these genes are more active in areas where these cells are enriched.

Fig. 9
figure 9

Spatial transcriptomics analysis shows the expression differences of 12 PCD genes in different lung adenocarcinoma tissue micro-regions. A The active landscape of the PCD gene set in the microenvironment, with the dots representing the microenvironment and the colour intensity indicating the expression activity (AUC score) of the PCD gene set in the microenvironment; B Comparison of the average AUC scores of the PCD gene set in different microenvironment types (malignant, mixed malignant, normal), with different colours indicating different tissue types. Different microenvironment types (malignant, mixed malignant, normal) in the PCD gene set, different colours represent different tissue types; C Spearman correlation between gene set expression and microenvironment components, different coloured lines and squares represent positive and negative correlations and significance, the thickness of the lines and the size of the squares represent the absolute value of the correlation coefficient and the significance of the p-value

Full size image

3.8 Expression of 12 PCDRGs in different cancers, identification of key genes for LUAD survival, and discussion of immune-relatedness

These 12 PCDRGs genes showed differential expression in BRCA, COAD, and LUAD (Fig. 10A), and were mainly risk factors in HNSC, LGG, and LUAD (Fig. 10B). In addition, these 12 genes were highly expressed in lung adenocarcinoma tumour tissues (Fig. 10C–D). The risk ratio of the 12 genes in the survival analysis of lung adenocarcinoma patients was calculated, and it was found that the HR of most genes was > 1, and these genes were associated with a poor survival prognosis in LUAD patients, especially DDIT4, BCL2L10 and CD27, which showed a higher risk ratio (Fig. 10E). The IHC results showed higher staining intensity of DDIT4 in tumor tissue as well (Fig. 10F); the survival rate of patients with high DDIT4 expression is significantly lower compared to those with low DDIT4 expression (Fig. 10G) and has significantly higher predictive efficacy (Fig. 10H). Additionally, the expression of DDIT4 is significantly positively correlated with neutrophils and CD8 T cells (Fig. 10I).

Fig. 10
figure 10

Expression of 12 PCD genes in pan-cancer analysis, and discussion of the impact of key genes on LUAD survival and immune-relatedness. A Expression of the 12 PCD genes in pan-cancer; B Heatmap showing the results of univariate Cox survival analysis of the 12 PCD genes in pan-cancer; C Expression differences of the 12 PCD genes between tumour and normal tissues in the TCGA dataset; D Expression differences of the 12 PCD genes between tumour and normal tissues in the TCGA and GTEx datasets; E Forest plot of the risk ratio of the 12 genes in lung Adenocarcinoma risk ratio forest plot of 12 genes; F IHC results showing the staining of DDIT4 in tumour and normal tissues; G KM curve showing the relationship between DDIT4 expression and LUAD patient survival; H time-dependent ROC showing the effectiveness of DDIT4 in predicting the prognosis of LUAD patients; I correlation lollipop chart showing the correlation between DDIT4 and immune infiltration

Full size image

4 Discussion

The apparent heterogeneity at the cellular and tissue levels together contributes to the complexity of LUAD in terms of diagnosis and treatment management [3, 41]. This heterogeneity is not only reflected in the biological behaviour of tumours, such as their growth rate and aggressiveness but also in the diversity within individual tumours. Exploring intratumour heterogeneity is key to understanding tumour progression [42]. Intratumoral heterogeneity often leads to treatment failure and disease recurrence, thereby impacting patient prognosis [43, 44]. In this context, the role of programmed cell death (PCD) is crucial, as it directly influences the survival and death of tumor cells [11]. Understanding the regulatory mechanisms of PCDR-dependent PCD, such as those involving the Bcl-2 gene, and their impact on cell death mechanisms in lung adenocarcinoma, as well as their correlation with tumor heterogeneity, is essential. These insights are not only key to predicting treatment response but also serve as significant factors in determining prognosis [45,46,47].

This study conducted an in-depth analysis of the expression and function of 1,254 PCDRGss in LUAD, identifying 113 differentially expressed genes between tumor and normal tissues. These genes were significantly enriched in pathways related to ferroptosis, non-small cell lung cancer, and IL-17 signaling. PCDRGs, which include SLC7A11, GCLC and GCLM, are a key component of redox homeostasis. It is overexpressed in cancer to promote tumour development and invasion and is directly related to the dependence or defence of the ferroptosis pathway in tumour cells [48,49,50]. Genes such as TRAF5 and TRAF2 are involved in the signalling of various cytokines, including IL-17 [51], Or, as in the case of JAK3, which is directly involved in IL-17 signalling [52], it can induce autoimmune reactions that promote cancer progression (31). On this basis, PCDI was successfully developed. This model shows significant prognostic predictive ability in both the training set (TCGA) and the external cohort (GSE58812, METABRIC, GSE21653). Moreover, PCDI has a higher accuracy than previous studies in different data sets. PCDI is not only a significant risk factor in LUAD, but Cox analysis also shows that PCDI can be used as a clinical prognostic decision-making tool at different time points, independent of clinical and pathological characteristics such as age and gender. Except for distant metastasis, the survival rate of patients in the High risk group was significantly lower than that of patients in the Low risk group in all clinical stages, which highlights the importance of PCDI in early risk assessment and intervention. However, the results for patients with M1 disease were not statistically significant due to the limited number of samples, and future studies need to further validate this finding in a larger patient population.

PCD triggers a specific immune response against tumours by regulating the enrichment of immune cells in the tumour microenvironment [47, 53]. PCD triggers a specific immune response against tumours by regulating the enrichment of immune cells in the tumour microenvironment [54, 55], PCD triggers a specific immune response against tumours by regulating the enrichment of immune cells in the tumour microenvironment. GSVA analysis showed that in the High risk group, the activity of G2M checkpoint, PI3K AKT and other signal pathways related to the cell cycle and proliferation was enhanced [56, 57], which suggests that High risk groups may be related to the activation of these pathways, promoting tumour progression and affecting patient prognosis. The High risk group is mainly involved in biological processes such as immune response and cytoskeleton organisation, and significant genes such as HLA-DRB1 and HLA-DRA show differences in expression in different risk groups, reflecting the heterogeneity of biological mechanisms. In addition, the pathways related to the cell cycle were enriched in the High risk group, while the metabolism of surface active substances showed a negative enrichment, revealing that these pathways may be potential mechanisms for regulating risk classification. These results provide important clues for understanding the differences in tumour biology and developing treatment strategies. Patients in the High risk group may be more sensitive to drugs such as BMS-754807, PRIMA-1MET, JQ1, GSK269962A, MK-2206, MK-2206 and CCT007093. BMS-754807 is an IGF-1R and IR inhibitor that has shown good tolerability and preliminary anti-tumour activity in multiple solid tumours in the study by Desai et al. [58], especially when blood glucose control is manageable. After this, Zhang et al. [59] also found in their research that the combination of dasatinib and BMS-754807 can enhance the anti-tumour effect by inducing autophagy and cell cycle arrest and downregulating the expression of key proteins, and has shown significant inhibitory effects on lung cancer cells in both in vitro and in vivo models.

We revealed the active distribution of 12 PCDRGs in the micro-regions of LUAD tumour tissue through spatial transcriptomics analysis, especially the high activity in malignant and mixed malignant regions. This finding may reveal that these genes are significantly positively correlated with specific cell components in the lung adenocarcinoma microenvironment, such as tumour cells, dendritic cells and neutrophils, indicating that they may play an important role in regulating the tumour microenvironment and pathological processes. Furthermore, the expression of these 12 PCDRGs is significantly different, usually appearing as a risk factor and showing high expression in lung adenocarcinoma. Among the identified genes, DDIT4 emerged as a key contributor. Its high expression is closely associated with poor survival outcomes in lung adenocarcinoma patients. DDIT4 expression levels show a significant correlation with neutrophils and CD8 + T cells within the lung adenocarcinoma microenvironment, a finding that aligns with previous research [60]. DDIT4 may influence patient prognosis by affecting tumor cell proliferation and invasion processes [61].

While this study offers a comprehensive analysis of the expression and function of PCDRGs in lung adenocarcinoma, highlighting their significant role in regulating the tumor microenvironment, there are still some limitations. Larger-scale datasets are urgently needed to validate these findings. Additionally, future research should delve deeper into the biological functions and clinical significance of PCDRGs by integrating proteomics data or clinical cohort samples. Lastly, a thorough analysis of the complex interactions between PCDRGs and their impact on the biological behavior of LUAD will be a key focus of our subsequent research.

In summary, this study provides insight into the expression of PCDRG in lung adenocarcinoma and its association with prognosis, revealing the key role of these genes in tumour biology and immune regulation. The PCDI model we constructed not only accurately predicts patient prognosis, but also provides new strategies for molecular targeted therapy of lung adenocarcinoma. This discovery opens up new possibilities for the future optimisation of treatment based on these genes, which is expected to improve patient outcomes.