Establishment and validation of an ADP-ribosylation-related gene signature for prognostic prediction in lung adenocarcinoma Authors: Xuan Luo Published: 2026 ABSTRACT AbstractBackgroundLung adenocarcinoma (LUAD) remains a leading cause of cancer mortality worldwide. Emerging evidence implicates ADP-ribosylation in cancer progression. We aimed to develop a prognostic model based on ADP-ribosylation to predict survival outcomes in LUAD.MethodsDifferentially expressed genes (DEGs) between normal and LUAD samples from TCGA were analyzed. Intersection with ADP-ribosylation-related genes (ADPRRGs) yielded DE-ADPRRGs. Univariate Cox, LASSO Cox, and multivariate Cox regression were employed to construct a risk model. A nomogram integrating clinical factors and ADP ribosylation-related risk score (ADPRRS) was developed. Predictive performance was evaluated using ROC curves. Immune infiltration, somatic mutations, and chemotherapeutic drug sensitivity were analyzed. In vitro experiments initially explored the potential functions of the key gene ARL6IP1.ResultsA risk model incorporating five signature genes (ARL6, GGA2, ARL14, PARP1, and ARL6IP1) demonstrated robust prognostic accuracy. The low-ADPRRS group exhibited greater immune cell infiltration (B cells, DCs, macrophages, and neutrophils) and a lower tumor mutational burden than the high-ADPRRS group. High ADPRRS patients exhibited potential sensitivity to cisplatin, docetaxel, and gefitinib, whereas low ADPRRS patients may respond better to parthenolide, roscovitine, and crizotinib. qRT-PCR confirmed the upregulated expression of ARL14, PARP1 and ARL6IP1 in A549 cell line. The in vitro functional experiments revealed that ARL6IP1 had significant effects on promoting proliferation, migration and invasion in LUAD cells, and it promoted tumor progression by inhibiting cell apoptosis.ConclusionThis ADP-ribosylation-based prognostic model reliably predicts LUAD survival and identifies potential biomarkers for tailored therapy.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1007/s12672-026-05353-x.Keywords: Lung adenocarcinoma, ADP-ribosylation, Prognosis, Tumor microenvironment, Drug sensitivity Introduction Lung adenocarcinoma (LUAD), the most prevalent histological subtype of lung cancer, contributes greatly to global cancer morbidity and mortality [1]. According to 2024 National Cancer Center statistics, China reports over 680,000 new lung cancer cases annually, with 478,000 deaths [2]. While early surgical resection remains standard treatment, adjuvant chemotherapy improves survivals [3]. Nonetheless, LUAD prognosis remains poor, with a 5-year survival rate below 20% [4]. Reliable biomarkers are needed to guide personalized therapy and improve outcomes. ADP-ribosylation, a post-translational modification involving the addition of ADP-ribose moieties to proteins, regulates diverse cellular processes, including DNA repair, gene expression, and cell death [5]. Poly (ADP-ribose) polymerases (PARPs) catalyze this modification, playing pivotal roles in tumorigenesis, progression, and drug resistance [6]. PARP7, a nuclear mono-ADP-ribosyltransferase, drives cancer growth by repressing IRF1/IRF3-dependent apoptosis via FRA1 ribosylation [7]. PARP12-mediated ADP-ribosylation modulates AKT activation and DNA damage response in breast cancer [8]. Ribosomal ADP-ribosylation maintains proteostasis in cancer by fine-tuning protein synthesis and preventing toxic aggregation [9]. The elevated expression of ARL14 is significantly associated with poor prognosis in non-small cell lung cancer (NSCLC) patients, potentially serving as a biomarker for disease prognosis [10]. While existing studies have primarily focused on individual PARP member, the comprehensive prognostic potential of ADP ribosylation-related genes (ADPRRGs) in LUAD remains underexplored. Here, we established and validated an ADP-ribosylation-derived prognostic model for LUAD, evaluating its utility in predicting survival, immune infiltration, mutational landscape, and therapeutic response. In addition, the expression levels of signature genes were verified by quantitative real-time PCR (qRT-PCR). This study provides a possible tool for risk stratification and informs precision therapy strategies. Materials and methods Data acquisition Transcriptomic data for LUAD were obtained from The Cancer Genome Atlas (TCGA) (https://portal.gdc.cancer.gov/), comprising 59 normal and 541 LUAD samples. GSE72094 (398 LUAD samples) and GSE26939 datasets (115 LUAD samples) were downloaded from Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/). ADPRRGs were retrieved from GeneCards (https://www.genecards.org/), retaining 90 genes with a Relevance Score > 10 (Supplementary Table S1). Prognostic model construction and validation Differentially expressed genes (DEGs) between normal and tumor samples in TCGA cohort were identified using “edgeR” package [11] with thresholds of FDR < 0.05 and |log₂FC| > 0.585. Specifically, prior to analysis initiation, the raw count data underwent preprocessing to eliminate low-expressing genes with average expression levels ≤ 1 across all samples. The data were then normalized using the TMM method to account for library size variations. For simulated biological variation, the common dispersion and tag-specific dispersion estimation methods were employed. In differential expression analysis, precision testing based on the negative binomial distribution was conducted to compare gene expression differences between tumor samples and normal controls. Significant criteria were established with a false discovery rate (FDR) threshold of 0.05 and an absolute log2-fold change threshold of 0.585 (equivalent to 1.5-fold change). Genes were classified into three categories according to the screening criteria: upregulated genes (FDR < 0.05 and log2FC > 0.585), downregulated genes (FDR < 0.05 and log2FC <-0.585), and genes with no significant differences. Subsequently, the intersection of DEGs and ADPRRGs yielded DE-ADPRRGs. Univariate Cox regression (“survival” R package [12]) was applied to screen prognosis-associated genes (p < 0.05), followed by multivariate Cox regression to build a prognostic model. ADP-ribosylation-related risk score (ADPRRS) for each patient was calculated as: Patients were stratified into high- and low-ADPRRS groups based on median ADPRRS. Kaplan-Meier survival analysis compared overall survival (OS) between groups, while time-dependent ROC curves (“timeROC” R package [13]) assessed predictive accuracy. Prognostic nomogram Independent prognostic factors were determined via univariate and multivariate Cox regression. A nomogram integrating ADPRRS and clinical factors (age, gender, TNM stage) was developed. Calibration curves evaluated the predictive performance of the nomogram at 1-, 3-, and 5-year intervals. Tumor microenvironment (TME) analysis “ESTIMATE” R package [14] computed stromal, immune, and ESTIMATE scores from TCGA-LUAD RNA-seq data. Wilcoxon tests compared scores between ADPRRS groups. Immune cell infiltration was quantified via ssGSEA and CIBERSORT. Differences in immune checkpoint expression were compared and illustrated by the box plot. TIDE scores (http://tide.dfci.harvard.edu/) were utilized to predict immunotherapy response. ESTIMATE analysis was based on gene expression data, with genes not expressed in all samples being filtered out prior to analysis. The preprocessed data underwent a filterCommonGenes step to match predefined matrix and immune feature gene sets in the ESTIMATE algorithm. Subsequently, the estimateScore function was used to calculate the ESTIMATE composite score, matrix score, and immune score of each sample by leveraging the expression levels of these signature genes. ssGSEA was employed to evaluate the pathway activity in predefined gene sets. Prior to analysis, the expression data underwent rigorous screening, retaining only protein-coding genes while eliminating low-expression ones. During computation, the getGmt function processed gene set files with the kCDF parameter set to ‘Gaussian’ to accommodate continuous data characteristics. The algorithm calculated independent enrichment scores for each sample, which are then standardized into a [0,1] range through min-max normalization. This process yielded a unified matrix of pathway activity metrics that could be directly compared across studies. CIBERSORT is a method for screening protein-coding genes from gene expression profiles. It integrates these genes with predefined risk cohort samples while excluding those not expressed across all samples. During analysis, gene expression data underwent voom conversion and subsequent quantile normalization. Using the LM22 feature matrix as a reference, this approach enabled estimation of the relative proportions of 22 human immune cell subpopulations in the sample. Mutation analysis Using “maftools" [15], somatic mutations (SNVs) were analyzed for variant types, SNV classes, and mutation rates. Tumor mutational burden (TMB) was compared between ADPRRS groups. Chemotherapeutic drug sensitivity Leveraging drug response data from Genomics of Drug Sensitivity in Cancer (GDSC, www.cancerrxgene.org/), differences in half-maximal inhibitory concentration (IC₅₀) values between groups were analyzed via “pRRophetic” R package [16]. Wilcoxon rank-sum tests were applied for statistical analysis. Cell culture The human LUAD cell line A549 and the human normal lung bronchial epithelial cell line BEAS-2B were purchased from iCell Bioscience (Shanghai, China). A549 cells were cultured in Ham’s F-12 K medium, while BEAS-2B cells were cultured in Dulbecco Modified Eagle Medium (DMEM). Both media contained 10% fetal bovine serum (10099141 C, Gibco, USA) and 1% penicillin-streptomycin (C0222, Beyotime, China). The cell culture environment was maintained at 37℃ with 5% CO2. Cell transfection To establish an A549 cell line with stable knockdown of ARL6IP1, shRNA lentiviral particles (GenePharma, China) targeting the human ARL6IP1 gene were used for transfection. The shRNA sequence used to target ARL6IP1 (corresponding to the target sequence inserted into the shRNA expression vector) is as follows: sh-ARL6IP1 Top Strand: 5’-CACCGGAGTAAATCCGAGCCACACCCTCGAGGTGTGGTGGCGTCGATTATCTC-3’; sh-ARL6IP1 Bottom Strand: 5’-AAAAGAGATAATCGACGCCACCACACCTCGAGGGTGTGGCTCGGATTTACTCC-3’. The cells were infected at an infection multiplicity (MOI = 30). The medium was replaced with fresh complete medium 8–12 h later. After 48 h of infection, 2 µg/mL puromycin (MCE, HY-K1057, USA) was added to continuously screen for 1–2 weeks to obtain a stable transfected cell pool. During this period, fresh medium containing puromycin was replaced every 2–3 days until the negative control group cells infected with the non-targeting control shRNA (shNC) completely died. The surviving cells were expanded and cultured in a maintenance medium containing 1 µg/mL puromycin for subsequent passages. The stable transfected cells were obtained for subsequent experiments. Real-time fluorescence quantitative PCR (qRT-PCR) To verify the knockdown efficiency of ARL6IP1 and to detect the expression of related genes, TRIzol reagent (Invitrogen, USA) was used to extract total RNA from human normal lung bronchial epithelial cells BEAS-2B, wild-type A549 cells, A549 cells with stable transfection of control shRNA (shNC), and A549 cells with stable transfection of ARL6IP1 shRNA (shARL6IP1). Subsequently, RNA was reverse-transcribed into cDNA using SuperScript II reverse transcriptase (Invitrogen, USA). qRT-PCR was performed with TB Green® Premix Ex Taq™ II (Takara, Japan). The reaction procedure is as follows: pre-denaturation at 95℃ for 30 s; denaturation at 95℃ for 10 s, annealing/extension at 60℃ for 30 s, a total of 40 cycles; finally, the melting curve analysis was performed (95℃ for 15 s; 60℃ for 1 min; the temperature was gradually increased at a rate of 0.3℃ per second to 95℃ and maintained for 15 s, then it was dropped to 60℃ for 15 s). Primer sequences are listed in Table 1. RNA levels were normalized using GAPDH expression, and relative expression levels were calculated using the 2−ΔΔCT method. Table 1Primer sequences for qRT-PCRGenePrimerPrimer sequenceARL14Forward5’-GGGCTGGTGTATGTTGTGGA-3’Reverse5’-TGCACATACCAGTTCCGGTC-3’PARP1Forward5’-CTCAGGGGAGGGTCTGATGA-3’Reverse5’-CTTTGACACTGTGCTTGCCC-3’ARL6IP1Forward5’-GTTTCGGTTGGAGGACTCGT-3’Reverse5’-AACCAGGCTCTTTCCCATCG-3’GAPDHForward5’-AATGGGCAGCCGTTAGGAAA-3’Reverse5’-GCGCCCAATACGACCAAATC-3’ Primer sequences for qRT-PCR Western blot (WB) To verify the knockdown of ARL6IP1 at the protein level, the cells were lysed with RIPA buffer (P0013B, Beyotime, China) to extract proteins. The protein concentration was measured using the BCA protein concentration assay kit (P0009, Beyotime, China). Equal amounts of protein (typically 20–30 µg) were separated by SDS-PAGE and transferred to a PVDF membrane. The PVDF membrane was sealed with 5% skimmed milk at room temperature for 2 h and then incubated with ARL6IP1 (1:500-1:2000, Rabbit, DF15567, AFFINITY, China) and GAPDH antibody (1:8000, Mouse, 60004-1, Proteintech, USA) at 4℃ overnight. The PVDF membrane was washed with TBST three times and then incubated with the corresponding secondary antibodies (Goat Anti-Rabbit HRP (1: 8000, 111-035-003, Jackson, USA), Goat Anti-Mouse HRP (1:8000, 115-035-003, Jackson, USA)) at room temperature for 2 h. The protein expression was visualized and analyzed using the ECL kit (34577, Thermo Fisher Scientific, USA). CCK-8 To detect the effect of ARL6IP1 knockdown on cell proliferation, shARL6IP1 and shNC A549 cells in the logarithmic growth phase were seeded in 96-well plates, with approximately 5 × 10³ cells per well. Three replicates were set up, and blank control wells without cells were also included. After the cells adhered overnight, 10 µL of CCK-8 reagent (C0038, Beyotime, China) was added to each well at the 0th, 24th, 48th, and 72nd hours of culture, and the cells were incubated in a dark cell culture box at 37℃. After incubation, the absorbance values (OD values) of each well were measured at 450 nm using a Multiskan SkyHigh microplate reader (Thermo Scientific, USA). Transwell The migration and invasion experiments were conducted using Transwell chambers (8.0 μm pore size). For the invasion experiment, the upper chamber membrane was pre-coated with diluted Matrigel matrix gel (354230, Corning, USA); for the migration experiment, the chamber was uncoated. 200 µL of DMEM medium was used to resuspend A549 cells, and 4 × 10⁴ cells (for migration) or 5 × 10⁴ cells (for invasion) were inoculated into the upper chamber. 600 µL complete medium was added to the lower chamber. After 24 h of culture at 37℃, 5% CO2, the cells were fixed with 4% PFA and stained with crystal violet solution for 20 min. Using a cotton swab, the cells that did not migrate or invade in the upper chamber were gently wiped off. Five fields were randomly selected under an inverted microscope for photography and counting of the cells that crossed the membrane. Flow cytometry Cell apoptosis was detected using the Annexin V-FITC/PI apoptosis detection kit (40302ES08, YESEN, China). Cells from each group were collected, washed twice with cold PBS, and resuspended in 1× binding buffer to a density of approximately 1 × 10⁶ cells/mL. 100 µL of the cell suspension was taken. 5 µL of Annexin V-FITC and 5 µL of PI staining solution were added and gently mixed to incubate at room temperature in the dark for 15 min. After incubation, 400 µL of 1× binding buffer was added, and a flow cytometer (BD Biosciences, USA) was used for detection. The FlowJo software was applied to analyze the proportions of early apoptotic cells (Annexin V⁺PI⁻) and late apoptotic/necrotic cells (Annexin V⁺PI⁺). Statistical analysis Data were analyzed using GraphPad Prism software. Results were presented as mean±standard deviation, with three independent experimental replicates. Two-tailed t-tests were employed to compare experimental groups. For multiple groups or multiple time points, one-way or two-way analysis of variance (ANOVA) was used for comparison, and Dunnett’s post hoc test was conducted. A p-value < 0.05 was considered statistically significant. Results Construction and validation of the prognostic model Differential expression analysis in the TCGA cohort identified 8,830 DEGs between normal and LUAD samples. Intersection with ADPRRGs yielded 26 DE-ADPRRGs (Fig. 1A, Supplementary Table S2). Univariate Cox regression identified 5 prognosis-associated genes (Supplementary Table S3), and multivariate Cox regression constructed a final prognostic model (Fig. 1B): . Kaplan-Meier analysis revealed significantly longer OS in the low-ADPRRS group versus the high-ADPRRS group in the TCGA training, GSE72094 and GSE26939 validation sets (p < 0.0001, Fig. 1C–E). ROC curves demonstrated robust predictive performance, with AUC values of 0.68 (1-year), 0.69 (3-year), and 0.67 (5-year) in the TCGA set (Fig. 1F), 0.69, 0.73, and 0.81, respectively, in GSE72094 (Fig. 1G), and 0.73, 0.66, and 0.66, respectively, in GSE26939 (Fig. 1H). Prognostic model construction and validation. A Venn diagram of DEGs and ADPRRGs. B Forest plot of multivariate Cox regression. C–E OS curves in TCGA (C), GSE72094 (D) and GSE26939 (E). F–H Time-dependent ROC curves in TCGA (F), GSE72094 (G) and GSE26939 (H) Development and validation of the nomogram Univariate and multivariate Cox regression confirmed ADPRRS as an independent prognostic factor (Fig. 2A and B). A nomogram integrating ADPRRS and clinical factors was developed to predict 1-, 3-, and 5-year OS (Fig. 2C). Calibration curves showed strong agreement between predicted and observed outcomes (Fig. 2D). Nomogram construction and validation. A, B Univariate (A) and multivariate (B) Cox regression of ADPRRS and clinical features (age, gender, TMN, and stage). C Prognostic nomogram. D Calibration curves for 1-, 3-, and 5-year OS Tumor immune microenvironment analysis To assess relationships between ADPRRS and tumor immune cell infiltration, differences in immune cell infiltration abundance between high ADPRRS and low ADPRRS groups were calculated by implementing multiple algorithms. The low-ADPRRS group exhibited significantly higher immune, ESTIMATE, and stromal scores (p < 0.01, Fig. 3A). ssGSEA revealed elevated infiltration of aDCs, B cells, DCs, macrophages, mast cells, neutrophils, T helper cells, and TILs in this group (p < 0.05, Fig. 3B). CIBERSORT showed enrichment of memory B cells, resting CD4 memory T cells, monocytes, M2 macrophages, resting dendritic cells, and resting Mast cells in low-ADPRRS patients, whereas the high-ADPRRS group had more activated CD4 memory T cells, resting NK cells, M0 macrophages, and M1 macrophages (p < 0.05, Fig. 3C). Immune checkpoint genes (CD28, ICOS, BTLA, CD27, CD40LG, CD96) were upregulated in low-ADPRRS group (p < 0.01, Fig. 3D). Lower TIDE scores were seen in this group (p < 0.01, Fig. 3E). These data suggested greater potential benefit from immunotherapy in low-ADPRRS patients. Immune microenvironment analysis. A Stromal, immune, and ESTIMATE scores. B, C Immune cell infiltration by ssGSEA (B) and CIBERSORT (C). D Immune checkpoint expression. E TIDE scores. (*p < 0.05; **p < 0.01; ***p < 0.001; ***p < 0.0001; ns not significant) Mutation analysis Both ADPRRS groups predominantly harbored missense mutations and C > A substitutions. Among the top 10 mutated genes between groups, TP53 (55% vs. 42%), TTN (52% vs. 35%), MUC16 (43% vs. 36%), and CSMD3 (44% vs. 33%) were the most frequently mutated genes (Fig. 4A and B). The high-ADPRRS group had significantly higher TMB (p < 0.0001, Fig. 4C). Mutation profiles. A, B Mutation landscapes in high- (A) and low-ADPRRS (B) groups. C TMB comparison Drug sensitivity analysis To explore chemotherapeutic agents that may be sensitive to patients with LUAD, a drug sensitivity analysis was performed. The high-ADPRRS group had increased sensitivity to 17-AAG, bryostatin-1, cisplatin, cytarabine, docetaxel, and gefitinib, whereas the low-ADPRRS group responded better to parthenolide, roscovitine, Z-LLNle-CHO, crizotinib, CMK, and TGX221 (Fig. 5). Chemotherapeutic drug sensitivity in high- vs. low-ADPRRS groups QRT-PCR verification of signature gene expression Using the TCGA dataset, we analyzed expression differences of five signature genes in both normal and LUAD samples. Results showed that ARL14, PARP1, and ARL6IP1 were significantly upregulated in LUAD compared to normal controls (p < 0.001, Fig. 6A). Furthermore, we used qRT-PCR to detect the expression of these three genes in BEAS-2B and A549 cell lines, and the results also confirmed that their expression was significantly upregulated in A549 cell lines (p < 0.05, Fig. 6B–D). Based on the above analysis, it was speculated that ARL14, PARP1 and ARL6IP1 genes may promote the development of LUAD. Analysis of expression patterns of signature genes in normal and LUAD samples. A Expression differences of five signature genes between normal and LUAD samples in the training set. B–D qRT-PCR detection of ARL14 (B), PARP1 (C), and ARL6IP1 (D) expression levels in BEAS-2B and A549 cell lines. (***p < 0.001; ***p < 0.0001) Function exploration of ARL6IP1 gene in LUAD Given the aforementioned results indicating that ARL6IP1 was significantly overexpressed in LUAD tissues and the A549 cell line (Fig. 6A, B), we further investigated the role of ARL6IP1 in the biological behavior of LUAD cells through in vitro functional experiments. Firstly, we constructed a stable knockdown model of ARL6IP1 in A549 cells using lentivirus-mediated shRNA technology. The qRT-PCR results showed that the mRNA expression level of ARL6IP1 in the shARL6IP1 group was significantly lower than that in the shNC group (p < 0.01) (Fig. 7A). The Western blot results further confirmed that the protein expression of ARL6IP1 was significantly downregulated in the knockdown group (Fig. 7B), indicating that the stable knockdown cell line of ARL6IP1 was successfully constructed. Subsequently, the results of the functional experiments showed that knockdown of ARL6IP1 significantly inhibited the proliferation, migration and invasion of A549 cells, and promoted their apoptosis (p < 0.05) (Fig. 7C–F). In summary, these in vitro experimental results indicate that ARL6IP1 has significant promoting effects on proliferation, migration and invasion in LUAD cells. It promotes tumor progression by inhibiting cell apoptosis. These findings further support its possibility as a potential oncogene and therapeutic target for LUAD. Functional exploration of the ARL6IP1 gene. A qRT-PCR was used to detect the mRNA expression level of ARL6IP1 in A549 cells after lentivirus-mediated shRNA transfection, verifying the knockdown efficiency. B Western blot was used to analyze the changes in ARL6IP1 protein expression levels in A549 cells after knockdown. C The CCK-8 assay was used to evaluate the effect of ARL6IP1 on the proliferation ability of A549 cells. D, E The Transwell assay was used to evaluate the effects of knockdown of ARL6IP1 on the migration (D) and invasion (E) abilities of A549 cells. F The effect of knockdown of ARL6IP1 on the apoptosis of A549 cells. **** indicates p < 0.0001 Discussion LUAD poses a substantial hazard to worldwide public health due to its high morbidity and mortality rates [1]. Despite tremendous advances in diagnostic tools and treatment options, the prognosis of LUAD patients exhibits substantial individual variability [17]. Therefore, more accurate biomarkers are needed to assess individualized patient risk and guide clinical decision-making. ADP-ribosylation is crucial to the onset and spread of cancers [5, 18]. Its precise roles in LUAD and connection to patient prognosis, however, are still not well understood. To uncover important prognostic genes, our work used mRNA expression data from the TCGA-LUAD dataset. We created a multi-biomarker prognosis model based on ADP-ribosylation-related signature genes, which may provide new perspectives and tools for prognostic prediction in LUAD patients. This study integrated ADP-ribosylation gene expression profiles from TCGA dataset and selected five signature genes (ARL6, GGA2, ARL14, PARP1, and ARL6IP1) to construct a prognostic model. Patients were stratified into two distinct prognostic subgroups based on median ADPRRS. Compared with conventional clinical indicators, ADPRRS demonstrated superior predictive performance, suggesting its potential as a theoretical framework for clinical decision-making. Among these signature genes, ARL6 and GGA2 functioned as protective genes (HR < 1), while ARL14, PARP1, and ARL6IP1 acted as risk genes (HR > 1). Furthermore, qRT-PCR was conducted to confirm the significantly upregulated expression of ARL14, PARP1, and ARL6IP1 genes in LUAD cell lines, with the results suggesting that these three genes may play a promoting role in the development of LUAD. ARL14 is a member of the ADP ribosylation factor (ARF) family of GTP-binding proteins of the Ras superfamily [10]. The primary function of ARF family proteins is to mediate the recruitment of cytosolic effectors to specific membrane compartments [19]. Relevant studies indicated that ARL14 emerged as a critical prognostic factor in LUAD [10]. ARL14 suppression represses cell proliferation and mitigates radiation-induced damage in LUAD cells. Silencing ARL14 blocks the extracellular signal-regulated kinase (ERK)/p38 signaling pathway and induces cell cycle arrest at G0 phase, ultimately leading to cellular dormancy [20]. ARL6, a member of the ARF (like) protein family, PARylation represents a post-translational modification involving linear or branched ADP-ribose polymer chains derived from NAD⁺. PARP1 serves as the primary enzyme for cellular PAR synthesis and the major target of PARylation during DNA damage [21]. In LUAD, PARP1 enhances metastasis by driving tumor cell invasion, anti-apoptosis, extravasation, self-renewal, and brain microenvironment remodeling [22]. ARL6IP1 is an endoplasmic reticulum protein that plays an important role in cell conduction and substance transport [23]. Zhou et al. [24] found that elevated expression of ARL6IP1 was associated with poor prognosis in breast cancer. Additionally, knockdown of ARL6IP1 significantly inhibited the growth of breast cancer cells while promoting apoptosis and oxidative stress [24]. In this study, we further investigated the biological role of ARL6IP1 in LUAD through in vitro functional experiments. The results of the cell experiments showed that the knockdown of ARL6IP1 could significantly inhibit the malignant progression of A549 cells. These results were highly consistent with previous studies in breast cancer. GGA2 interacts with the intracellular domain of epidermal growth factor receptor, stabilizing receptor expression and promoting cell growth [25]. Li et al. demonstrated that GGA2 serves as a favorable prognostic factor for NSCLC [26], a finding consistent with our research findings. Similar to ARL14, ARL6 also belongs to the ARF protein family. However, its functional roles in cancers such as LUAD require further investigation. TME analysis revealed significant infiltration of T cells, B cells, and NK cells in the low-ADPRRS subgroup. These immune populations enhance effector T cell activity and promote anti-tumor immunity, rendering low-ADPRRS LUAD more likely to exhibit a “hot tumor” phenotype [27, 28]. In contrast, the high-ADPRRS group showed marked enrichment of M0 macrophages. Tumor-associated macrophages recruited to TME can polarize into M1 or M2 subtypes in response to microenvironmental stimuli; M2 macrophages suppress effector T cell activity and facilitate immune escape, contributing to a “cold tumor” phenotype [29, 30]. This differential immune cell infiltration may partially explain the observed disparities in malignancy and prognosis between subgroups, underscoring the pivotal role of TME in LUAD progression. ADPRRS correlated with expression of multiple immune checkpoints (e.g., ICOS, BTLA, CD96). Elevated checkpoint expression can impair immune cell function and mediate tumor immune evasion [31, 32]. LUAD cells may suppress T cell responses via CD155/CD96 interactions within TME [33]. These findings highlight the potential for developing personalized immunotherapeutic strategies tailored to specific patient subgroups. Given the inherent chemoresistance of LUAD [34], we investigated potential sensitive agents for each ADPRRS subgroup. High-ADPRRS patients exhibited greater sensitivity to 17-AAG, bryostatin 1, cisplatin, cytarabine, docetaxel, and gefitinib, whereas low-ADPRRS patients responded better to parthenolide, roscovitine, Z-LLNle-CHO, crizotinib, CMK, and TGX221. 17-AAG (a HSP inhibitor) suppresses LUAD growth/invasion via LATS1/YAP pathway modulation [35]. Platinum-based chemotherapy remains the standard-of-care for LUAD, with cisplatin being the most utilized platinum agent. Cisplatin exerts its therapeutic effect by disrupting DNA structure and function [36]. Cisplatin effectively exerts anti-proliferation effects on LUAD via miR-93-mediated downregulation of cyclin D2 [37]. miR-383 increases cisplatin sensitivity by repressing RBM24/NF-κB signaling [38]. Docetaxel with sodium butyrate synergistically inhibits A549 proliferation via Gli1 suppression, while facilitating apoptosis [39]. Dihydroartemisinin potentiates gefitinib cytotoxicity through ROS-dependent apoptosis/ferroptosis [40]. Crizotinib, approved for advanced patients with ALK/c-ROS-1-rearranged LUAD [41], accelerates alleviation of lung cancer symptoms and improvement in quality of life versus standard first-line pemetrexed-plus-platinum chemotherapy [42]. Furthermore, the low ADPRRS group may benefit more from immunotherapy. Therefore, we speculate that patients in the low ADPRRS group may be more suitable for combination therapy with immunotherapy and targeted drugs (e.g., Crizotinib). We established a robust 5-gene ADP-ribosylation-related prognostic signature for LUAD risk stratification. The ADPRRS-integrated nomogram provides a clinically applicable prognostic tool. While our findings identify potential diagnostic/therapeutic biomarkers, limitations exist. First, our analysis primarily utilized publicly available GEO datasets. While these datasets provide extensive data, inherent biases and limitations may persist. The samples within these datasets might not fully represent the entire LUAD population, and variations in data quality and experimental methodologies across studies could influence results. Therefore, future research should focus on evaluating these effects within larger-scale, prospective, multi-center cohorts. Secondly, proteomic-level validation is essential to enhance clinical translatability. The precise molecular mechanisms through which ADPRRGs modulate patient outcomes and therapeutic responses remain to be fully elucidated, warranting mechanistic investigations. Moreover, this study has only utilized bioinformatics methods for predicting drug sensitivity, and future clinical trials will be required to validate its efficacy and safety in actual treatment. Finally, although we have preliminarily investigated the potential role of the key gene ARL6IP1 in the LUAD cell lines, the detailed mechanism still needs to be explored in the future through more detailed functional experiments. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1. Supplementary Table S1: 90 ADPRRGs. Supplementary Material 2. Supplementary Table S2: 26 DE-ADPRRGs. Supplementary Material 3. Supplementary Table S3: Five prognosis-related genes screened by univariate Cox regression analysis. Author contributions The author contributed to data analysis, drafting and revising the article, gave final approval of the version to be published, and agreed to be accountable for all aspects of the work. Funding Not applicable. Data availability The data and materials in the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate Ethical approval was not required as this study was based on publicly available data. Consents were not required as this study was based on publicly available data. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Supplementary Table S1: 90 ADPRRGs. Supplementary Material 2. Supplementary Table S2: 26 DE-ADPRRGs. Supplementary Material 3. Supplementary Table S3: Five prognosis-related genes screened by univariate Cox regression analysis. Data Availability Statement The data and materials in the current study are available from the corresponding author on reasonable request.