This retrospective investigation included 273 patients with LUAD from Shandong Provincial Hospital, an affiliated hospital of Shandong First Medical University, who received immunotherapy between 2020 and 2025. Post-immunotherapy, patients were monitored in the outpatient clinics every three months—data collected during follow-up encompassed health history, survival status, and other relevant details. The conclusive further investigation date was May 31, 2025. The criteria for patient inclusion were as follows: (1) diagnosed with lung adenocarcinoma and (2) underwent immunotherapy. The study’s exclusion criteria were: (1) patients who had not undergone immunotherapy, and (2) individuals who could not be monitored consistently and lacked clinical data. After applying the exclusion criteria, 58 of 273 patients were excluded from the study.

The Ethics Committee of Shandong Provincial Hospital, affiliated with Shandong First Medical University, approved this study, and informed consent was obtained from all patients (Ethical Review Number: 2023-119). All tests were conducted in accordance with authorized protocols.

Immunohistochemical staining was conducted utilizing paraffin-embedded tissue sections. Following methanol-hydrogen peroxide blocking and antigen retrieval, each sample was incubated with the primary and secondary antibodies, then stained with DAB (3,3’-Diaminobenzidine) and hematoxylin (41). Antibodies used for immunohistochemistry included IGF2BP1 (ZSGBio, bs-8683R).

The expression of IGF2BP1 in IHC samples was assessed based on the degree of cytoplasmic and nuclear staining. Semi-quantitative grading is conducted based on staining intensity and the proportion of positive cells. Two pathologists, separately and in double-anonymized fashion, assess immunohistochemically stained slides under the microscope. The ratings for unstained, weakly positive (pale yellow granules), positive (yellow granules), and strongly positive (dark granules) are scored as 0, 1, 2, and 3, correspondingly. The scoring system based on the proportion of positively staining cells relative to the total cell count is as follows: 0% corresponds to 0 points, 1% to 25% equates to 1 point, 26% to 50% yields 2 points, 51% to 75% results in 3 points, and greater than 75% awards 4 points. The ultimate expression score of IGF2BP1 for every sample is determined by the intensity of staining of the slicing multiplied by the percentage of positive cells. A final score of < 6 points indicates low expression, whereas > 6 points signifies high expression. Select 5 random 400x high-power mirrored fields for each slice, evaluate the staining intensity and proportion of positive cells in every region, and calculate the average as the scoring outcome.

Approximately 20 regulators for m6A were gathered from prior research publications (42–46). The expression matrices and associated clinical features of LUAD samples were acquired from the Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/) and The Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov/) databases. We excluded patients who lacked survival information, had survival times of less than 30 days, did not receive immunotherapy, or had incomplete clinicopathological features from further evaluation. A total of 826 individuals were included, comprising the two groups GSE31210 (N = 226) and TCGA-LUAD (N = 600). The expression matrix data of the TCGA-LUAD cohort (in FPKM format) were obtained from the Genomic Data Commons platform. Subsequently, we converted the RNA-seq data from the GEO and TCGA-LUAD datasets from fragments per kilobase million (FPKM) to transcripts per kilobase million (TPM) and to the log2(TPM + 1) scale. The “Normalized Between Arrays” functionality of the R package “Limma” (47) was utilized for data normalization.

We carefully analyzed the mRNA expression levels of twenty m6A regulatory genes: METTL3, METTL14, METTL15, WTAP, RBM15, RBM15B, KIAA1429, ZC3H13, FTO, ALKBH5, YTHDC1, YTHDC2, YTHDF1, YTHDF2, YTHDF3, IGF2BP1, IGF2BP2, IGF2BP3, HNRNPA2B1, and HNRNPC (48–51). In 744 LUAD tissues and 82 healthy lung tissue samples from the TCGA and GEO datasets. The genes with differential expression were identified using the raw p-value adjusted for false discovery rate (FDR). The Limma (47) software tool was employed to assess the differential expression of m6A-related genes between LUAD and typical tissues, identifying genes with |log2FC| > 1 and FDR < 0.05 as differentiating genes (fold change [FC]; false discovery rate [FDR]). Clustering analysis of differential m6A-associated genes was conducted utilizing the pheatmap R package (52).

The protein–protein interaction (PPI) network that includes these m6A-related genes was constructed using the STRING database (53) (https://string-db.org/). The network was integrated into Cytoscape (version 3.7.2) (54), and the cytoHubba plug-in was installed to identify all m6A-related genes using the “Degree” algorithm.

The PD-1/PD-L1 and CTLA-4 pathways in cancer facilitate tumor evasion of immunological destruction; hence, immune checkpoint drugs targeting PD-1 and CTLA-4 augment anti-tumor immunity (55). We utilized the Tumor Immune Dysfunction and Exclusion (TIDE) methodology and subclass mapping to predict clinical response to immune checkpoint inhibitors, as previously detailed (56). Given that chemotherapy is a prevalent clinical approach for treating NSCLC, we used the R package oncoPredict (57) to assess chemotherapeutic response, quantified by the half-maximum inhibitory concentration (IC50), for every LUAD patient on the General Data Science Center website (58).

Utilizing a framework of differential RNA expression associated with m6A, LUAD patients were classified into two subgroups. A comparison across subcategories was performed to evaluate the m6A mutation pattern in TCGA and GEO LUAD. The R ESTIMATE (59) software was used to quantify the proportions of immune cells, stromal elements, and tumor cells in each sample, yielding the immune system score, stromal score, and ESTIMATE score (60). The acquired data were processed using the CIBERSORT (61) method, yielding percentages for 22 immune cell types (61). The R ggpubr package was used to assess the immunological penetrance of the two immune cell categories by determining the proportions of each immune cell type within the sample.

To investigate the relationship among m6A-related RNA in order and immune checkpoints, we picked 47 pivotal immune checkpoints (IDO1, LAG3, CTLA4, TNFRSF9, ICOS, CD80, PDCD1LG2, TIGIT, CD70, TNFSF9, ICOSLG, KIR3DL1, CD86, PDCD1, LAIR1, TNFRSF8, TNFSF15, TNFRSF14, IDO2, CD276, CD40, TNFRSF4, TNFSF14, HHLA2, CD244, CD274, HAVCR2, CD27, BTLA, LGALS9, TMIGD2, CD28, CD48, TNFRSF25, CD40LG, ADORA2A, VTCN1, CD160, CD44, TNFSF18, TNFRSF18, BTNL2, C10orf54, CD200R1, TNFSF4, CD200, NRP1), which are linked to currently utilized tumor immune checkpoint inhibitors (62). The expression levels of checkpoint members were assessed in both the high- and low-expression groups using the “limma” (47) package and the Wilcoxon test.

Principal component analysis (PCA), a multivariate regression technique, was employed to validate differential infiltration of immune cells in tumor samples compared with standard control samples (63). A PCA plot was generated with ggplot2 (64) in the R programming language. The PCA graph was subsequently generated, with infiltrated immune cells treated as variables, and the differences between tumor and healthy control samples analyzed.

This study encompassed twenty genes associated with m6A modification, comprising eight methylation transferases (METTL3, METTL14, METTL15, WTAP, KIAA1429, ZC3H13, RBM15, RBM15B), ten methylation reading proteins (YTHDC1, YTHDC2, YTHDF1, YTHDF2, YTHDF3, HNRNPC, HNRNPA2B1, IGFBP1, IGFBP2, IGFBP3), and two demethylases (FTO, ALKBH5). The copy number of m6A regulators was retrieved from TCGA-LUAD using Perl, and a histogram was created visually in R. The RCircos (65) program was used to illustrate the correlation between the number of copies of 20 regulators in m6A and chromosomes, showing differences in copy number across 23 chromosomal pairs. The Wilcox test was used to assess variable expression of m6A regulators in TCGA-LUAD, using the limma program. The Limma package provides a comprehensive solution for microarray analysis and RNA-Seq differential analysis (47). Waterfall charts were generated using the maftools (66) program to illustrate the mutation frequency of regulators m6A in LUAD. The m6A regulators with elevated mutation rates were selected to categorize the collected specimens into normal and mutant groups for analysis of gene expression levels. A P-value of less than 0.05 was considered to have statistical significance, and the corresponding box plot was generated utilizing the ggpubr program.

According to the established predictive model, the LUAD samples were classified into high-expression (high risk) and low-expression (low risk) groups using screening criteria of |log2FC| > 1 and FDR < 0.05. To investigate the functional and route distinctions between the low-risk and high-risk groups, we used the “GSVA” R package (67) (version 1.38.2) to conduct Gene Set Variation Analysis (GSVA) and examine diverse biological processes across all m6A-related clusters (67). The Hallmarker gene collection, utilized as biological signatures, was sourced from the MSigDB database version 7.2 (68). Gene Set Enrichment Analysis (GSEA) was conducted using the “clusterProfiler” R package (version 3.18.1), with P-values adjusted for multiple comparisons < 0.05 deemed statistically significant (69). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were conducted using the “clusterProfiler” R package (version 3.18.1). The threshold for the GO examination was P.adjust < 0.05, while the thresholds for the KEGG analysis were P < 0.05 and P.adjust < 0.2. Subsequent functional enrichment analyses of DEGs were conducted to assess differences in biological processes (BP), cellular components (CC), and molecular functions (MF) between the low- and high-risk groupings.

We analyzed gene expression levels of 20 m6A genes associated with regulation in 744 LUAD samples and 82 healthy lung tissue samples using the Wilcoxon-Mann-Whitney test. The Kruskal-Wallis test was used to compare m6A regulatory gene expression levels across American Joint Committee on Cancer (AJCC) stages. A Spearman correlation study was conducted to examine the links among several m6A regulatory genes. A univariate Cox regression framework was employed to identify predictive m6A-regulating genes. The Least Absolute Shrinkage and Selection Operator (LASSO) Cox regression framework was utilized to develop an optimal predictive risk signature. We computed the lambda value associated with the minimum mean-squared error (lambda.min) for the five m6A-regulating genes and assessed their coefficients using 10-fold cross-validation. The risk score for each patient cohort was computed as the sum of the expression levels of each gene, each multiplied by its respective coefficient. The patients were categorized into low-risk and high-risk groups based on the median risk ratings. The predictive accuracy of the predictive risk signature and AJCC stages was evaluated using a Receiver Operating Characteristic (ROC) curve. Kaplan-Meier survival curves and the log-rank test were used to compare survival durations between low- and high-risk cohorts. A univariate Cox regression model was used to evaluate the relationships among risk score, clinicopathological features, and overall survival in the training set. A multivariate Cox regression model was used to identify variables independently associated with overall survival (P < 0.05). Subsequently, forest plots were created to enhance the visualization of the relationship between each prognostic indicator and overall survival (OS). A nomogram was developed to estimate the 1-, 3-, and 5-year survival probabilities for LUAD patients. Additionally, the performance of the prognostic model in the training sets was assessed using concordance index (c-index) values, area under the ROC curve (AUC) values, and calibration curves. All statistical analyses were conducted utilizing R statistical software version 4.4.1 (R Foundation, Vienna, Austria). A two-tailed P<0.05 was deemed statistically significant.

Figure 1 illustrates the comprehensive workflow of the current investigation. To investigate the function of m6A regulator modifications in LUAD carcinogenesis, we conducted a thorough analysis of gene expression patterns of these regulators in LUAD tumors and normal lung tissues using the TCGA database. This investigation comprised 20 m6A regulators and revealed substantial differences in protein expression levels between the normal (82) and tumor (744) sample groups. Tumor samples were differentiated from normal ones by three-dimensional principal component analysis (3D-PCA) of the 20 m6A regulators. The findings indicated that the two subgroups were entirely distinct (Figure 2a). The Wilcoxon test was used for differential expression analysis, identifying 16 of the 20 m6A-related genes as differentially expressed. The heat map (Figure 2b) indicates that 14 regulators of m6A (METTL3, KIAA1429, RBM15, YTHDF2, HNRNPC, HNRNPA2B1, IGFBP1, IGFBP3) were elevated, whereas 6 regulators (FTO, ZC3H13, METTL14, WTAP, ALKBH5, YTHDC1) were decreased in tumor samples relative to controls. Furthermore, six proteins (METTL15, RBM15B, YTHDF1, YTHDF3, IGF2BP2, YTHDC2) showed no significant differences between the two groups.

As illustrated in Figure 2c, we initially investigated the incidence and kinds of somatic mutations among the 20 m6A regulators to examine mutation features. In this research, 120 (19.51%) of the 615 individuals with mutation data exhibited alterations, and 16 (80%) of the 20 regulators for m6A were modified, with frequencies ranging from 1% to 4%. Subsequently, we examined copy number variation (CNV) in the 20 m6A regulators identified in the LUAD cohort. Analysis of copy number variation indicated that IGF2BP3, HNRNPA2B1, KIAA1429, IGF2BP1, YTHDF1, YTHDF3, METTL3, HNRNPC, and IGF2BP2 exhibited the most significant increases in copy number alterations. In contrast, FTO, METTL15, YTHDF2, RBM15, YTHDC1, YTHDC2, METTL14, ALKBH5, RBM15B, WTAP, and ZC3H13 demonstrated decreased copy number variation (Figure 2d), which largely aligned with the expression changes observed between normal and tumor sample groups. The CNV of the 20 m6A regulatory genes and their chromosomal sites were illustrated using a circos plot (Figure 2e). LUAD tissues and surrounding non-cancerous tissues can be distinguished based on CNV abnormalities in chromosomes.

Copy number variation may alter the expression levels of m6A regulators, whereas distinct epigenetic modifications in m6A regulators occur within tumors and surrounding non-cancerous tissues. The results indicated that CNV significantly influences the expression levels of m6A regulators. Numerous m6A regulators exhibited significant expression alterations in LUAD, indicating that the aberrant condition of m6A regulators contributes to the progression of LUAD.

The protein-protein interaction (PPI) analysis of networks of the 20 m6A regulatory genes indicated that RBM15, YTHDF2, YTHDF3, and KIAA1429 (VIRMA) functioned as hub genes (Figure 2f). In contrast, YTHDF3, IGF2BP2, and METTL15 were exclusively associated with MERRL14 and WTAP, lacking connections with additional m6A regulatory genes. Overall, the “writer” m6A regulatory genes exhibited a markedly greater number of connections, whereas the “reader” m6A regulatory genes displayed a dramatically decreased amount of potential interaction partners. Furthermore, all “writers” except METTL15 exhibited interactions with one another, but METTL16 showed no interactions with other writers aside from METTL14. Subsequently, we examined the correlation between writing and erasing phrases and created a scatter plot (Figures 3, 4). Correlation analysis revealed weak to moderate correlations between the eight writes and the two eraser genes. The strongest link was seen between FTO and METTL15 (r = 0.518), while the association between ALKBH5 and RBM15 was the most negative (r = 0.005). Moreover, FTO was the only eraser gene to exhibit a substantial correlation with all eight writer genes in the PPI network.

We deduced that m6A methylation correlates with an unfavorable prognosis in LUAD, based on differences in overall survival across subtypes. To elucidate the prognostic significance of m6A methylation genes in LUAD patients, we subsequently examined the predictive capabilities of the 16 differentially expressed m6A methylation genes previously identified (METTL3, METTL15, KIAA1429, RBM15, RBM15B, YTHDC1, YTHDC2, YTHDF1, YTHDF2, YTHDF3, HNRNPC, HNRNPA2B1, IGFBP1, IGFBP2, IGFBP3, FTO) in this patient cohort. Univariate Cox regression analysis of the differentially expressed genes indicated that four out of a total of sixteen m6A regulating genes—WTAP (adjusted HR = 1.02, 95% CI = 1.01–1.04, p=0.003), IGF2BP1 (adjusted HR = 1.01, 95% CI = 1.01–1.02, p=0.0001), IGF2BP3 (adjusted HR = 1.02, 95% CI = 1.01–1.03, p=0.01), and KIAA1429 (adjusted HR = 1.02, 95% CI = 1.00–1.04, p=0.01)—were significantly associated with overall survival (P < 0.05) (Figure 5a). The comprehensive survival analysis indicated that the expression levels of four genes were strongly correlated with an unfavorable prognosis in LUAD. Lasso-Cox regression was employed to precisely identify the most significant prognostic m6A-related genes as predictors of overall survival using LASSO-penalized regression. The dashed perpendicular line represents the primary value of log l with the minimal segment likelihood bias. When Lambda is designated as Lambda-min, the partial likelihood deviation is minimized, and the variable count is one (Figures 5b, c). The variable IGF2BP1 has a coefficient of 0.00699, indicating that it exerts the most significant influence on the prognosis of individuals with LUAD among the m6A-related genes. The results validated the predictive significance of modifications to m6A regulators in LUAD.

Consequently, our findings indicated that elevated IGF2BP1 expression constituted an independent prognostic risk factor for worse outcomes in LUAD. Based on the LASSO regression results, we categorized individuals with LUAD into high-expression (high risk) and low-expression (low risk) groups using the median IGF2BP1 gene expression (counts) as the cutoff.

We analyzed high- and low-expression LUAD patient cohorts in the TCGA dataset and identified 5083 differentially expressed genes (DEGs; 3019 upregulated and 2064 downregulated). Figure 5d displays a graphical representation of 50 upregulated and downregulated genes. KEGG and GO enrichment analyses were conducted on 5083 DEGs to investigate the functional and pathway differences between high- and low-expression groups.

Gene Ontology enrichment analyses revealed that the ten most upregulated genes in patients with high expression were significantly associated with biological processes about proliferation and differentiation, including chromosome segregation, organelle fission, nuclear division, and nuclear chromosome segregation (Figure 5e), thereby affirming the involvement of IGF2BP1 high expression patterns in RNA modification and immune system control mechanisms. The KEGG results (Figure 5f) indicated that IGF2BP1 is closely associated with pathways such as the cell cycle, progesterone-mediated oocyte maturation, and the p53 signaling pathway, suggesting that elevated gene expression may be linked to tumor progression.

To further examine the correlation between high- and low-expression groups in LUAD prognosis, we used GSEA and discovered 27 differential pathways. Fourteen pathways were linked to tumorigenesis and progression, comprising eight related to biological functions and six related to metabolism. The high-expression group showed positive correlations with eight KEGG pathways, including cell cycle, DNA replication, pyrimidine metabolism, pentose and glucuronate interconversions, arginine and proline metabolism, ascorbate and aldarate metabolism, steroid hormone biosynthesis, and drug metabolism involving other enzymes. The remaining six KEGG pathways exhibited negative correlations, comprising hematopoietic cell lineage, natural killer cell-mediated cytotoxicity, chemokine signaling pathway, cell adhesion molecules (CAMs), endocytosis, and cytokine-cytokine receptor interaction (Figure 6).

The aforementioned findings indicate that high IGF2BP1 expression correlates with the onset, progression, metastasis, and resistance to treatment in lung cancer. Based on the preceding analysis, we postulated the existence of distinct immunological subtypes characterized by varying immune processes and applications, hence validating the credibility of our research.

The tumor microenvironment is now recognized as a significant regulator of cancer development and treatment response (70). To assess the correlation between our risk model and the immunological microenvironment and to provide insights into immunotherapy response, we examined the association between risk score and immune cell infiltration. Consequently, we examined the correlation between IGF2BP1 expression and immune infiltration levels. The immunological distributions of 22 leukocyte subtypes in every LUAD sample were determined using the CIBERSORT algorithm. Figure 7a illustrates the significance of 22 immune cell types across all LUAD patients. Figure 7b illustrates the disparities in infiltration of immune cells between the high-expression and low-expression groups using a heat map. The proportion of 22 immune cell types between the high- and low-expression groups was examined using the Wilcoxon test (Figure 7c). A total of 12 immune cell types exhibit varying levels of invasion. The high-expression group showed significantly elevated infiltration of cells from the plasma membrane, activated CD4 memory T cells, follicular helper T cells, M0 macrophages, and M1 macrophages. In contrast, the low-expression group showed a stronger positive correlation with memory B cells, resting CD4 memory T cells, monocytes, M2 macrophages, resting dendritic cells, activated dendritic cells, and resting mast cells. These results illustrate varying levels of immune infiltration across several subgroups.

Immune checkpoints are critical regulatory molecules that uphold self-tolerance, avert autoimmune reactions, and mitigate tissue damage by regulating the duration and magnitude of immune responses, thus facilitating an efficient antitumor immune response (71). Given the significance of checkpoint inhibitor-based immunotherapy, the variations in immune checkpoint gene activity between the two groups were also examined. We systematically analyzed the expression of 47 immune checkpoint genes, including IDO1, LAG3, CTLA4, TNFRSF9, ICOS, CD80, PDCD1LG2, TIGIT, CD70, TNFSF9, ICOSLG, KIR3DL1, CD86, PDCD1, LAIR1, TNFRSF8, TNFSF15, TNFRSF14, IDO2, CD276, CD40, TNFRSF4, TNFSF14, HHLA2, CD244, CD274, HAVCR2, CD27, BTLA, LGALS9, TMIGD2, CD28, CD48, TNFRSF25, CD40LG, ADORA2A, VTCN1, CD160, CD44, TNFSF18, TNFRSF18, BTNL2, C10orf54, CD200R1, TNFSF4, CD200, NRP1. The expression of 12 immune checkpoint genes demonstrates variability, despite the absence of a substantial variation in PDL1 expression between the two groups (Figure 7d). The expression of six immunological checkpoints, namely CD276, IDO1, TNFRSF8, TNFRSF4, VTCN1, and TNFRSF18, was considerably elevated in the high expression group. The other markers (BTLA, CD200R1, CD40LG, CD44, CD48, and TNFRSF14) showed lower levels in the high-expression group.

Following the previously mentioned research on the relationship between IGF2BP1 expression levels and tumor immunity, the immune status of LUAD patients was assessed utilizing a prognostic model. The stromal score, immune score, and ESTIMATE score for patients with high and low expression were computed using the ESTIMATE algorithm (Figure 8a). The Estimated score is the sum of the stromal and immune scores. Moreover, an elevated stromal score correlates with a higher immunological score and a better prognosis. No significant difference in stromal score was observed between patients with high and low expression. Nonetheless, a considerable disparity was apparent between actual scores and estimated scores. The Estimated score for the low-expression group was markedly higher than that for the high-expression group, indicating a more favorable prognosis.

Noting variations in survival during subgroup analysis of TCGA-LUAD patients treated with various therapies, we postulated that IGF2BP1 expression could be associated with drug sensitivity and conducted a drug-sensitivity assessment. Immune checkpoint inhibition with immunotherapies targeting CTLA-4 and PD-1 has emerged as a viable strategy for treating various cancers. Consequently, we employed the TIDE method with subclass mapping to assess patients’ responses to immune checkpoint inhibitors (CTLA-4 and PD-1). A higher TIDE score indicates more susceptibility to immune escape and a lower probability of benefitting from immunotherapy. A score above 0 is deemed non-responsive, whilst a score below 0 is considered responsive. Notably, the TIDE score for the low expression group was considerably lower than that for the high expression group, suggesting that immunotherapy was more effective in the low expression group and that IGF2BP1 may serve as a biomarker to identify patients likely to benefit from immunotherapy. (Figure 8b).

To achieve a more thorough analysis of chemotherapy response between high- and low-expression LUAD patients, we employed the oncoPredict algorithm to assess chemotherapeutic response using half-maximal inhibitory concentration (IC50) data from the Genomics of Drug Sensitivity in Cancer (GDSC) database for each TCGA sample. 198 substances were evaluated to assess any significant differences in IC50 estimates between the two classes. Of the 198 compounds, 77 exhibited variable IC50 values, and the high-expression group showed increased sensitivity to nearly all of them (Figure 8c). In the high expression group, the sensitivity of 70 molecules to these prospective medicines was elevated. EGFR-tyrosine kinase inhibitors (EGFR-TKIs; Osimertinib and Erlotinib) demonstrated increased sensitivity in the high expression group, while sensitivity to Afatinib and an ALK-TKI (Crizotinib) remained consistent across both groups. The cytotoxic medicines Cisplatin, Oxaliplatin, Gemcitabine, and Epirubicin showed improved efficacy in the high expression group. Notably, the anti-estrogenic agents Tamoxifen and Fulvestrant showed significant responsiveness in the high-IGF2BP1 expression cohort, suggesting that IGF2BP1 expression may influence endocrine therapy outcomes in breast cancer.

Collectively, these findings demonstrated that IGF2BP1 expression was associated with clinical outcomes. Consequently, the expression pattern of IGF2BP1 may serve as a biomarker to inform appropriate therapeutic strategies, potentially offering further insights into individualized treatment approaches for LUAD patients.

The results from the entire dataset were subjected to univariate Cox regression and Lasso Cox regression analyses, focusing on 20 m6A-related RNAs from the TCGA database. indicating that IGF2BP1 overexpression is a significant predictive determinant. Subsequently, we examined whether IGF2BP1 expression serves as an independent predictive biomarker for LUAD to elucidate, in greater detail, the relationships between clinical factors and overall survival in LUAD patients. Cox logistic regression studies, both univariate and multivariate, were conducted in the univariate paradigm. Patients with LUAD were classified into low- and high-expression groups based on the median IGF2BP1 expression level. The univariate and multivariate Cox regression analyses, incorporating clinical and demographic variables (such as sex, age, and TNM stage), showed that IGF2BP1 expression is an independent and significant prognostic marker for LUAD patient outcomes, with an inverse correlation with LUAD risk. In the univariate framework, Stage II (HR = 2.48; p < 0.001), Stage III (HR = 4.72; p < 0.001), Stage IV (HR = 3.17; p = 0.001), T2 stage (HR = 1.63; p= 0.044), T3 stage (HR = 3.85; p < 0.001), T4 stage (HR = 4.28; p < 0.001), N1 stage (HR = 2.47; p < 0.001), N2 stage (HR = 3.60; p < 0.001), and IGF2BP1 expression (HR = 1.37; p= 0.005) exhibited significant associations with overall survival (OS). Multivariate Cox regression analysis identified Stage III (HR = 4.47; p=0.011), Stage IV (HR = 2.57; p=0.027), T3 stage (HR = 2.51; p=0.028), and IGF2BP1 expression (HR = 1.54; p<0.001) as distinct prognostic indicators for LUAD patients (Table 1). These data suggest that IGF2BP1 is highly effective at predicting outcomes in LUAD patients.

The data in Figure 9a show that Kaplan–Meier survival curves indicate a significant reduction in overall survival (OS) for patients with high IGF2BP1 expression compared to those with low expression (p = 0.0071) as IGF2BP1 expression increases. Additionally, to assess the predictive model’s influence, patients were classified into distinct subgroups based on their clinical characteristics, including sex, age, T stage, N stage, and tumor stage. A comparison of patients’ predictions across various expression subgroups was conducted. The Kaplan-Meier curves for overall survival indicated that the survival rate of the low-expression group, as defined by the risk model, was significantly higher than that of the high-expression group in the subgroups of individuals aged 65 years or older, aged 65 years or younger, T1–2 stage, and N0 stage (Figures 9b–e). It is important to note that a statistically significant difference in overall survival (OS) was observed only between the high- and low-expression categories in the T1–2 and N0 stage groups. In contrast, no such difference was observed in the later tumour staging groups (T3-4, N1, and N2). This implies that the influence of IGF2BP1 on patient outcomes may be confined to early stages, whereas in late tumors, the tumor itself exerts a greater impact than IGF2BP1. Furthermore, there was no statistically significant difference in overall survival between the high- and low-expression groups by gender, suggesting that IGF2BP1 expression may be consistent across genders. We deduced that patients with high expression had a worse prognosis, based on differences in overall survival across subtypes.

The predictive efficacy of the risk model was assessed using time-varying ROC curves. The AUCs for overall survival (OS) were 0.740 at 1 year, 0.743 at 3 years, and 0.764 at 5 years (Figure 10a), respectively. The predictive accuracy at the five-year interval was determined to be the highest. The results indicate that the prognostic model exhibits exceptional sensitivity and specificity. In summary, IGF2BP1 expression shows the highest predictive value for LUAD prognosis. Additionally, we evaluated the ROC curves for IGF2BP1 expression in relation to other clinicopathological features. The AUC of IGF2BP1 expression exceeded that of clinical and pathological features (age, T, N, M stage) (Figure 10b), suggesting that IGF2BP1 expression more effectively predicts the occurrence and progression of LUAD, thereby indicating that this prognostic model based on IGF2BP1 expression may be comparatively reliable.

Consequently, our data provide initial evidence of IGF2BP1’s prognostic significance in LUAD patients, suggesting substantial predictive value.

Univariate Cox regression analysis (Table 1) incorporating clinical and demographic variables revealed that age, T stage, N stage, M stage, TNM stage, and IGF2BP1 expression are associated with patient prognosis. To facilitate the prediction of survival prognosis for LUAD patients, we developed a nomogram that integrates IGF2BP1 expression with five additional prognostic risk factors: age, T stage, N stage, M stage, and TNM stage. This predictor estimates 1-, 3-, and 5-year survival probabilities based on the training cohort (Figure 10c). This nomogram can assess a patient’s survival probability at 1, 3, and 5 years. Moreover, compared with clinical criteria, IGF2BP1 expression in the prognostic model exhibited superior predictive capability in the nomogram. The entire nomogram score was derived from the aggregate of each person’s scores of the three components. An elevated overall patient score was associated with a poorer outcome. The calibration curve showed that the projected overall survival for patients closely matched the observed outcomes, confirming the high accuracy of this predictive model (Figure 10d). The data suggest that the nomogram is a superior model for forecasting survival rates compared to individual characteristics, and the risk model is a significant indicator of prognosis in persons with LUAD.

To validate the predictive significance of IGF2BP1 in patients with LUAD, as indicated in the TCGA database, survival analysis was conducted using data from the Shandong Provincial Hospital Affiliated to Shandong First Medical University (SPH) cohort. We performed IHC staining on 215 samples from the Shandong Provincial Hospital Affiliated to Shandong First Medical University (SPH) cohort. Additionally, we classified 89 samples as the high-expression group, 103 as the low-expression group, and 23 as negative-expressors. Figures 11a–f displays representative micrographs of IGF2BP1 IHC staining.

Table 2 indicates that IGF2BP1 expression was strongly correlated with gender and TNM stage (p < 0.05), but not with age, radiation, smoking history, T stage, N stage, or treatment lines. A greater number of samples with high IGF2BP1 expression were identified among individuals with Stage III LUAD. The results indicate high IGF2BP1 expression in LUAD, with statistically significant correlations to specific clinicopathological characteristics of the patients.

To validate the predictive significance of IGF2BP1 in LUAD patients, as observed in the TCGA database, survival analysis was conducted using data from the SPH cohort. The Kaplan–Meier survival curve indicated that high IGF2BP1 expression was substantially associated with an unfavorable prognosis in LUAD (p=0.0221, HR = 1.815), suggesting that IGF2BP1 possesses prognostic significance in LUAD (Figure 11g).

Univariate and multivariate Cox logistic regression analyses were conducted to elucidate the relationships between clinical factors and progression-free survival (PFS) in patients with lung adenocarcinoma (LUAD). In the univariate model, Stage II (HR = 1.23; p < 0.001), Stage III (HR = 1.34; p < 0.001), Stage IV (HR = 2.89; p = 0.001), T2 stage (HR = 1.45; p= 0.036), T3 stage (HR = 2.75; p < 0.001), T4 stage (HR = 3.66; p < 0.001), N1 stage (HR = 1.22; p < 0.001), N2 stage (HR = 2.49; p < 0.001), N3 stage (HR = 3.06; p < 0.001), and IGF2BP1 expression (low/high, HR = 0.25; p < 0.001) were all significantly associated with progression-free survival (PFS). Multivariate Cox regression analysis identified Stage III (HR = 2.57; p=0.034), Stage IV (HR = 1.75; p=0.027), T3 stage (HR = 1.45; p=0.028), T4 stage (HR = 2.84; p=0.033), N1 stage (HR = 1.95; p=0.047), N2 stage (HR = 2.88; p=0.046), N3 stage (HR = 3.79; p=0.021), and IGF2BP1 expression (low/high, HR = 0.24; p< 0.001) as independent prognostic factors for LUAD patients (Table 3). These data suggest that IGF2BP1 is highly effective at predicting outcomes in LUAD patients.