We obtained transcriptomic data for 526 LUAD samples and 59 normal samples using the TCGA database (https://portal.gdc.cancer.gov/). Of these, 493 LUAD patients with complete clinical information were selected for the study of OS and clinical relevance. Information on 447 LUAD patients with disease-free survival (DFS) was selected for DFS analysis. The GSE32863 (20), GSE31210 (21) and GSE13213 (22) datasets are from the GEO database (https://www.ncbi.nlm.nih.gov/geo/). To validate METTL7A expression in paracancer and LUAD samples, GSE32863 dataset was used. The GSE31210 and GSE13213 datasets were used to validate the results of survival analysis.

At the People’s Hospital Affiliated to Jiangsu University, we collected surgically resected frozen tumors and paracancerous tissue from 20 patients with pathological diagnosis of LUAD. Immediately after removal, surgical specimens were placed in liquid nitrogen and frozen to -80°C. The Ethics Committee of the People’s Hospital Affiliated to Jiangsu University supervised and approved the research project (approval number K-20220088-W). Both patients and volunteers gave express written consent to participate in this research investigation.

Both the TIMER database (23) (http://timer.cistrome.org/) and the GEPIA database (24) (http://gepia.cancer-pku.cn/) were searched for information on the expression of METTL7A in pan-cancer.

LUAD patients were divided into two groups, namely high- and low-expression groups, with the ideal cut-off value of gene expression serving as the threshold. This was done so that the function that METTL7A plays in the survival of patients could be determined. For the purpose of conducting a survival analysis, the ‘survival’ R package (version 3.1.12) was employed, and a Kaplan–Meier curve was produced. When the P-value was less than 0.05, researchers considered there to be a significant difference in survival rates between the high expression and low expression groups.

The Wilcoxon test or the Kruskal-Wallis test was used to assess whether a link existed between the expression of METTL7A and the related clinicopathological features. Gender, age, pathological T, N, M stage, and tumor stage were among these variables. This association was represented using a violin plot. In addition, a K-M survival analysis was performed on the identified subtypes with distinct clinicopathological features.

In the TCGA-LUAD dataset, we performed a univariate cox regression analysis followed by a multivariate Cox regression analysis to identify independent predictors of OS or DFS. After that, the ‘cph’ function in R was used to generate a nomogram that included the independent prognostic variables. The calibration curves and ROC curves for one, three, and five years were constructed in order to further validate the nomogram’s capacity for accurate prediction and efficiency.

We used the “GSVA” R package (25) to analyze changes in enriched pathways in high and low expression groups based on Hallmark genes downloaded from MsigDB (http://www.gsea-msigdb.org/gsea/msigdb/index.jsp) set. In addition, we compared the enrichment analysis of the high and low expression groups. The value of |t| must be > 1 for the Hallmark genome to be considered significantly enriched in the pathway. Further calculations were performed to determine differences in enriched pathways in the low and high expression groups, and matching heatmaps are shown.

The ‘Estimation of Stromal and Immune cells in MAlignant Tumours using Expression data’ (ESTIMATE) method was performed to determine the stromal-, immune- and estimate-scores of each LUAD individual in both the low-expression and high-expression groups (26). The enrich score of immunological infiltration cells was computed using single-sample gene set enrichment analysis (ssGSEA) (27) and the Microenvironment Cell Populations-counter (MCP-counter) method (28). The violin plot was developed so that the changes in immune cell infiltration that occurred between the low-expression and high-expression groups could be shown. As was done in the previous publication (29), the geometric mean of the expression of GZMA and PRF1 in transcripts per million (TPM) was used to compute the cytolytic activity (CYT) of each LUAD patient in the high- and low-expression subgroups.

The ‘limma’ package was utilized with the screening criteria of |log2FC|>0.5 and FDR<0.05 in order to identify the differentially expressed microRNAs (DE-miRNAs) and long noncoding RNAs (DE-lncRNAs) between LUAD samples and normal samples. This was the first step in the process. After that, we predicted the interaction between METTL7A and DE-miRNAs by using the miRWalk website with a confidence level of 1, and we only kept the relationship pairings that had opposing expression patterns in accordance with the ceRNA mechanism. In a similar manner, we utilized StarBase to predict the interaction between DE-lncRNAs and DE-miRNAs. However, we only kept the pair combinations that showed opposing expression patterns. Cytoscape (30) was used to design the METTL7A-DElncRNA-DEmiRNA ceRNA network, which was the last step.

Using the Nuclezol LS RNA Isolation Reagent and according to the protocols provided by the manufacturer (ABP Biosciences Inc.), total RNA was extracted from both the 10 normal tissue samples and the 10 LUAD tissue samples. Next, using the SureScript-First-strand-cDNA-synthesis-kit (GeneCopoeia) and following the instructions provided by the manufacturer, total RNA was converted into cDNA by the process of reverse transcription. qPCR was then carried out with the use of the BlazeTaq™ SYBR ^® Green qPCR Mix 2.0 (GeneCopoeia). The following thermocycling conditions were used for quantitative polymerase chain reaction (qPCR): one cycle at 95°C for one minute (initial denaturation), followed by forty cycles of twenty seconds at 95°C (denaturation), twenty seconds at 55°C(annealing), and thirty seconds at 72°C (extension). Table 1 included a rundown of the primer sequences for your perusal. Calculations were done with the 2−ΔΔCq technique (31) after the relative expression level was first standardized with respect to the endogenous control GAPDH. The student t-test d was utilized to analyze the differences that were found between the two groups. In statistical analysis, a result was considered statistically significant if the two-tailed P-value < 0.05.

The programming language R was used to carry out all of the studies, and the Wilcoxon test and the Kruskal-Wallis test were utilized in order to compare the data obtained from the various groups. In the event that it was not indicated above, a P-value of less than 0.05 was regarded as statistically significant.

Compared to corresponding normal samples, we detected that METTL7A was down-regulated in18 types of cancer, which include BLCA, BRCA, CESC, CHOL, COAD, ESCA, HNSC, KICH, KIRC, KIRP, LIHC, LUAD, LUSC, PRAD, READ, STAD, THCA, and UCEC, based on the data from TIMER and GEPIA database ( Figures 1A, B ). To further investigate the expression level of METTL7A in LUAD individuals, we examined the expression of METTL7A in normal and LUAD individuals, as well as matched paracancer and LUAD samples from the TCGA-LUAD and GSE32863 datasets. As demonstrated in Figures 2A-D , METTL7A was definitely down-regulated in LUAD individuals relative to normal individuals or matched paracancer samples, implying a suppressive effect in LUAD development. To further confirm the expression level of METTL7A in clinical LUAD and paracancer individuals, 10 pairs of clinical samples were obtained, their RNA was extracted, and RT-qPCR was conducted. METTL7A was down-regulated in the LUAD samples, as shown in Figure 2E , similar with the results from the public database.

To identify the relationship between METTL7A expression level and clinicopathological characteristics, we analyzed the variances in METTL7A expression among subgroups of clinical factors. As shown in Figure 3 , METTL7A expression was linked with age and T stage (p < 0.05). The expression of METTL7A was higher in patients aged > 65 than in those aged ≤ 65, and the expression of M gradually decreased from the T1 stage to the T3 stage.

Next, we investigated the role of METTL7A in LUAD prognosis. We discovered that the overall survival (OS) rate and disease-free survival (DFS) rate of the METTL7A high-expression subgroup in the TCGA-LUAD dataset were both substantially greater than those of the METTL7A low-expression subgroup ( Figures 4A, B ). In accordance with the findings from the TCGA-LUAD database, the OS of the METTL7A high-expression subgroup was greater compared to the METTL7A low expression group in both the GSE31210 and GSE13213 datasets ( Figures 4C, D ). The above results suggested that METTL7A was a protective factor in the prognosis of LUAD.

In the following, we performed the survival analysis of LUAD patients under different subgroups of clinical factors, and the result was displayed in Supplementary Figure S1 . After that, we utilized stratified survival analysis to determine whether or not the METTL7A was connected to the survival of LUAD individuals under various clinicopathological features. As shown in Figure 5 , we discovered that the level of expression of METTL7A was capable of effectively predict OS for all subgroups based on various clinical features, with the exception of M1 stage.

We conducted univariate and multivariate Cox regression analyses in order to further assess whether the expression of METTL7A was independent of those other clinical features (age, gender, pathologic M, pathologic N, pathologic T, and pathologic stage) as a prognostic factor for LUAD individuals. According to the results of the univariate analysis, the pathologic N, pathologic T, pathologic stage, and METTL7A were linked with the patients’ overall survival and disease-free survival ( Figures 6A, B ). According to the findings of the multivariate Cox regression studies, the METTL7A can be utilized independently to predict the overall survival and disease-free survival of LUAD individuals ( Figures 6C, D ). In addition, the multivariate analysis revealed that pathologic N and pathologic T are both independent prognostic variables. Following that, we developed a nomogram to predict the 1-, 3-, and 5-year overall survival of individuals based on independent prognostic markers ( Figure 6E ). According to the calibration plots, the nomogram was effective for predicting the 1-, 3-, and 5-year survival probability in LUAD individuals ( Figures 6F–H ).

To explore the molecular mechanisms underlying the differential prognosis of the METTL7A high- and low-expression groups, we performed a GSVA analysis based on the hallmark gene set. The results showed that 20 pathways were enriched in the high-expression group and 15 pathways were enriched in the low-expression group ( Figures 7A, B ). ‘MTORC1 signaling’, ‘glycolysis’, ‘DNA repair’, ‘inflammatory response’, ‘G2M checkpoint’, ‘oxidative phosphorylation’, ‘TNFA signaling via NFKB’, ‘IL2 STAT5 signaling’, and ‘WNT beta catenin signaling’ were enriched in the high-expression group. ‘Mitotic spindle’, ‘HEDGEHOG signaling’, ‘bile acid metabolism’, ‘NOTCH signaling’, ‘KRAS signaling up’, ‘TGF beta signaling’, ‘hypoxia’, ‘cholesterol homeostasis’, ‘PI3K AKT MTOR signaling’, ‘interferon alpha response’, and ‘p53 pathway’ were enriched in the low-expression group.

Because multiple immune-related pathways were significantly enriched in both the low- and high-expression subgroups, and the tumor microenvironment (TME) is an essential component in the development of tumors. Following that, we carried out a comparative study between the METTL7A low-expression and high-expression subgroups by performing a differential analysis of immune-, stromal, and estimate- scores. As displayed in Figure 8A , the scores for immune, stromal, and estimation were all considerably higher in the high-expression group than they were in the low-expression group. After that, we performed an analysis on the immune infiltrating cells in both the low-expression and high-expression subgroups of METTL7A. The findings of the MCP-counter indicated that the percentage of all cell types, with the exception of fibroblasts, was higher in the subgroup with higher levels of METTL7A expression than low expression subgroup ( Figures 8B, D ). According to the findings of ssGSEA, the enrich score of high-expression subgroup of aDC, B cells, CD8 T cells, Cytotoxic cells, DC, Eosinophils, iDC, Macrophages, Mast cells, Neutrophils, NK cells, pDC, T cells, T helper cells, Tcm, Tem, TFH, Tgd, and Th1 cells was significantly higher than the enrich score of low-expression subgroup. However, compared to the high-expression group, the low-expression group had a considerably higher enrich score for Th2 cells ( Figures 8C, D ). There was an association between the cytolytic activity and inflammation. In addition, the CYT of each patient with LUAD in the low-expression and high-expression group was estimated and exhibited in Figure 8E . Our findings also showed that the CYT of the high-expression subgroup was considerably greater than the low-expression subgroup. The above data revealed that the immunological microenvironment of the low-expression subgroup was suppressed, which may be connected to the poor prognosis.

To explore the ceRNA regulatory mechanism of METTL7A, we finally constructed a regulatory network containing 1 mRNA, 2 miRNAs, and 5 lncRNAs as described in Materials and Methods 2.9 ( Figure 9 ). The hsa-miR-5698 and hsa-miR-4788 regulated the expression of METTL7A. C14orf132, SPATA13, MTHFS, and TSTD3 targeted the hsa-miR-5698. MTHFS and SHANK3 targeted the hsa-miR-4788.