Prognostic Modeling of Lung Adenocarcinoma Based on Hypoxia and Ferroptosis-Related Genes

Background It is well known that hypoxia and ferroptosis are intimately connected with tumor development. The purpose of this investigation was to identify whether they have a prognostic signature. To this end, genes related to hypoxia and ferroptosis scores were investigated using bioinformatics analysis to stratify the risk of lung adenocarcinoma. Methods Hypoxia and ferroptosis scores were estimated using The Cancer Genome Atlas (TCGA) database-derived cohort transcriptome profiles via the single sample gene set enrichment analysis (ssGSEA) algorithm. The candidate genes associated with hypoxia and ferroptosis scores were identified using weighted correlation network analysis (WGCNA) and differential expression analysis. The prognostic genes in this study were discovered using the Cox regression (CR) model in conjunction with the LASSO method, which was then utilized to create a prognostic signature. The efficacy, accuracy, and clinical value of the prognostic model were evaluated using an independent validation cohort, Receiver Operator Characteristic (ROC) curve, and nomogram. The analysis of function and immune cell infiltration was also carried out. Results Here, we appraised 152 candidate genes expressed not the same, which were related to hypoxia and ferroptosis for prognostic modeling in The Cancer Genome Atlas Lung Adenocarcinoma (TCGA-LUAD) cohort, and these genes were further validated in the GSE31210 cohort. We found that the 14-gene-based prognostic model, utilizing MAPK4, TNS4, WFDC2, FSTL3, ITGA2, KLK11, PHLDB2, VGLL3, SNX30, KCNQ3, SMAD9, ANGPTL4, LAMA3, and STK32A, performed well in predicting the prognosis in lung adenocarcinoma. ROC and nomogram analyses showed that risk scores based on prognostic signatures provided desirable predictive accuracy and clinical utility. Moreover, gene set variance analysis showed differential enrichment of 33 hallmark gene sets between different risk groups. Additionally, our results indicated that a higher risk score will lead to more fibroblasts and activated CD4 T  cells but fewer myeloid dendritic cells, endothelial cells, eosinophils, immature dendritic cells, and neutrophils. Conclusion Our research found a 14-gene signature and established a nomogram that accurately predicted the prognosis in patients with lung adenocarcinoma. Clinical decision-making and therapeutic customization may benefit from these results, which may serve as a valuable reference in the future.


Introduction
Lung cancer is one of the most frequent malignancies with high mortality and poor prognosis [1,2]; 80% of lung malignancies diagnosed were NSCLC [3]. LUAD accounts subtypes, the rate of survival for LUAD is still poor [9]. According to statistics, LUAD has a poor prognosis that only 18% could survive longer than 5 years [10]. As a result, the search for valid biomarkers might lead to the establishment of individualized diagnosis and therapy for LUAD patients [11]. e cancer tissue has many specific characteristics, including accelerated cell cycle, alterations of the genome, increase in cell mobility and invasive growth of the cells, incapable of going through normal apoptosis process, and depletion of normal cell functions. Because of these physiological and pathological characteristics, it is difficult for tumors to be treated.
Recently, it has been studied that ferroptosis is a relatively new type of cell death. is process is often accompanied by significant iron buildup and lipid peroxidation in dying cells [12]. It can be distinguished from apoptosis, necrosis, and autophagy by certain key characteristics. Firstly, it is iron-dependent and is induced by the buildup of harmful lipid reactive oxygen species. In addition, polyunsaturated fatty acids are consumed during the process [12]. With the rapid development of the role of iron ions in cancer, new prospects have emerged for their use in cancer therapy [13]. e expression of the S100 calcium-binding protein A4 (FSP1) in lung cancer cell lines is related to resistance to ferroptosis, suggesting that overexpression of FSP1 may be a method for ferroptosis escape [14]. In addition, MAPK pathway activation is associated with the susceptibility to ferroptosis triggered by cystine deprivation in NSCLC cell lines [15] . Alvarez et al. [16] recently found that inhibiting the iron-sulfur cluster biosynthesis enzyme NFS1 induced ferroptosis in vitro and slowed tumor development in LUAD. Additionally, Liu et al. [17] discovered that brusatol, an inhibitor of NRF2, increased the response rate of cystine deprivation-triggered ferroptosis through the FOCAD-FAK signaling pathway in NSCLC cell lines. What is more surprising is that the merger of brusatol and erastin demonstrated a superior therapeutic effect on NSCLC. e findings in these prior studies suggest that ferroptosis is quite important for lung cancer treatment. Based on the above research, we made the following hypothesis that ferroptosis is connected with the prognosis of LUAD, and thus ferroptosis-related genes may function as prognostic biomarkers.
Hypoxia or oxygen deprivation is a feature of most solid tumors because the growth of a tumor requires a large amount of oxygen. As the rapid tumor growth outstrips the supply of oxygen, an imbalance between decreased oxygen supply and increased oxygen demand was formed. is is a typical feature observed in the tumor microenvironment (TME) that increases the aggressiveness of many tumors and also causes abnormal blood vessel formation due to impaired blood supply, leading to poorer clinical outcomes [18][19][20]. Many transcription factors are active in tumor cells when the environment is hypoxic, and these transcription factors regulate cell proliferation, motility, and apoptosis via a variety of downstream signaling mechanisms [21]. is leads to an immunosuppressive TME that reduces the effectiveness of immunotherapy [22] and upregulates the expression of PD-L1, further supporting cancer escape [23,24].
Although several studies have shown that intratumoral hypoxia and HIF1A expression affect overall survival (OS) in LUAD [25][26][27], hypoxia-based cannot be used to estimate who are at a high risk very early.
According to recent research, HIF1A may influence lipid metabolism and cause lipids to be stored in droplets, which reduces peroxidation-mediated endosomal damage and limits cellular ferroptosis [28]. Additionally, HIF-2α has been reported to activate hypoxia-inducible lipid dropletassociated (HILPDA) expression and selectively enrich polyunsaturated lipids, thus promoting cellular ferroptosis [29]. Furthermore, increased ferritin heavy chains under hypoxic conditions can protect HT1080 tumor cells from ferroptosis [30]. ese findings suggest a potential relationship between ferroptosis and hypoxia. But more research is needed to further investigate how ferroptosis and hypoxia interact with each other and how they can affect LUAD patients' prognosis.
A variety of models have been created to predict the prognostication in LUAD according to the TME [31], ferroptosis [32], hypoxia [33], and tumor immunology [34]. However, to our knowledge, there is no reported prognostic role of hypoxia and ferroptosis-interrelated features in LUAD. To fill the gap and broaden the diagnostic and therapeutic potential of LUAD, we performed a comprehensive analysis using TCGA and Gene Expression Omnibus (GEO), aiming to endorse the least prognostic genes for LUAD. Finally, a signature on hypoxia-and ferroptosisinterrelated genes was constructed to know the prognostic value in LUAD patients.

Data Source.
Transcriptomic data from 593 samples, composed of 59 normal and 534 LUAD, from TCGA database were used in this study. A total of 476 LUAD samples had available survival data. e GSE31210 dataset [35,36] (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi? acc�GSE31210), containing transcriptomic data and survival information for 226 LUAD patients, was obtained from the GEO database to validate the established model.

Single Sample Gene Set Enrichment Analysis.
e MSigDB (https://www.gsea-msigdb.org/gsea/msigdb/) was performed to acquire the hallmark gene sets of hypoxias, which consisted of 200 genes. e results show that there are 259 genes related to ferroptosis in total, which were gathered from the FerrDb database (https://www.zhounan.org/ ferrdb/). e TCGA-LUAD database matched the expression patterns of the aforementioned genes. e ssGSEA method (from the R package GSVA) was performed to analyze all samples, and the hypoxia and ferroptosis scores for each sample were then calculated [37].

Coexpression Network Construction.
e TCGA-LUAD transcriptome data were selected for the establishment of gene coexpression networks using the R package WGCNA [38]. Hypoxia and ferroptosis scores were used as phenotypic characteristics. To assess the correlation of all samples in the TCGA-LUAD database, we performed a cluster analysis to ensure the completeness of the samples. As shown in Supplementary Figure 1(a), TCGA-44-3917-01A-01R-A278-07 was identified as an outlier and therefore was not included in this section of the subsequent analysis. During the network construction phase, the soft thresholding power β was obtained above 0.90, with the fit index of the scale-free topology. A dendrogram of all genes was established using the dissimilarity measure to group them together (1-TOM) (Supplementary Figure 1(b)). We set 30 as the minimum module size, and modules with similar gene expressions were clustered and displayed in a tree diagram with color assignments according to the dynamic treecutting algorithm. To identify the modules associated with hypoxia and ferroptosis scores, a heatmap of module-feature relationships with correlation coefficients and P-values was drawn. Modules that had a strong dependency on both scores were identified as modules of interest, and the genes in these modules of interest were defined as hub genes.

Analysis of Differentially Expressed Genes (DEGs).
Transcriptome data from 53 normal and 539 LUAD samples were used as the foundation for comparison to analyze genes expressed differently. DEGs were analyzed using the R package limma, with significance criteria of |log 2 fold change (FC)| > 1 and P < 0.05 as significance thresholds.

Overlap Analysis.
Overlap analysis was used to identify common genes between the identified hub genes and DEGs, which were defined as DE-hypoxia and ferroptosis scorerelated genes for the subsequent analysis.
2.6. Functional Enrichment. Using Metascape (https:// metascape.org) [39], the researchers were able to confirm the functional enrichment of DE-hypoxia and ferroptosis score-related genes in this investigation. P < 0.05 was the significant threshold. e active signaling was analyzed using gene set variation analysis (GSVA) [37], which could compute sample gene set enrichment using a Kolmogorov-Smirnov-like rank statistical analysis. In the present study, a GSVA assessment was used to establish the t score and to allocate 50 hallmark gene signature activity conditions to the groups with high or low risk. At last, we compared the values. e cutoff value was set to |t| > 2.

Identification and Establishment of the Gene Signature.
TCGA's 476 LUAD cases were randomly separated into two groups by using a 7 : 3 split ratio. One group was used for training and another one for testing. e DE-hypoxia and ferroptosis score-interrelated genes that are related to OS were discovered using the TCGA training dataset. e characteristics related to LUAD prognosis were determined by using univariate Cox regression (UCR) analysis. P < 0.05 was considered as significant. After the LASSO-penalized Cox regression (LCR) analysis of the proposed predictive panels, 10-fold cross-validation was used. Risk scores can be generated by using prognostic gene signature. In accordance with the appropriate cutoff of the risk score, patients from the TCGA training and TCGA test sets, as well as GSE31210, were split into two groups. e AUC of the ROC curve and Kaplan-Meier (KM) analyses were applied. External validation was performed using the GSE31210 dataset.

Nomogram Construction and Validation.
To identify whether the risk model can be influenced by clinical factors, UCR and MCR analysis together with the survival R package were performed. Following those analyses, a nomogram was obtained using MCR coefficients of the risk score and clinical variables in the TCGA cohort, which was then analyzed. It was necessary to create calibration curves to determine whether OS for one, three, or five years were consistent with the actual findings (bootstrap-based 1000 iterations resampling validations). We developed these analyses based on the R package rms.

Immune Cells Infiltration (ICI).
e ICI into two groups was determined using the ssGSEA method and the R software [40]. e analysis considered only values with a P < 0.05. e violin diagrams used to illustrate the changes in ICI between separate categories were drawn with the ggplot2 package.

Patients and Tissue Samples.
We performed experimental validation on specimens from five LUAD patients who underwent surgery at Yan'an Affiliated Hospital, Kunming Medical University, to validate 14 hypoxia and ferroptosis score-related signature expression status in LUAD and adjacent normal tissues (ANT). ANTs were used as controls. e institutional and national research committees were followed in the conduct of all procedures, as well as the Helsinki Declaration. e hospital's Ethics Committee gave its approval before any of the operations could be carried out (Permit No. 2017-014-01). All of the patients who took part in the trial gave their informed permission before participation.

Statistical
Analysis. Statistical analysis was performed using R 3.4.3 and GraphPad Prism V9. P-value <0.05 means significant difference. To evaluate survival, both UCR and MCR analyzes were used. Both hazard ratios (HRs) and 95 percent CIs were reckoned to identify genes that were related to OS. Paired t-tests were performed for statistical differences in this study using GraphPad Prism V9.

Filtering for Hypoxia Score-and Ferroptosis Score-Related
Genes in TCGA-LUAD Database. A total of 200 hypoxiainterrelated and 259 ferroptosis-interrelated genes were gained from MSigDB and FerrDB, respectively. e expression conditions of these genes in 593 samples (normal: 59, LUAD: 534) were then matched and utilized as the basis for ssGSEA, which aimed to derive the hypoxia and ferroptosis scores in TCGA database. e ssGSEA outputs for the detailed score results are shown in Supplementary  Table 1. WGCNA was performed by applying the obtained hypoxia and ferroptosis scores as phenotypic data. After excluding the outlier samples, we constructed a sampleclustering tree (Figure 1(a)). Herein, a scale-free network was built when β � 3, which was defined as a soft threshold parameter (Figure 1(b)). Finally, 23 modules were identified according to the dynamic tree-cutting algorithm and were labeled with different colors (Figure 1(c)). e turquoise module was most irrelevant to ferroptosis score (cor � −0.69, P � 3e − 10) and hypoxia score (cor � −0.63, P � 8e − 68), whereas the red module correlated more strongly with both ferroptosis score (cor of −0.47, P � 6e − 34) and hypoxia score (cor � −0.49, P � 2e − 36) (Figure 1(d)). erefore, these two models were identified as the modules of interest. Collectively, 8314 genes (Supplementary Table 2) and 660 genes (Supplementary Table 3) were identified as hub genes and considered as hypoxia and ferroptosis score-related genes for subsequent analysis.

Identification of LUAD-Related
DEGs. Differential expression analysis was used to acquire transcriptome data from TCGA (59 normal and 534 LUAD samples), which was produced using the R program limma. When LUAD samples were compared to normal samples, a total of 1,969 eligible DEGs were obtained, among which 906 were significantly increased in LUAD samples, and 1,063 were significantly decreased (Figure 2(a); Supplementary Table 4).

Analysis of DE-Hypoxia and Ferroptosis Score-Related
Genes. Based on the overlap analysis, we identified 152 common genes from the list of 8,974 hypoxia and ferroptosis score-related genes and the list of 1,969 LUAD-related DEGs, which were defined as DE-hypoxia and ferroptosis score-related genes (Figure 2(b)). In LUAD, 86 of these genes were upregulated, while 66 were inversed. e expression patterns of DE-hypoxia and ferroptosis score-related genes in the TCGA-LUAD database are described in Supplementary Table 5.
Functional annotations obtained from Metascape indicated that DE-hypoxia and ferroptosis score-related genes were mainly augmented in "transcriptional misregulation in cancer," "spermatogenesis," and "positive regulation of cell projection organization" (Figures 2(c) and 2(d)).

Establishment of the Hypoxia and Ferroptosis
Score-Related Signature. In the TCGA training set (n � 334), the association of the 152 identified DE-hypoxia and ferroptosis score-related genes with survival in LUAD patients was analyzed using UCR. As shown in Table 1 Table 6). e patients with LUAD in the TCGA training set were separated into two groups with the cutoff value at 1.0803 (Supplementary Table 7). e allocation of risk scores is shown in Figure 3(d). Association analyses revealed a significant correlation (P < 0.05) between the T stage and various risk groups in the TCGA training set ( Table 2). A significant association between a high-risk score and a poor outcome (P < 0.0001; Figure 3(e)) was shown in the Kaplan-Meier survival curves. ROC curves indicated that hypoxia and ferroptosis score-related signature could be used to predict OS in the TCGA training group (Figure 3(f )). Additionally, the heatmap indicated that the expression levels of KCNQ3, ITGA2, ANGPTL4, TNS4, FSTL3, LAMA3, MAPK4, PHLDB2, and VGLL3 were upregulated with enhancing risk score, but the expression levels of KLK11, SMAD9, WFDC2, SNX30, and STK32A were reduced. Additionally, in individuals with LUAD, T stages are also relevant to these genes expression ( Figure 4).

Validation Prognostic Signature with 14
Genes. We used the same algorithm to compute the risk scores for the patients in the TCGA test cohort (n � 142; Supplementary Table 8) and the GSE31210 dataset (n � 226; Supplementary Table 9). According to cutoff values determined for each dataset, patients were separated into two risk groups. e results corroborated those from the TCGA training set. e risk scores of AUCs for 1-, 3-, and 5-year OS predictions were 0.666, 0.652, and 0.637 in the TCGA test set, respectively ( Figure 5(c)), while the AUCs of the 14-gene signature were 0.741, 0.648, and 0.677 for the three kinds of OS predictions, respectively, using the GSE31210 dataset ( Figure 5(f )). e distribution of LUAD patients with different groups according to each clinical feature in the TCGA test set is shown in Table 3. Association studies revealed a significant (P < 0.01) correlation between the clinical stage and different risk groups in the GSE31210 dataset (Table 4).

Correlation Analysis of Risk Score with Clinical Characteristics of LUAD.
We observed the allocation of patient risk scores according to different clinical characteristics.   Journal of Oncology Interestingly, the distribution of patient risk scores was highly related to the stages of the patients. Risk scores in patients in stage III were increased compared to those in stage I (P < 0.05; Figure 6(a)). In terms of the T stage ( Figure 6(b)), patients with LUAD in T4 had the highest risk scores, which have a significant difference in T1 and T2, but comparable to T3. Patients with LUAD in the T3 stage had slightly higher risk scores than those in the T1 stage (P < 0.01); in the N stage ( Figure 6(c)), patients in the N2 stage had higher risk scores than those in the N0 stage endopeptidase inhibitor activity Transcridtional misreaulation in cancer NABA MATRISOME ASSOCATED multicellular organismal water homeostasis MET promotescell motility ABC transporters CtBP complex kidney development BMP signaling glycoprotein biosynthetic process Vitamin D receptor pathway epithelial cell apoptotic process Plasma lipoprotein assembly, remodeling, and cleara      (P < 0.01). Although the risk score in stage N3 was lower than in stage N2 (P < 0.05), the sample size in stage N3 was too small to be considered valid. Subsequently, the impact of clinical characteristics on the OS in LUAD patients was investigated using KM survival analysis. Specifically, in the stratified analysis of stage ( Figure 6(d)), patients with a lower stage are more likely to have a better prognosis, which showed the same trend with distribution of risk score levels. In the stratified analysis of the T stage ( Figure 6(e)), the T1 stage had a better OS, whereas T3 and T4 stages exhibited a poor prognosis. e worst prognosis in LUAD patients in the T4 stage was consistent with the previous result that patients with T4 stage had the highest risk score. In terms of the N stage ( Figure 6(f )), the N3 stage contained only one LUAD sample, and therefore, its impact on patient prognosis was ignored. Patients with the N0 stage had the longest survival time compared with those with the N2 stage who had the shortest survival time.
e allocation of risk scores and stratified prognosis according to other clinical characteristics, including age, sex, and M stage, are detailed in Supplementary Figure 2.

Subgroup Analysis of the Prognostic Signature.
After establishing a correlation between hypoxia and ferroptosis score-related gene signatures and the aforementioned clinicopathological traits, we aimed to measure whether our model's prognostic efficacy can be utilized for clinical factors. Five patients were separated according to the indicated subgroups, and then data stratification was executed according to age, sex, pathological tumor stage, pathological T stage, pathological N stage, and pathological M stage. e hypoxia and ferroptosis score-related gene signature was able to differentiate between prognoses in all subgroups except for T3-T4 and M1 features, implying a clinically and statistically significant prognostic value (Figure 7).

Independent Prognostic Role of Risk Scores.
We investigated whether the risk score could be the only prognostic factor in LUAD patients using UCR and MCR. Based on the data from in TCGA set, UCR analyses showed that the risk score, stage, T stage, and N stage were significantly related to LUAD prognosis (Figure 8(a)). Subsequently, the abovementioned variables (P < 0.05) were subjected to MCR analysis. e results identified hypoxia and ferroptosis score-related gene signature (risk score) and stage as two independent prognostic factors predicting prognosis in LUAD patients (Figure 8(b)). LUAD patients' OS were predicted using a compound nomogram incorporating the risk score and stage. is approach was developed to provide a more accurate prediction tool for clinical practice (Figure 8(c)). It was evident from the calibration plots that the prognostic nomogram model accurately predicted patient survival with only a slight divergence from the actual outcomes (Figure 8(d)).

Differences in Hallmark Gene Sets between Two Group
Patients. According to the results of the analysis of signature gene sets, signaling pathways converging in numerous biological processes were found to vary in two groups. Notably, hypoxia, TNFα signaling via NF-κB, mitotic spindle, and glycolysis were decreased in the low-risk group. On the other hand, the other group was preferentially associated with bile acid metabolism, pancreatic beta cells, and KRAS signaling (Figure 9 and Supplementary Table 10).

TME Infiltration Pattern of LUAD Based on Risk Score.
e ssGSEA algorithms were used on the data to investigate how risk scores affect TME components. As the results of heatmaps and Wilcoxon tests performed on TCGA-LUAD datasets, the infiltration of several TME contents, such as      eosinophils and immature dendritic cells, was increased in the less-risk group, whereas the ICI of activated CD4 T cells and others was more in the other group, as depicted in Figure 10.

Validation of Seven Selected Prognostic Genes Based on qRT-PCR.
According to the expression profiles of the identified DEGs (Supplementary Table 5), TNS4, WFDC2, and ITGA2 were revealed to be all highly expressed, while MAPK4, SMAD9, KLK11, and LAMA3 were all downregulated in LUAD samples from the TCGA dataset. As shown in Figure 11, the high expression of TNS4, WFDC2, and ITGA2 and the low expression of MAPK4, SMAD9, KLK11, and LAMA3 in LUAD tissues (n � 10) were confirmed compared with the expression levels in the ANTs (n � 10).

Discussion
As well known, lung cancer is one of the general forms of malignancy globally. Nearly 80% of lung cancer patients have NSCLC, and nearly 50% have LUAD [42]. LUAD is a malignant tumor that affects the lungs and has a poor prognosis [43]. Although there have been breakthroughs in the treatment of patients with LUAD, the OS rate in these individuals remains low. Ferroptosis is a particular kind of programmed cell death [17]. Ferroptosis-related research on lung cancer has mostly focused on the identification of related biomarkers that could induce ferroptosis [16,[44][45][46]. Hypoxia is also related to high proliferation rates in tumor cells [47]. Tumor hypoxia has a broad range of consequences, affecting a variety of biological systems, including metabolic changes, angiogenesis, and metastasis [48][49][50]. Numerous hypoxia-associated genes are associated with lung adenocarcinoma [51,52]. However, no high-throughput research has been conducted to date to explore the possible prognostic value of them in LUAD.
Here, the ferroptosis and hypoxia Z-scores of each sample were estimated as clinical features based on the expression of ferroptosis and hypoxia-related genes identified in each sample, respectively. We obtained 23 modules, and the turquoise module showed no relationship with ferroptosis scores (cor � −0.69, P � 3e − 10) and hypoxia scores (cor � −0.63, P � 8e − 68), while the red module correlated more strongly with both scoring phenotypes, with ferroptosis score and hypoxia score. We then identified 152 common genes from the list of 8,974 hypoxia and ferroptosis  score-related genes and 1,969 LUAD-related DEGs, which were defined as DE-hypoxia and ferroptosis score-related genes, respectively. Functional annotations obtained from Metascape indicated that DE-hypoxia and ferroptosis score-related genes were mainly enriched in "transcriptional misregulation in cancer," "endopeptidase inhibitor activity," and "positive regulation of cell projection organization." Overexpression of oncogenic transcription factors has been proven in recent research to change cells' core autoregulatory circuitry, which has long been recognized to induce tumorigenesis due to mutations in transcription factor genes [53]. erefore, it is possible to intervene in this pathway to prevent the development of LUAD.
Of the 152 DE-hypoxia and ferroptosis score-related genes, 7.3% (17/152) were associated with prognosis in univariate Cox analysis. In addition, univariate Cox analysis identified six genes as protective markers and 11 genes as risk factors for patients with LUAD. Fourteen genes were identified using LASSO Cox regression (MAPK4, TNS4, WFDC2, FSTL3, ITGA2, KLK11, PHLDB2, VGLL3, SNX30, KCNQ3, SMAD9, ANGPTL4, LAMA3, and STK32A) to construct prognostic-related gene signatures and develop prognostic models to classify LUAD patients into two groups with various risks. Herein, we suggested that lower-risk patients seem to live longer. Additionally, we built a nomogram using MCR analysis and proved its predictive ability using ROC curves, calibration plots, and decision curves.
MAPK4 overexpression promotes LUAD progression [54]. Tensin 4 (TNS4) is involved in MET-induced cell motility and is connected to the GPCR signaling pathway. According to one study, increased TNS4 expression leads to poor treatment outcomes in gastric cancer patients [55]. WFDC2 is upregulated in lung cancer [56][57][58] and has thus recognized the clinical application of WFDC2 as a serum tumor marker in the early diagnosis and efficacy monitoring of lung cancer [59]. In addition, in a study of individuals with LUAD, Song et al. [34] reported that WFDC2 was substantially related to the TNM stage of LUAD and prognosis of patients. Recent studies have reported substantial overexpression of FSTL3 in a subset of cancers [60][61][62]. Additionally, in patients with NSCLC and thyroid carcinoma, FSLT3 expression is substantially linked to lymph node metastasis and poor prognosis [60,61]. ITGA2 overexpression is essential for tumor development, metastasis, and motility, and this molecule triggers the overexpression of the STAT3 signaling pathway, thus promoting tumor progression [63]. KLK11 protein is expressed more in NSCLC serum, although KLK11 mRNA levels are lower in cancerous lung tissues than in ANTs [64]. Leakage of these secreted proteins into the systemic circulation due to disruption of lung structure during angiogenesis or development may be the reason for this discrepancy between low mRNA levels and elevated serum protein levels in lung cancer [65]. It has been well studied that PHLDB2 is linked to a variety of malignancies  Figure 10: (a, b) Heatmap illustrating the distributions of immune cell subsets, fibroblasts, and endothelial cells assessed via MCP-counter (a) and ssGSEA (b) algorithms in the TCGA-LUAD cohort. (c, d) Wilcoxon analysis of the differing TME subtype distributions between two groups in the TCGA-LUAD cohort. * P < 0.05, * * P < 0.01, and * * * P < 0.001. [66,67]. PHLDB2's primary role is to control migration through interacting with the transcription factors CLASPS, prickle 1, and liprin 1 [68,69]. According to Ge et al., patients with lower PHLDB2 expression have a better prognosis [70]. VGLL3 is a unique Ets1 interacting partner that inhibits adipocyte differentiation and controls trigeminal nerve development [71]. VGLL3 acts as a coactivator of mammalian toxicity equivalency factors and is implicated in many kinds of cancers, including breast, colon, and lung cancers [72,73]. Methylation, phosphorylation [74], and dephosphorylation of SMAD9 may function in the progression of lung cancer [75]. Tumor cell-derived human angiopoietin-like protein 4 (ANGPTL4) has been shown to disrupt vascular endothelial cell connections, enhance pulmonary capillary permeability, and facilitate tumor cell protrusion through the vascular endothelium, which is involved in lung cancer [76]. rough the synergistic action of AP-1 binding sites [77], the epithelial enhancer mediates the production of laminin subunit alpha 3 (LAMA3), which is associated with tumor progression. Xu et al. [78] reported that it was discovered that the inhibition of LINC00628 decreased LUAD cell proliferation and drug resistance by lowering the methylation of the LAMA3 promoter. STK32A is important in cellular balance and transcription factor phosphorylation, together with cell cycle regulation, and its overexpression leads to enhanced NSCLC cell progression, as well as enhanced NF-κB p65 phosphorylation and inhibition of apoptosis [79]. SNX30 encodes sorted nexin-30 protein, a member of the sorted nexin, which a large class of proteins localized in the cytoplasm with membrane-bound potential via a phospholipid-binding domain [80]. KCNQ3 encodes a protein that regulates neuronal excitability, and GCSH encodes a mitochondrial protein that forms the glycine cleavage system [81]. However, there is a lack of research on the mechanisms of action of these two genes in cancer.
Following this assessment, KM survival studies demonstrated that the 14 prognosis-associated genes may have a contribution to the initiation and development of LUAD in certain individuals. It came as a surprise to observe that risk scores for the 14-gene prognostic profile were shown to be strongly correlated with the OS in LUAD patients in two cohorts split by the TCGA and one GEO validation cohort. We discovered that modulation of the prognostic gene profile was linked with the LUAD survival models (T, N, M, stage, sex, and age) in our study. Furthermore, the nomogram of independent risk factors, which included risk score models, had a good predictive value and might assist clinicians in making optimum treatment choices to enhance the OS rates of patients with LUAD in the future. ese results suggest that hypoxia-and ferroptosis-related genes were indispensable in the construction of prognostic models for LUAD development and that they may have the potential to act as OS biomarkers. Our findings suggested that the signaling pathways that converge in various biological processes differ between two groups, and the hypoxia, TNFα, signaling via NF-κB, mitotic spindle, and glycolysis were significantly downregulated in the less-risk group. Additionally, 14 prognosisrelated genes in LUAD, including one hypoxia-related gene, ANGPTL4, were significantly expressed in the tumor tissues. is finding reflects the dependence of LUAD on hypoxia and the heterogeneity of hypoxia responses in the low-and high-risk groups. Hypoxia heterogeneity indicates its involvement in promoting a phenotypic variety of cancer cells in the TME, which promotes metastasis and therapeutic resistance. Li et al. [82] demonstrated that suppressing NLRP2 boosted cell proliferation through NF-κB signaling activation, thus resulting in an EMT phenotype in LUAD cells. erefore, the regulatory pathways involved in NF-κB also function in the progression of LUAD. e evidence implies that LUAD pathogenesis is a complicated biological process involving multiple genes. Apart from that, dysregulation of multiple genes may contribute to the progression of LUAD by a variety of distinct processes. e differences in GSVA signatures and prognostic genes between the two groups have the potential to be explored in a more in-depth study. ese discoveries may, in general, open new avenues of investigation of additional molecular mechanisms of LUAD for academics and physicians.
Significant differences in immune infiltrating cell types between two groups were shown in this study. Interestingly, the enrichment fraction of activated CD4 T cells and neutrophils was enhanced in the high-risk group, whereas the enrichment fraction of eosinophil and immature dendritic cells was found in the low-risk group. Immune cells, neutrophils that infiltrate tumor tissue, called TANs, also play a role in antitumor immunity. TANs stimulate T cell responses in lung cancer rather than have an immunosuppressive effect [83]. In LUAD, overexpression of bridging granule genes is associated with a significant enhancement in infiltration of activated CD4 and CD8 T cells [84]. We hypothesize that the inflammatory response induced by immune cells may function in accelerating tumor cell mutations, which in turn may affect patient prognosis. e specific mechanisms by which the tumor immune microenvironment affects prognosis remain to be explored.
Here, a prognostic model of LUAD with general applicability was successfully developed and validated based on hypoxia and ferroptosis. In addition, we performed experiments to validate the 14 molecules in the model. Of these, seven molecules were validated by qRT-PCR to be significantly different between tumor and paracancerous tissues. However, our study has some limitations. Due to the lack of studies on hypoxia and ferroptosis in tumors, the information provided by MSigDB and FerrDB websites may be inaccurate, as the references were manually obtained from previous studies. More studies will do to validate the roles of these fundamental prognostic genes' hypoxia-and ferroptosis regulation roles in LUAD [3]. Both cohorts (TCGA-LUAD and 1 GEO cohort) were used to construct predictive signature.
is hypoxia-and ferroptosis-predictive signal may be more reliable if examined in our research center's prospective clinical trial cohort.

Conclusion
Hypoxia and ferroptosis are two major mechanisms associated with lung adenocarcinoma development. In this research, the candidate genes associated with hypoxia and ferroptosis scores were identified; as a result, we have found a 14-gene signature and developed a predictive nomogram that could accurately predict OS in individuals with LUAD. ese results may be useful in facilitating the making of medical decisions and personalizing therapeutic interventions.

Supplementary Materials
Supplementary Figure 1. Hypoxia and ferroptosis scores were used as phenotypic characteristics. To assess the correlation of all samples in the TCGA-LUAD database, we performed a cluster analysis to ensure the completeness of the samples. Supplementary Figure 2.