Construction of NEIL3 as a Prognostic Biomarker and Co-expressed Prognostic Signature in Lung Adenocarcinoma

Background: Lung adenocarcinoma (LUAD) is the leading cause of cancer-related death. This study aimed to develop and validate reliable prognostic biomarkers and signature. Methods: Differentially expressed genes were identied based on three Gene Expression Omnibus (GEO) datasets. Based on three LUAD cohorts’ data included 1052 samples and extracted from our cohort, GEO, The Cancer Genome Atlas, we explored clinicopathological features and the expression of NEIL3 to determine its clinical effect in LUAD. Western blotting, Real-time quantitative PCR; (22 pairs of tumor and normal tissues), and immunohistochemical analyses (406-tumor tissues subjected to microarray) were conducted. TIMER and ImmuCellAI analyzed relationship between NEIL3 expression and the abundance of tumor-inltrating immune cells in LUAD. The co-expressed-gene prognostic signature was established based on the Cox regression analysis. Results: This study identied 502 common differentially expressed genes and conrmed that NEIL3 was signicantly overexpressed in LUAD samples(P<0.001). Increased NEIL3 expression was related to advanced stage, larger tumor size and poor overall survival (p < 0.001) in three LUAD cohorts. The proportions of natural T regulatory cells and induced T regulatory cells increased in the high NEIL3 group, whereas those of B cells, Th17 cells and dendritic cells decreased. Gene set enrichment analysis indicated that NEIL3 may activate cell cycle progression and P53 signaling pathway, leading to poor outcomes. We identied nine prognosis-associated hub genes among 370 genes co-expressed with NEIL3. A 10-gene prognostic signature including NEIL3 and nine key co-expressed genes was constructed. Higher risk-score was correlated with more advanced stage, larger tumor size and worse outcome (p<0.05). Finally, the signature was veried in test cohort (GSE50081) with superior diagnostic accuracy. Conclusions: This study suggested that NEIL3 has the potential to be an immune-related therapeutic target and an independent predictor for LUAD. We also developed a prognostic signature for LUAD with a precise diagnostic accuracy.


Background
Lung cancer, accounting for almost one-quarter of all cancer deaths, has become more common than breast, brain, colorectal, and prostate cancers combined and is now the leading cause of cancer-related death globally [1]. Lung adenocarcinoma (LUAD) is the most frequent pathological subtype, accounting for nearly 45% of lung cancer cases. [2] The 5-year relative survival rate of LUAD is only 5% because about 57% of patients are diagnosed with advanced stage and metastatic disease [3,4]. Patients with LUAD have improved overall survival (OS) after surgery and radiotherapy when diagnosed sooner. With the development of immune checkpoint inhibitors (atezolizumab, pembrolizumab, nivolumab, etc.), the survival of LUAD patients has signi cantly improved; thus, such treatments have attracted considerable attention [5,6]. However, limited data are available regarding the relationship between biomarkers and immune responses. Therefore, exploring and identifying effective immune-related biomarkers for LUAD to reduce the mortality rates and develop innovative targeted therapies remains crucial.
In the past ten years, researchers have conducted new studies and explored meaningful genes using bioinformatics techniques. Here we identi ed 502 commonly expressed genes from three LUAD datasets (GSE32863, GSE33532, and GSE43458). After an extensive literature review, we found that DNA endonuclease VIII-like 3 (NEIL3) is a very promising DNA repair gene, but few studies have examined its role in cancer development. NEIL3, a member of the Nei-like (NEIL) DNA glycosylase family, was the rst enzyme to be identi ed and from which the damaged bases in the base excision repair (BER) pathway were excised [7,8]. NEIL3 has a glycolysis domain and apurinic/apyrimidinic lyase activity that excises damaged bases and generates a single-strand break [8,9]. NEIL3 repairs telomere oxidative damage and protects telomere integrity during the S phase to enable accurate chromosome segregation in actively dividing cells [10]. Some ndings based on a Neil3-/-mice model indicated that the NEIL3 mutation was linked to impaired B cell function and severe autoimmunity [11]. NEIL3 promotes the proliferation of cardiac broblasts and neural progenitor cells in the brain and tends to be overexpressed in cells with a high proliferative capacity, such as cancer cells and those in the bone marrow [7,[12][13][14]. The transcription of NEIL3 as a cell cycle dependent gene is regulated by BRG1 in breast cancer cells [15,16].
Moreover, NEIL3 highly expressed in primary malignant melanomas, inducing metastatic tumors [17]. Much evidence suggests that NEIL3 as a DNA repair gene is connected to tumorigenesis, therapy resistance, and poor prognosis in astrocytoma [18,19]. Evidence has shown that NEIL3 participates in regulating the cell proliferation process. However, the association between NEIL3, cancer prognosis, and immune response in LUAD is unclear.
Here we analyzed the correlation between NEIL3 expression and the clinical characteristics, and the prognosis of LUAD patients based on the gene expression and clinicopathology data of three different cohorts (The Cancer Genome Atlas [TCGA], Gene Expression Omnibus [GEO], and our patient group). We also investigated the relationship between NEIL3 expression and tumor-in ltrating immune cells using Immune Cell Abundance Identi er (ImmuCellAI) and Tumor Immune Estimation Resource (TIMER). To better understand the role of NEIL3 in LUAD development, we performed a gene set enrichment analysis (GSEA) and screened out 370 NEIL3 co-expressed genes. Importantly, nine hub genes were independent predictors for LUAD; thus, we attempted to construct a prognostic signature and validate its prognostic accuracy in other cohorts.

Human tissue samples
Twenty-two pairs of frozen LUAD and adjacent noncancerous tissues obtained from the A liated Hospital of Nantong University between March 2018 and June 2019 were subjected to real-time quantitative polymerase chain reaction (RT-qPCR) and western blotting. LUAD tissue microarray and clinical data for 406 tumor samples and 50 normal lung tissue samples were derived from the Pathology Department of the A liated Hospital of Nantong University and used as our cohort. Informed consent was obtained from all patients before the study. The ethical committee of A liated Hospital of Nantong University approved the study.

Data Resources And Preprocessing
The GEO cohort included ve gene expression datasets (http://www.ncbi.nlm.nih.gov/geo). GSE33532, GSE30219, and GSE43458 were used to screen out differentially expressed genes (DEGs) between LUAD and adjacent lung tissues by GEO2R (logFC > 1; adjusted p < 0.001), which included 40 LUAD tissues and 20 adjacent lung tissues, 85 LUAD tissues and 14 adjacent lung tissues and 80 LUAD tissues and 30 adjacent lung tissues, respectively. The GEO cohort also included GSE31210 (226 LUAD tissues and 20 adjacent lung tissues) and GSE50081 (127 LUAD tissues of stage I and II), which were used to explore the relationship between NEIL3 expression and the clinical outcomes of LUAD patients.
We utilized the gene expression pro le and clinical information of the TCGA cohort (work ow type: HTSeq-FPKM; https://portal.gdc.cancer.gov/projects), and determined NEIL3 gene expression using R software (version: 3.5.3) and Strawberry Perl (version: 5.30.2.1). The TCGA cohort included 293 tumor samples and 54 normal lung tissues after he elimination of samples for which key clinical information was missing such as age, sex, Tumor Node Metastasis stage, overall survival, distant and lymph node metastasis. Meanwhile cases with a follow-up time of less than 90 days were deleted. Our work was performed in accordance with the TCGA publication requirements.
Functional enrichment analyses GSEA, a powerful calculation software (http://software.broadinstitute.org/gsea/ index.jsp) based on Gene Ontology (GO) and the Kyoto Gene and Genomic Encyclopedia (KEGG), was used to investigate the possible biological functions of NEIL3. The cohort of LUAD patients was divided into high and low NEIL3 expression groups by median values. The gene sets used in this work (c2.cp.kegg.v5.2.symbols.gmt) were downloaded from the Molecular Signatures Database ( http://software.broadinstitute.org/gsea/msigdb/index.jsp) [20]. We performed functional enrichment of the NEIL3 and co-expressed genes using the Bohao Online Enrichment Tool (http://enrich. shbio.com/) [21]. When a false discovery rate (FDR) and the nominal p were less than 0.05, the enrichment results were deemed statistically signi cant.

Immune in ltrates analysis
TIMER is a user-friendly web tool that is used to investigate the molecular characterization of tumor immune system interactions including six major analytic modules (https://cistrome.shinyapps.io/timer/). We evaluated the correlation between NEIL3 expression and the abundance of the six tumor-in ltrating immune subsets in LUAD: B cells, CD4 + T cells, neutrophils, CD8 + T cells, dendritic cells(DCs), and macrophages [22]. ImmuCellAI has a powerful ability for tumor-immune in ltration estimation, especially in the abundance of 18 T-cell subsets (http://bioinfo.life.hust.edu.cn/ImmuCellAI#!/) [23]. A gene set signature of TCGA LUAD patients was uploaded to the website. ImmuCellAI predicted the abundance of 24 immune cells in the sample including 18 T-cell subsets. We measured the immune response of 24 immune cells to evaluate their association with NEIL3 expression in LUAD and performed "vioplot" package to visualize the data. At values of p < 0.05, the results were considered statistically signi cant.
Predictive signature construction and risk score calculation The Cox' proportional hazards model is often applied to survival analyses. We selected out nine prognostic signatures from among the 10 hub genes via "Survival" package and Cox regression analysis. We also performed a multivariate Cox regression analysis including NEIL3 and nine prognostic signatures, and constructed a 10-gene prognostic model to evaluate individual survival risk as follows: risk score = 0.00043 × NEIL3 expression level + (-0.00907) × CCNB2 expression level + 0.00357 × CDK1 expression level + (-0.04486) × CDC45 expression level + 0.05000 × BUB1B expression level + (-0.04195) × BUB1 expression level + (-0.0003) × KIF23 expression level + 0.047506 × CCNA2 expression level + (-0.00037) × UBE2C expression level + 0.1721 × NCAPG expression level. The optimal cutoffs and related speci city and sensitivity from receiver operating characteristic (ROC) curves were determined using a conventional method. Values of p < 0.05 were considered statistically signi cant.

Immunohistochemistry assay
The slides of LUAD tissues were stained with anti-human NEIL3, followed by horseradish peroxidase secondary antibody (cat. #ab205718; Abcam) and diaminobenzidine treatment. Immunohistochemistry score (IHC-score) = staining intensity score × staining area ratio score of positive cells. The staining intensity was scored as 0 for negative, 1 for weakly positive, 2 for medium positive, and 3 for strongly positive. The positive area ratio was scored as 0 (0%), 1 (1-25%), 2 (26-50%), 3 (51-75%), or 4 (76-100%) [24]. When the IHC-score was less than 6, the case was classi ed into the low expression group; otherwise, the case was classi ed into the high expression group. The staining results were independently scored by two pathologists.

Statistical analysis
All statistical analyses were performed using R (v.3.5.3) and SPSS version 15.0 (SPSS Inc., Chicago, IL, USA). To evaluate the correlation between NEIL3 expression and the other variables (sex, age, clinical stage, tumor size, distant metastasis, and OS), we performed Spearman correlation test, Chi-square Tests and Wilcoxon/Kruskal-Wallis test. We strati ed the cohort into patients with a high or low median NEIL3 expression value or median risk score as the cut-off value. Multivariate Cox regression analysis was conducted to identify independent prognostic factors. Values of p < 0.05 were considered as statistically signi cant.

NEIL3 overexpression in LUAD
We thoroughly screened the LUAD data in the GEO database and selected three miRNA-sequencing datasets: GSE30219, GSE43458, and GSE33532. We identi ed 502 common DEGs among them (Fig. 1a), including 126 up-regulated DEGs in cancer tissues (Fig. 1b). After an extensive literature review, we found that NEIL3 is a very promising DNA repair gene, but few studies reported its role in the development and mechanism of cancer. TIMER data showed that NEIL3 expression increased in 19 kinds of tumor tissues compared to adjacent normal tissues, especially in LUAD (Fig. 1c).
Analysis of the NEIL3 gene expression data in the TCGA and GSE31210 cohorts con rmed that NEIL3 is overexpressed in LUAD tissue (Fig. 2a, b). Therefore, RT-qPCR and western blotting were veri ed that the NEIL3 protein and mRNA levels were up-regulated in 22 LUAD tissues compared with the matched normal tissues (Fig. 2c, d). NEIL3 staining was positive in LUAD tissues but negative in normal lung tissue after immunohistochemistry staining (Fig. 2e). The results consistently showed that NEIL3 was obviously overexpressed in LUAD compared with the matched normal tissues (p < 0.001) (Fig. 2).
Correlation between NEIL3 expression and clinicopathological characteristics of LUAD patients The characteristics of our 406 LUAD cohort subjects and 293 TCGA LUAD cohort subjects are summarized in Tables 1 and 2. All samples were divided into low and high expression groups by IHCscore and median expression level. The results revealed that NEIL3 overexpression was closely associated with advanced clinical stage (p < 0.05; Fig. 3a) and larger tumor size (p < 0.05; Fig. 3b, c). No signi cant correlation was noted between NEIL3 expression and other clinicopathological characteristics. Previous research has proved that NEIL3 could protect telomere integrity during the S phase and accurately segregate chromosomes in actively dividing cells, which may explain the advanced T classi cation in NEIL3-overexpressing patients.

NEIL3 overexpression predicts poor prognosis in LUAD patients
This study included all follow-up survival information of the GSE50081, GSE31210, TCGA datasets and our cohort. As is shown in Fig. 3d-g, Kaplan-Meier curve and log-rank test analyses indicated that upregulated NEIL3 expression signi cantly reduced the OS and RFS of LUAD patients (p < 0.05). All risk factors and NEIL3 expression levels were subjected in the univariate Cox regression analysis (  (Table 3 and Fig. 3h).   NEIL3 expression is associated with immune cell in ltration in LUAD Patients with the same histological type of cancer may have different degree of immune in ltration cells that lead to diverse clinical outcomes [22,25]. The fact that an increased number of tumor-in ltrating lymphocytes in primary tumor tissue relates to good prognosis has been reported in several cancers, including LUAD [26]. The TIMER result showed that NEIL3 expression had a signi cant negative correlation with the in ltration of B cells, CD4 + T cells, and DCs (p < 0.05; Fig. 4a). The TIMER "Survival" module showed that high in ltrating levels of B cells bene t OS in contrast to NEIL3 expression (p < 0.05; Fig. 4b). We speculated that NEIL3 overexpression could affect OS by regulating the degree of B-cell in ltration in LUAD.
The ImmuCellAI analysis showed that the frequencies of natural T regulatory(nTreg), induced T regulatory(iTreg), Th1 cells, exhausted T cells, NK cells, effector memory and Gamma delta T cells exhibited a positive correlation with NEIL3 expression (p < 0.001), whereas the proportion of CD4 naïve cells, Tregulatory1 (Tr1), Th17, follicular helper T cell (Tfh), NK T, DC, and CD4 T cells exhibited a negative correlation (p < 0.001) (Fig. 4c). This funding suggested that NEIL3 has a regulatory effect on the formation of the LUAD immune microenvironment, especially on T-cell subsets, NK cells, and DCs. Figure 4d shows the correlation between different types of immune cell subsets. Exhausted T cells had the strongest positive correlation with Cytotoxic T lymphocyte (Pearson correlation = 0.73), while Cytotoxic T lymphocyte had the strongest negative correlation with CD4 naïve cells (Pearson correlation = -0.66). Taken together, these ndings indicate that NEIL3 plays a key role in the regulation of immunein ltrating cells in LUAD.

KEGG and GO enrichment analysis of NEIL3 and co-expressed genes in LUAD
According to FDR < 0.050 and normalized enrichment score, the GSEA analysis revealed that the pathways of the cell cycle, nucleotide excision repair, DNA replication, mismatch repair, and the P53 signaling pathway were enriched in the high NEIL3 expression phenotype (Fig. 5a). The asthma and aldosterone-regulated sodium reabsorption pathways were enriched in the low NEIL3 expression cohort (Fig. 5a). It is worth noting that the cell cycle pathway, playing a crucial part in tumorigenesis and development, is associated with NEIL3 expression.
Subsequently, we analyzed mRNA sequencing data of TCGA cohort and acquired 370 genes co-expressed with NEIL3 (Supplementary Table 1, absolute Pearson correlation coe cient > 0.5, p < 0.001). As showed in the protein-protein interaction network (PPI), there were 323 genes (yellow dots) positively related to NEIL3, versus 47 genes (blue dots) negatively related to NEIL3 (Fig. 5b). Next, we performed GO and KEGG enrichment analyses and displayed the top 30 items. The GO analysis indicated that the coexpressed genes were signi cantly enriched in nuclear cell cycle DNA replication, DNA strand elongation, and DNA replication initiation in the biological process group; condensed chromosome outer kinetochore and spindle midzone in the cellular component group; 3'−5' DNA helicase activity in the molecular function group (Fig. 5c). Similarly, the KEGG pathway enrichment analyses showed that these genes were mainly enriched in the cell cycle, oocyte meiosis and DNA replication (Fig. 5d). Taking 370 co-expressed genes into analysis via the Cytoscape software cytoHubba plugin, we ltered 10 hub genes according to node degree: TOP2A, CCNA2, BUB1B, CDC45, BUB1, CDK1, NCAPG, KIF23, UBE2C, and CCNB2 (Fig. 5e).
Several hub genes (CDK1, CCNA2, CDC45, and CCNB2) played a vital role in cell cycle progression, whereas genomic instability contributed to potentiate tumorigenesis [27]. Establishment and evaluation of prognostic signature for LUAD patients Based on the TCGA LUAD cohorts, we veri ed that all 10 of hub genes were conspicuously overexpressed in LUAD tissues versus normal lung tissues (p < 0.05; Supplementary Fig. 1). In addition, the high expressions of nine hub genes (all but TOP2A) were remarkably associated with poor prognosis in LUAD patients (p < 0.05; Supplementary Fig. 2).
Next, we attempted to establish a prognostic signature based on the expressions of NEIL3 and the other nine hub genes. According to the median risk score (cut-off = 0.926), the LUAD patients were divided into two groups with discrete clinical outcomes for OS. Figure 6a shows the distribution of risk scores in the LUAD dataset. More deaths occurred in the high risk-score group than in the low risk score group (Fig. 6b). Meanwhile, the expressions of these 10 genes were up-regulated in the high-risk group (Fig. 6c). Figure 6D shows that patients in the low-risk group had signi cantly better OS than others in the Kaplan-Meier analysis (p < 0.05). Analysis of the association between risk core and various clinical features revealed that an increased risk score was correlated with more advanced stage, the larger tumor size and poor outcome (Fig. 6e, f, g).
The univariate and multivariate Cox analyses revealed that the risk score of a 10-gene signature was an independent prognostic factor for LUAD patients (p < 0.05, Fig. 7a, b). After ROC analysis, we observed a marked predictive advantage of the risk-score signature (area under the curve [AUC] = 0.679), stage (AUC = 0.739) and lymph node metastasis (AUC = 0.680) (Fig. 7c).
Finally, we selected a test cohort based on GSE50081(127 patients with stage I & II disease). In line with the TCGA cohort, the risk score of the 10-gene prognostic signature could be an independent prognostic factor for LUAD patients (p < 0.001; Fig. 7e). Patients in the high risk score group had poor OS (p < 0.001; Fig. 6d), and the AUC values of the 1-, 3-, and 5-year ROC curves were 0.676, 0.788, and 0.766, respectively (p < 0.001; Fig. 7f). This evidences strongly suggests that the 10-gene prognostic signature has superior diagnostic accuracy and may bene t LUAD patients in the early stage.

Discussion
LUAD, the most common type of malignant tumors, has signi cant morbidity and mortality rates [28]. Growing scienti c research has focused on searching for effective treatment methods and sensitive biomarkers to improve the 5-year survival rate and life quality of LUAD patients. NEIL3, a DNA glycosylase of the BER pathway, repairs telomere oxidative damage and protects telomere integrity in cells with a high proliferative capacity during the S phase, which may explain the advanced T classi cation in NEIL3 overexpressing patients [17]. In addition, some studies indicated NEIL3 as a cell cycle dependent gene was regulated by BRG1 in breast cancer cells and that NEIL3 overexpression may facilitate distant metastasis in primary melanoma [15,19]. As mentioned in the literature review, NEIL3 plays a crucial role in preventing autoimmunity and cell proliferation [11,13]. Here a comprehensive bioinformatics analysis was performed based on gene transcript pro les of LUAD from TCGA and GEO databases. Combined with the IHC-scores of our 406 patients cohort, NEIL3 expression was up-regulated in LUAD tissues and correlated with clinicopathological characteristics, especially advanced clinical stage and large tumor size. Meanwhile, the Cox regression analysis results demonstrated that NEIL3 may serve as an independent prognostic predictor in LUAD patients.
Over the past decade, immunotherapy has been a well-known promising cancer treatment with amazing achievements in treating various refractory malignancies. The pivotal strategy of immunotherapy is to interfere with immune checkpoints expressed in immune cells [29]. Immune cells are a crucial part of the immune microenvironment, including tumor in ltrating lymphocytes (TILs), tumor-associated macrophages (TAM), dendritic cells (DC), and myeloid-derived suppressor cells (MDSCs) [30,31]. As reported in the immunoediting theory, tumor invasive immune cells (TIICs) play a "double-edged sword" role in the development of cancers. A large number of studies have found the occurrence and development of lung adenocarcinoma not only depends on the lung cancer cells themselves but also is regulated by the tumor-in ltrating immune cells in the lung cancer microenvironment [32].
Based on the TIMER database review, NEIL3 expression had a signi cant negative correlation with B cells, CD4 + T cells, and DCs. The ImmuCellAI analysis revealed nTreg, iTreg, and Exhausted T cells were increased in the high NEIL3 expression group, whereas Th17 cells, DCs and CD4 + T cells were decreased. Th17 cells play a contradictory role in tumorigenesis and might be associated with secreting cytokines such as IL-17A, IL-17F, IL-21, and IFN-γ + Th17 cells as well as effector lymphocytes, including Th1, Tc1, and NK cells [33][34][35]. Treg cells could inhibit the activation of T lymphocytes by secreting cytokines such as IL-4, IL-10, and TGF-β, which regulate the immune function of tumor patients and promote the proliferation of lung cancer cells [36]. Evidence indicates that DCs induced antitumor immunity and inhibit the formation of new blood vessels in tumors in the lung cancer microenvironment [37]. These ndings suggest that NEIL3 overexpression could increase the proportion of T regulatory cells and inhibit the antitumor function of Th17 cells and DCs, which predicted a poor OS and advanced clinical stage.
Together this evidence demonstrates that NEIL3 plays a pivotal role in the regulation of immune in ltrating cells and could be considered a novel immune-related therapeutic target in LUAD.
Equally important, GSEA analysis showed that the cell cycle and P53 signaling pathway were enriched in cases of high NEIL3 expression. As we all know, cell cycle proteins dysregulations is related to uncontrolled proliferation of a malignant tumor and becomes an attractive target in cancer therapy [38].
CDK1, a cyclin-dependent kinase, plays an important part in regulating cell cycle progression and reportedly increases cellular proliferation in various cancers if dysregulated [39]. CCNA2 and CCNB2 are cyclin family proteins. CCNA2 was recognized as an effective prognostic marker in prostate, colon, lung, and liver cancers [40,41]. In line with CCNA2, CCNB2 increased the risk of multiple cancer prognoses such as such as adrenocortical carcinoma [42], lung cancer [43], breast cancer [44], and colorectal adenocarcinoma [45]. Our result showed that NEIL3 and nine hub genes were in a tight co-expressed relationship. Combining the expression levels of the nine hub genes and NEIL3, a 10-gene prognostic signature was constructed to accurately predict the prognostic risk of LUAD patients. Multivariate Cox regression analyses proved that the risk score was an independent prognostic factor correlated with clinical stage and tumor size. Analysis of the test cohort data achieved the same results. In recent years, scholars have performed a great many studies to explore new predictive biomarkers. One type of effective biomarkers was constructed by several prognostic genes through bioinformatics analysis. This 10-gene signature could effectively predict the prognostic risk of LUAD patients and provide a theoretical basis for the development of new targeted therapies.
NEIL3 expression and the 10-gene signature were proven to be independent predictors for LUAD patients using bioinformatics technology in this study. However, identifying NEIL3 in LUAD cell lines, exploring the underlying tumorigenesis mechanism, and designing novel potent selective NEIL3-targeted drugs require further research.

Conclusion
In summary, this study is the rst to reveal that NEIL3 is overexpressed in LUAD and identify it as a new diagnostic biomarker for LUAD patients. Increased NEIL3 expression was related to advanced stage and larger tumor size as an independent diagnostic factor of poor prognosis in LUAD patients. Moreover, NEIL3 may play an important role in regulating immune-in ltrating cells such as regulatory T cells and DCs and may affect LUAD cell proliferations. The cell cycle and P53 signaling pathway is the major pathway affected by NEIL3 in LUAD. And nally, we constructed a 10-gene prognostic signature (NEIL3 and nine co-expressed hub genes) to accurately predict prognostic risk of LUAD patients. The results based on the test cohort proved that the 10-gene signature had precise diagnostic accuracy. Meanwhile, NEIL3 could be a promising biomarker for diagnosis and treatment and correlates with immune in ltration in LUAD. Availability of data and materials

Ethics approval and consent to participate
The studies were approved and consented to by the Ethics Committee of the A liated Hospital of Nantong University. The patients provided their written informed consent to participate in this research.

Consent for publication
Not applicable.