ROS is produced in many cellular compartments including mitochondria, which are the major source of ROS (mROS) . Superoxide anion (•O2-), hydrogen peroxide (H2O2) and hydroxyl radical (•OH) belong to a group of highly reactive and heterogeneous molecules derived from oxygen (O2) and are the main forms of ROS in biological systems. Many factors in the tumor microenvironment, including the presence of ROS, promote the progress of solid tumors. The increase of ROS level, the imbalance of redox homeostasis and the enhancement of antioxidant capacity are some of the many signs in cancer cells. Therefore, the understanding and elucidating the role of ROS in the tumor microenvironment is essential for developing new methods to combat this disease . Various tumors, including LUAD, possess high levels of ROS with abnormal metabolism and constitutive carcinogenic signals. ROS are the main effectors of DNA damage associated with cancer and is accompanied by tumor suppression [14-15]. Therefore, tumor cells adapt to the oxidative DNA damage to prevent cell destruction by regulating cell necrosis through the modification in the expression of some genes, thereby inducing the aberrant expression of signaling networks that cause tumorigenesis and metastasis .
8-hydroxyguanine is the strongest product of oxidative stress in cells, and is mostly closely related to the occurrence and development of tumors. The DNA repair gene can hydrolyze 8-hydroxyguanine in the base pool to avoid base mismatch and replacement. Once the 8-hydroxyguanine in tumor cells is hydrolyzed by the DNA repair gene, it promotes tumor cell growth. Certain protective effects lead to a malignant phenotype, poor cancer prognosis, or resistance to treatment. [17-18] In some cases, tumors up-regulate the mutagenic repair pathways to survive. Therefore, cancer cells generally rely more on repair pathways than normal cells. In addition, cancer cells often have dysfunctional redox homeostasis, and therefore once again, they rely heavily on mechanisms that repair oxidative DNA damage and inhibit enzymes that modify compounds, which can then be incorporated into genomic DNA in their unmodified form. Processes such as replication and oxidative stress provide a background for ongoing DNA damage in cancer cells and can provide a potential therapeutic window for compounds that exacerbate these processes. Such compounds can accomplish by further emphasizing replication, weakening the ability of cancer cells to handle high levels of replication or oxidative stress, or potentially inhibiting DNA repair and related processes [19-21].
Therefore, in this work, the synergistic tumorigenic effect of ROS-related genes and DNA repair genes was evaluated, and the regulatory relationship between the two groups of genes was further explored. It is important to consider whether it is better to use ROS to kill cancer cells or to inhibit the DNA repair in cancer cells to improve patient prognosis.
The expression of ROS-related genes and DNA repair genes was used to cluster TCGA tumor samples uniformly. ROS-related genes divided tumors into classes, and DNA repair genes divided tumor samples into classes. Significant differences in survival between the internal classifications were obtained by the two clustering methods, and the differentially expressed genes were further screened. Our analysis found that the samples that originally belonged to the ROS classification partial overlapped in the classification of DNA repair genes. After reclassifying the samples according to the two classifications, the prognosis of patients changed when the expression of ROS-related genes and DNA repair genes in the samples changed. Thus, our hypothesis was that ROS-related genes and DNA repair genes might have a mutual regulatory relationship, which in turn affected the occurrence and development of tumors. A total of 29 differential genes were finally identified and included 5 ROS-related genes and 24 DNA repair genes. STRING analysis of the regulatory relationship found that 3 ROS-related genes (NQO1, TXNRD1, and PRDX4) can be repaired by the DNA repair gene NEIL3 and other DNA repair genes.
A large amount of evidence showed that NQO1 has a "Janus" effect in cancer biology, playing a role in suppressing cancer and promoting tumors . NQO1 is constitutively expressed at a relatively low level in various normal tissues. Under oxidative stress, NF-E2 p45-related factor 2 (Nrf2)/Kelch-like ECH-related protein 1 (Keap1) signaling pathway can cooperate to transcribe a series of defense genes and provide cells with multiple layers of protection against carcinogenesis. These measures include the immediate elimination of ROS . The expression of NQO1 is considered as a practical and economical way to control cancer. NQO1 is abnormally up-regulated in solid tumors, and high levels of NQO1 are associated with poor patient prognosis. It is known that cancer cells have a significant increase in ROS production compared to normal cells. In this case, high levels of NQO1 in cancer can help cancer cells to cope with the increased ROS just like normal cells, thus, tumor growth and metastasis is not only not compromised, but promoted . Our results showed that NQO1 was correlated with the expression of the DNA repair gene NEIL3 (Pearson correlation coefficient), suggesting its role as a tumor control gene
The cytoplasmic selenoprotein thioredoxin reductase 1 (TXNRD1) has several different effects related to cancer including the protection of normal cells to evolve into cancer cells or the protection against the promotion of cancer progression. TXNRD1 has a unique connection with Nrf2 signaling and ribonucleotide reductase-dependent deoxyribonucleotide production and it supports a variety of antioxidant systems against oxidative stress. Thus, it is essential that metabolic pathways regulated by TrxR1 are affected in cancer . Our regulatory network suggested that TXNRD1 had a significant correlation with the DNA repair gene NEIL3, thus, it might be considered as a potential targeted gene in a combination therapy affecting ROS-related genes and DNA repair genes.
Peroxiredoxin 4 is a typical peroxidase 2-Cys antioxidant in the endoplasmic reticulum, which protect cells against oxidative stress by detoxifying hydrogen peroxide, thus promoting cell survival . The role of PRDX4 in cancer received considerable attention. The expression of PRDX4 in NSCLC-derived endothelial cells is higher than that in normal cells . Sulfiredoxin is an antioxidant protein induced by H2O2 that acts as a catalyst for reducing the peroxidized PRDXs to reduce their peroxidase activity. Sulfiredoxin is more inclined to combine with PRDX4 than other PRDXs. The up-regulation or down-regulation of the sulfiredoxin-PRDX4 axis can affect the mitogen-activated protein kinase pathway, cAMP response element binding protein and activator protein-1/matrix metalloproteinase axis pathway . Furthermore, another study revealed that the expression of PRDX4 is closely related to the disease-free survival time and short recurrence time of patients with early-stage lung squamous cell carcinoma undergoing early radical surgery. 
Endonuclease VIII-like 3 (NEIL3) is a DNA glycosylase protein that is involved in oxidative and interstrand crosslink DNA damage repair.  NEIL3 is highly expressed in various human cancer cells and is associated with metastatic cancer, indicating that it may be necessary to maintain cancer cell growth or malignant progression. [31,32] NEIL3 overexpression is positively correlated with homologous recombination and mismatch repair gene expression. High NEIL3 expression may promote cancer phenotype by increasing genomic instability and/or interfering with other DNA repair . Our analysis found that NEIL3 played a pivotal role in the connection between DNA repair genes and ROS-related genes. Therefore, the mutual regulation of ROS-related genes and DNA repair genes centered on NEIL3 might become an important topic for further studies.
A prognostic model based on all differentially expressed ROS-related genes and DNA repair genes was constructed and combined with the clinical data of the samples, and finally 9 genes were selected to calculate the risk score. The results revealed that the prognosis of patients in the high- and low-risk groups was significantly different, and the GEO data verified this result. The multivariate analysis suggested that the risk score could be used as an independent prognostic factor to evaluate patient prognosis. The above mentioned model genes included 3 ROS-related genes and 6 DNA repair genes, and TXNRD1 gene played an important role in the regulatory network of the two groups of genes, as revealed by previous studies.
Thus, this study might highlight the significance of ROS-related genes and DNA repair genes in LUAD, and the combined target of ROS and DNA repair genes might be a promising strategy in the treatment of LUAD, although further studies should be performed to validate these findings.