Lung cancer is regarded as a leading cause of cancer-related deaths across the globe (Malhotra et al. 2016). The pathogenesis of LUAD is not well understood and effective therapeutic approaches are lacking (Bade et al. 2020). In comparison to the single clinical biomarker, the integration of multiple biomarkers into one model improves predictive accuracy and facilitates the development of individualized treatment plans. Recently, second-generation sequencing technology has revolutionized the prognostic prediction of cancer (Sears and Mazzone 2020; Hirsch et al. 2017). In routine clinical practice, pathological staging is an important prognostic determinant of LUAD. However, many differences were observed in the clinical findings of same-stage patients, suggesting that traditional staging systems are unable to adequately predict patient outcomes. Hence, there is a considerable need to identify and develop biomarkers related to tumor diagnosis and prognosis.

Mitochondria are not only energy factories but are also involved in cell growth, differentiation, senescence, and different cell death processes (Bock and Tait 2020). In the past few years, many studies have highlighted the involvement of mitochondrial metabolism and mitochondrial dysfunction in the development of cancer (Srinivasan et al. 2017; Porporato et al. 2018). LncRNAs were recognized as effective biomarkers that were involved in the initiation and progression of different tumors like LUAD (Zhao et al. 2018b). Song et al. found that the mitochondria-related lncRNA, i.e., MDL1, controls nuclear gene expression by regulating the subcellular localization of the transcription factor (p53 protein), thus leading to a retrograde regulation of nuclear gene expression by the mitochondria (Li et al. 2022). LncRNAs are an emerging biomarker and a current research hotspot for non-coding RNAs, which are seen to be involved in cell life activities and play critical roles (Zhao et al. 2018b; Wei et al. 2020). LncRNAs can participate in cellular biological functions by exerting their endogenous “Mirna sponge” function and RNA target interaction (Batista and Chang 2013). Studies in the past have confirmed that the lncRNA expression was significantly up- or down-regulated and was involved in several malignant biological processes like invasion, proliferation, and apoptosis of tumor cells, and was also closely associated with drug sensitivity (Huang et al. 2017; Yang et al. 2022; Li et al. 2018). Some studies have identified a group of lncRNAs that could help in differentiating between the early LUAD tissue and normal lung tissue with high sensitivity and specificity. It is suggested that abnormally expressed lncRNAs may be used as a potential biomarker for diagnosing early-stage LUAD patients (Wang et al. 2015). Nevertheless, it is uncertain whether mitochondria-related lncRNA could be utilized to anticipate the prognosis of LUAD patients. In this study, a novel predictive model for LUAD was developed that displayed a better patient survival probability rate by screening for mitochondria-related lncRNAs.

In this study, 147 mitochondria-related genes and 2175 mitochondria-related lncRNAs were identified. Subsequently, 863 mitochondria-associated lncRNAs were identified by differential expression analysis. LUAD patients with complete clinical data were randomly classified into two different sets, i.e., training and testing sets. Then, univariate regression analysis was carried out in this study for identifying 39 mitochondria-related lncRNAs in the training set. Furthermore, LASSO regression was used for dimensionality reduction of the data to prevent overfitting. This yielded 23 mitochondria-related lncRNAs that were closely associated with OS in LUAD patients. Finally, 13 mitochondria-related lncRNAs were detected by performing multivariate Cox regression analysis, and a novel prognostic model was constructed. Some of these mitochondria-related lncRNAs have been reported in the past and were seen to be closely related to tumor initiation and progression. In their study, Wu et al. noted that AC092168.2 was a member of the immune-related-lncRNA prognostic signature of LUAD (Wu et al. 2021). Additionally, several studies revealed that AC026355.2 was engaged in a variety of processes, including immunomodulation, autophagy, pyroptosis, necroptosis, etc., which may have contributed to the onset and progression of LUAD (Lu et al. 2022; He et al. 2021; Liu et al. 2022; Gong et al. 2022). However, ZNF571-AS1 was first reported in solid tumors, and earlier studies have highlighted its role in different diseases such as dilated cardiomyopathy, acute myeloid leukemia, and Alzheimer's disease (Chen et al. 2021; Pan et al. 2017; Li et al. 2022). However, their specific mechanism of action in LUAD is not fully understood and needs further investigation. Very few studies have investigated the involvement of the remaining 10-mitochondria-related lncRNAs, and, although little is known about them, their importance should not be underestimated.

Furthermore, a prognostic model was developed to predict the survival of LUAD patients based on 13 mitochondria-related lncRNAs, which could be independently used as a prognostic marker for LUAD. The model accurately classifies the LUAD patients into two risk groups: low-risk and high-risk groups. The low-risk patients showed a better outcome in the training, testing, and entire sets. As shown in the results, both risk groups presented significant differences in their survival curves. The findings indicated that the high-risk score patients showed higher mortality rates in the training, testing, and entire sets. Furthermore, it was noted that an increase in the risk score led to a subsequent increase in the probability of patient death and a decrease in their survival time. On the other hand, low-risk patients showed a longer survival time. The training set data could therefore be used to construct a model based on 13 mitochondria-related lncRNAs, and it could accurately identify the prognosis of the patients and display good predictive power for the prognosis of LUAD patients.

In clinical practice, the tumor stage is determined by the tumor size, node, and metastasis (TNM) method, which is generally used for evaluating the prognosis of tumor patients. With the advent of precision medicine, an increasing number of studies have suggested that lncRNAs could have some predictive significance for tumor prognosis (Bhan et al. 2017; Lv et al. 2022; Yan et al. 2021). Therefore, the findings revealed that the prognostic model constructed by combining lncRNAs with tumor stage showed a higher predictive accuracy compared to the currently available methods for predicting prognosis. Nomograms are an easy-to-understand prognostic prediction model that is easy to operate, presents a high prediction accuracy, and is increasingly being used in medical research and clinical practice (Balachandran et al. 2015). In this study, a nomogram was constructed by integrating several clinical factors like age, TNM stage, and 13-mitochondria-associated lncRNA models based on the independent predictors generated by multivariate regression analysis. The ROC curves showed that nomograms were better than the individual prognostic factors (such as tumor stage) in predicting the disease prognosis. These results suggest that nomograms constructed based on the 13-mitochondria-associated lncRNA model may have more reliable clinical applicability and improve the accuracy of predicting prognosis.

Owing to differences in the genetic composition of the patients, targeted therapy can become an accurate and personalized treatment strategy. Identification of the molecular pathways related to LUAD could help in discovering new therapeutic targets. Therefore, GSEA was employed to screen for signaling pathways associated with 13 mitochondria-related lncRNA models, and the results indicated that these signaling pathways were associated with several cellular life activities. The P53 signaling pathway is closely associated with apoptosis and progression of the lung cancer cells (Wang, et al. 2019) and it is seen to regulate the immune responses (Muñoz-Fontela et al. 2016). Hence, the 13-mitochondria-associated lncRNA model that was constructed in this study was seen to be involved in a few cancer-related signaling pathways.

The lack of knowledge regarding the complexity, heterogeneity, and immune evasion mechanisms of tumors is one of the main challenges that affect the development of immunotherapy strategies for LUAD patients. In addition, there is a lack of specific biomarkers used to assess the benefit of tumor immunotherapy. Consequently, it is crucial to identify new immunotherapy targets and prognostic markers (Galon et al. 2014). The combination of immunosuppressive drugs can be regarded as an effective approach for treating several malignancies, and the activated TME is associated with a good response to immune checkpoint inhibitors (Yang 2015; Bersuker et al. 2019). The low-risk patients expressed high levels of the eight immunological checkpoints determined in this study, suggesting that these patients could be more benefitted by the use of immunosuppressive drugs. It was further indicated that the prognostic model developed in this study could help in predicting the effectiveness of immunosuppressive therapy.

Molecular targeted therapy has improved the treatment of lung cancer. The first molecularly-targeted therapeutic drug, i.e., gefitinib, has increased the survival duration of NSCLC patients by two times (Sun et al. 2020). Gefitinib and erlotinib, which are seen to be two small-molecule first-generation EGFR tyrosine kinase inhibitors (EGFR-TKI), were approved more than a decade ago and have been popularly used as a first-line treatment option for advanced NSCLC (Gelatti et al. 2019). The IC50 values of targeted drugs and chemotherapeutic drugs such as gefitinib, erlotinib, cisplatin, etoposide, paclitaxel, docetaxel, and Gemcitabine were analyzed using the pRRophetic tool. The findings suggest that high-risk patients may be more sensitive to chemotherapy and targeted therapy and could be used to determine the efficacy of targeted therapy in LUAD patients. These findings have potential implications for guiding the treatment and prognostic assessment of LUAD patients and may help in describing the relationship between the model and the drug to more accurately guide subsequent targeted drug therapy.

In conclusion, our study was a retrospective analysis based on the TCGA data set. It used a single sample source and could display a probable bias in the analysis results. This study on mitochondria-related lncRNAs lacks clinical and experimental validation, but future experiments would be conducted on gene expression and lncRNA function. Despite the aforementioned drawbacks, developing LUAD prediction models based on mitochondria-related lncRNAs may help in predicting the OS of LUAD more accurately than conventional pathological staging. This model could help in screening the high-risk LUAD population, and provide important references for the individualized treatment of all identified people. This would improve the research direction and offer a theoretical basis for follow-up clinical work and experimental development.