Machine learning-based lung adenocarcinoma survival prognosis model method for constructing ferroptosis-immune feature score and immunotherapy application of lung adenocarcinoma survival prognosis model method

Abstract

The invention discloses a method for constructing a lung adenocarcinoma survival prognosis model based on ferroptosis-immune feature scoring based on machine learning and immunotherapy application thereof, after the expression levels of NR5A2 and LIFR are detected, the scoring result is calculated through (NR5A2 * 0.04531979) + (LIFR *-0.007303573), patients are divided into a low-risk group and a high-risk group according to the scoring result, verification is carried out on a GEO data set, and the survival prognosis model of the lung adenocarcinoma based on ferroptosis-immune feature scoring is obtained. The result shows that the patient with high score has poor prognosis. Compared with the consensus clustering analysis, the grouping of the model can more accurately predict the prognosis of the patient and analyze the prepotency of the immune-related reaction in LUAD, which is better than the individual ferroptosis and immunization, so that the LIFR and NR5A2 are determined as promising prognosis biomarkers in LUAD, and the deep research of the ferroptosis-immune mechanism in LUAD is facilitated.

Description

Technical Field

[0001] The present invention belongs to the technical field of lung adenocarcinoma prognosis, and specifically relates to a method for constructing an iron death-immune signature score-based lung adenocarcinoma survival prognosis model by machine learning and its application in immunotherapy. Background Art

[0002] Lung cancer is the most commonly diagnosed cancer globally. It is an extremely prevalent disease, with over 2 million new cases detected globally each year. Lung cancer is typically classified into small cell and non-small cell histological types; non-small cell lung cancer (NSCLC) accounts for more than 85% of all cases and can be further classified by histological subtype. The most common histological subtype in NSCLC is adenocarcinoma, accounting for 40% of cases. Adenocarcinoma surpassed squamous cell carcinoma in the 1990s to become the most common subtype and is also the most prevalent subtype in women and non-smokers. Despite progress in prevention, detection, and treatment, the prognosis of patients with lung adenocarcinoma (LUAD) remains poor, and the overall survival rate of LUAD is still unsatisfactory. A large number of studies have investigated the key determinants of lung cancer development and progression, focusing on processes such as endocytosis, the dynamics of the immune microenvironment, and the complexity of cellular heterogeneity. Therefore, there is an urgent need to develop new lung cancer treatment strategies, with the ultimate goal of improving treatment outcomes and enhancing patients' chances of survival.

[0003] Recent clinical and preclinical studies have begun to reveal a series of systemic immune disruptions that occur during tumor development and the key contribution of peripheral immune cells to the anti-cancer immune response. Since patients with refractory malignancies, including lung cancer, benefit greatly from immune checkpoint inhibitors, immunotherapy is emerging as a new treatment option for cancer patients. Therefore, immunotherapy targeting immune-related antigens generated by LUAD may be a potentially effective treatment method. However, due to the complexity of tumor heterogeneity and carcinogenic mechanisms, immunotherapy is only applicable to a limited number of patients, and there are significant individual differences in treatment effects. Evaluating the individual response of LUAD patients to immunotherapy is very important clinically because it helps to develop personalized treatment strategies to improve patient prognosis and optimize treatment efficacy.

[0004] Ferroptosis is a form of iron-dependent programmed cell death. Its general mechanism is the lipid peroxidation of unsaturated fatty acids highly expressed on the phospholipid bilayer, leading to cell death, in which ferrous ions play a catalytic role. Ferroptosis is characterized by the accumulation of reactive oxygen species (ROS), which are molecules capable of causing oxidative damage to cells. Iron plays a crucial role in the initiation and progression of this form of cell death. ROS can enhance lipid peroxidation, especially polyunsaturated fatty acids, leading to the death of tumor cells through ferroptosis. ACSL4 is an important regulator in this process for multiple cancer types. ICD is a type of cancer cell death triggered by certain treatments such as chemotherapeutic agents, oncolytic viruses, therapies, and radiotherapy. After ICD, cells release molecules called DAMPs (damage-associated molecular patterns), TAAs (tumor-associated antigens), and proinflammatory cytokines. These molecules are captured, processed, and presented to immune cells by dendritic cells and macrophages, ultimately leading to an antigen-specific immune response. Although the immunogenicity of ferroptosis has not been widely studied, preliminary evidence suggests that it may trigger an immune response by releasing DAMPs. DAMPs are signals released by dying cells that can warn the immune system of danger and may lead to an immune response against tumor cells. Modulating the TME and TIL composition by inducing immunogenic cell death (e.g., necrosis, ferroptosis, or pyroptosis) shows promise in enhancing anti-tumor immunity.

[0005] Therefore, regulating anti-tumor immunity through ferroptosis is a key aspect of cancer immunotherapy. In the field of bioinformatics, elucidating new prognostic biomarkers and developing predictive models are emerging areas of research. However, in the context of lung adenocarcinoma (LUAD), constructing a prognostic model integrating ferroptosis and immunity (FER-IMMU) remains an unexplored area.

[0006] Although targeted therapy and immunotherapy have become the main treatment options, the treatment of LUAD still faces challenges. Efficacy and drug resistance remain the focus of attention. Tumor heterogeneity hinders personalized treatment strategies, while the evolution of tumor cells affects short-term treatment efficacy and patient prognosis. There is an urgent need to explore new and more effective treatment strategies. Ferroptosis plays an important role in cancer treatment. Some recent studies have emphasized that ferroptosis is an important cell death mechanism that can trigger a strong and persistent tumor-specific immune response. Current literature shows that ferroptosis promotes cell death in various cancers, such as colorectal cancer, malignant malaria, and breast cancer. Some studies have explored ferroptosis in LUAD and the prognostic modeling of immune-related lncRNAs and miRNAs. However, it is worth noting that lncRNAs and miRNAs regulate gene expression and function by modulating the expression levels of mRNAs. As key molecules in intracellular information transmission, mRNAs play a crucial role in the immune response. It transcribes genetic information from DNA into RNA and translates it into proteins in the cytoplasm. These proteins play important roles in immune cells, including regulating the immune response, signal transduction, and cell-cell interactions. However, despite the central role of mRNAs in cellular immune activities, their specific role in the LUAD mechanism related to FER-IMMU remains poorly understood. Summary of the Invention

[0007] The purpose of the present invention is to provide a method for constructing a ferroptosis-immune feature score-based lung adenocarcinoma survival prognosis model using machine learning and its application in immunotherapy, providing a more reliable basis for predicting the prognosis and immunotherapy response of LUAD patients and for the application of individualized immunotherapy for LUAD.

[0008] To achieve the above object, the present invention adopts the following technical solutions:

[0009] Application of LIFR and NR5A2 as biomarkers in LUAD.

[0010] A lung adenocarcinoma survival prognosis model based on machine learning for constructing a ferroptosis-immune feature score is established according to the following scoring formula by detecting the expression levels of two genes, NR5A2 and LIFR:

[0011] FER-IMMU score = (NR5A2 × 0.04531979) + (LIFR × -0.007303573),

[0012] where FER-IMMU score is the scoring result, NR5A2 is the gene expression level, and LIFR is the gene expression level.

[0013] Based on the calculation results of the scoring formula, the model classifies patients into a low-risk group and a high-risk group, and the division standard line is determined according to the median after scoring (an interval value obtained by calculation) to determine the high and low expression groups.

[0014] A computer software product comprising the model for predicting the prognosis of lung adenocarcinoma through ferroptosis-immunity as described in claim 2 or 3.

[0015] A method for the prognosis of lung adenocarcinoma patients, the method comprising the following steps:

[0016] (1) Obtain the gene expression levels of NR5A2 and LIFR

[0017] (2) Input the obtained gene expression levels of NR5A2 and LIFR into the model for predicting the prognosis of lung adenocarcinoma through ferroptosis-immunity;

[0018] (3) Evaluate the prognosis of the patient according to the scoring result obtained from the model for predicting the prognosis of lung adenocarcinoma through ferroptosis-immunity or the high or low risk of the patient.

[0019] The method for constructing the model for predicting the prognosis of lung adenocarcinoma through ferroptosis-immunity includes the following steps:

[0020] (1) Data acquisition and preprocessing:

[0021] Download the RNA sequencing data and clinicopathological information of lung adenocarcinoma (LUAD) samples from The Cancer Genome Atlas (TCGA) and the Gene Expression Omnibus (GEO); use the "ComBat" tool to correct the systematic batch effects between datasets; exclude patient data with a follow-up or survival time of less than 30 days;

[0022] (2) Creation of the FER-IMMU gene list:

[0023] Combine the ferroptosis gene list and the immune gene list to create the FER-IMMU gene list; download the ferroptosis gene list from the FerrDb database and the immune gene list from the ImmPort database; remove the overlapping genes in the two lists;

[0024] (3) Survival analysis and construction of the FER-IMMU score:

[0025] Perform Kaplan-Meier survival analysis and log-rank test using "survminer";

[0026] Identify prognostic genes and independent predictors through univariate and multivariate Cox proportional hazards regression analysis; perform LASSO regression modeling analysis using the "glmnet" and "survival" software packages to determine the prognostic central genes;

[0027] Calculate the coefficients of each gene through multivariate Cox analysis and construct the FER-IMMU gene signature;

[0028] Calculate the FER-IMMU score based on the sum of gene expression and the corresponding coefficients;

[0029] (4) Somatic mutation analysis:

[0030] Analyze the somatic mutation data downloaded from the TCGA database using "maftools"; calculate the tumor mutation burden TMB;

[0031] (5) Immunotherapy response prediction and validation:

[0032] Use the TIDE scoring system to evaluate the possibility of tumor immune escape; predict the potential benefits of various cancer types' responses to immune checkpoint blockade;

[0033] (6) Model validation:

[0034] Validate the FER-IMMU score as a reliable scoring tool on the GEO dataset; compare the survival outcomes of different risk groups through Kaplan-Meier analysis and survival curves.

[0035] The beneficial effects obtained by the present invention are as follows:

[0036] The present invention reveals the differences in immune cell infiltration and functional characteristics among lung adenocarcinoma patients in different FER-IMMU score groups, and also emphasizes the important role of the FER-IMMU score in exploring the tumor immune microenvironment, predicting immunotherapy responses, and guiding clinical treatment. These findings provide a new perspective for understanding the immune escape mechanism in LUAD and lay a foundation for the development of more effective and personalized treatment methods.

[0037] Identify LIFR and NR5A2 as promising prognostic biomarkers in LUAD. This is the first study reporting genes related to FER-IMMU in LUAD and the first study applying convolutional neural networks to FER-IMMU prognostic modeling. The results show that the prognostic efficacy of FER-IMMU in LUAD is better than that of ferroptosis and immunity alone.

[0038] Construct a FER-IMMU score system to predict the prognosis and immunotherapy responses of LUAD patients. The results show that patients with a low FER-IMMU score have a better prognosis. Grouping by machine learning methods can more accurately predict the prognosis of patients and analyze immune-related responses than using consensus clustering analysis, which provides a more reliable basis for the application of individualized immunotherapy for LUAD.

[0039] Emphasized the important roles of ferroptosis and immunity in the occurrence and development of lung adenocarcinoma (LUAD). Brief Description of the Drawings

[0040] Figure 1 is the flowchart of the present invention.

[0041] Figure 2A is the Venn diagram of 80 genes related to the immune system and ferroptosis; Figure 2B is the PPI network diagram of 80 differentially expressed genes carefully constructed using the STRING database; Figure 2C is the result diagram of dividing 522 lung adenocarcinoma patients into two FER-IMMU patterns using gene expression data; Figure 2D is the difference in transcriptome profiles between two FER-IMMU patterns; Figure 2E is the result diagram of Kaplan-Meier survival analysis.

[0042] Figure 3A is the heatmap generated by the FER-IMMU cluster 1 group. Figure 3B is the heatmap generated by the FER-IMMU cluster 2 group. Figure 3C is the heatmap generated from the clinical data of two clusters, FER-IMMU cluster 1 and FER-IMMU cluster 2.

[0043] Figure 4A is the box plot and scatter plot of FER-IMMU cluster 1 and FER-IMMU cluster 2, illustrating the differential analysis of tumor mutation burden (TMB) between FER-IMMU cluster 1 and FER-IMMU cluster 2. Figure 4B is the violin plot, depicting the comparative analysis of microsatellite instability (MSI) between FER-IMMU cluster 1 and FER-IMMU cluster 2. Figure 4C are the results of gene set GO enrichment analysis of FER-IMMU cluster 1 and FER-IMMU cluster 2 respectively, including the enriched biological processes, cellular components, and molecular functions. Figure 4D-1 is the gene function enrichment analysis: biological process, cellular component, and molecular function; gene function enrichment analysis, Figure 4D-2 is the gene function enrichment analysis: biological process, cellular component, and molecular function; continuation of gene function enrichment analysis; Figure 4E quantifies the expression levels of immune checkpoint genes under different FER-IMMU molecular patterns.

[0044] Figure 5AUnivariate Cox regression analysis identified 11 prognostic FER-IMMU signature genes. Figure 5B Figure showing the results of Kaplan-Meier survival analysis of the NOX4 gene obtained from TCGA-LUAD data analysis; Figure 5C Figure showing the results of Kaplan-Meier survival analysis of the LIFR gene obtained from TCGA-LUAD data analysis; Figure 5D Figure showing the results of Kaplan-Meier survival analysis of the IL6 gene obtained from TCGA-LUAD data analysis; Figure 5E Figure showing the results of Kaplan-Meier survival analysis of the NOX5 gene obtained from TCGA-LUAD data analysis; Figure 5F The waterfall plot shows the overall mutation status of the 11 genes that make up the FER-IMMU prognostic signature.

[0045] Figure 6A Results of LASSO regression analysis of these 11 FER-IMMU prognostic genes using survival data from the TCGA-LUAD training cohort; Figure 6B Cross-validation error path plot. Helps select the optimal regularization parameter λ for the best-fitting Lasso regression model for you. Figure 6C Figure showing the results of Kaplan-Meier survival analysis of the TCGA-LUAD training set (model group); Figure 6D Figure showing the results of Kaplan-Meier survival analysis of the GEO validation set GSE68465 (validation group); Figure 6E Figure showing the results of Kaplan-Meier survival analysis of the GEO validation set GSE30219 (validation group).

[0046] Figure 7A The waterfall plot shows the overall mutation status of the genes that make up the FER-IMMU high-risk group.

[0047] Figure 7B The waterfall plot shows the overall mutation status of the genes that make up the FER-IMMU high-risk group. Figure 7C Forest plot showing the results of univariate Cox proportional hazards regression analysis, indicating that different FER-IMMU score stratifications have independent prognostic significance compared to clinical parameters. Figure 7D Forest plot showing the results of multivariate Cox proportional hazards regression analysis, indicating that different FER-IMMU score stratifications have independent prognostic significance compared to clinical parameters. Figure 7E GSEA enrichment difference plot of the KEGG pathways in the high FER-IMMU score group. Figure 7FIt is the GSEA enrichment difference map of KEGG pathways in the low FER-IMMUscore group. Figure 7G It is the GSEA enrichment difference map of GO functions of KEGG pathways in the high FER-IMMUscore group. Figure 7H It is the GSEA enrichment difference map of GO functions of KEGG pathways in the low FER-IMMUscore group.

[0048] Figure 8 A-D stratified by FER-IMMUscore analyzed the clinical and pathological characteristics of meta-LUAD TNM stage. Figure 8 E-F stratified by FER-IMMUscore analyzed the clinical and pathological characteristics of gender and age in meta-LUAD.

[0049] Figure 9A The expression differences of immune checkpoints and regulators between the high FER-IMMUscore group and the low FER-IMMUscore group were analyzed. Figure 9B -C Box plot and scatter plot illustrate the differential analysis of tumor mutational burden (TMB) between the high FER-IMMUscore group and the low FER-IMMUscore group. Figure 9D The violin plot describes the comparative analysis of microsatellite instability (MSI) between the high FER-IMMUscore group and the low FER-IMMUscore group.

[0050] Figure 10A It is a box plot; Figure 10B It is the Kaplan-Meier analysis showing the immune cell infiltration in different FER-IMMUscore groups. Figure 10C Box plot showing the immune function in different FER-IMMUscore groups; the results of Kaplan-Meier analysis in Figure 10D. Detailed implementation manners

[0051] The present invention will be described below with reference to the accompanying drawings.

[0052] RNA sequencing data of 522 lung adenocarcinoma (LUAD) samples and their corresponding clinicopathological information were downloaded from the The Cancer Genome Atlas (TCGA) database (link: https: / / portal.gdc.cancer.gov / projects / TCGA-STAD). Clinicopathological information and microarray expression analysis data of LUAD cohorts GSE68465 and GSE30219 were downloaded from the Gene Expression Omnibus (GEO).

[0053] The "ComBat" tool in the R package "sva" was used to correct the systematic batch effects between the TCGA and GEO datasets. Patients with a follow-up or survival time of less than 30 days were excluded to eliminate bias caused by loss to follow-up or perioperative death.

[0054] Based on genomic analysis, we identified 1,794 different immune-related genes from the ImmPort database. In the FerrDb database, we identified 677 ferroptosis-related genes. Then, a Venn diagram was made, and the results showed that 80 genes intersected between immunity and ferroptosis. The Venn diagram is as Figure 2A shown.

[0055] To further explore the functional interactions between these differentially expressed genes (DEGs), we used the STRING database to construct a protein-protein interaction (PPI) network ( Figure 2B ). We collected 522 lung adenocarcinoma patients from the TCGA-LUAD cohort to form a meta-LUAD cohort for FER-IMMU classification.

[0056] By applying the consensus clustering algorithm to the FER-IMMU gene expression profiles, we divided the cohort into two different molecular patterns: FER-IMMU clusters, named FER-IMMU cluster 1 and FER-IMMU cluster 2, respectively. Specifically, FER-IMMU cluster 1 included 509 patients, while FER-IMMU cluster 2 included 32 patients ( Figure 2C ). Principal component analysis (PCA) showed different transcriptomic profiles between the two clusters ( Figure 2D ). Kaplan-Meier survival analysis showed that the survival outcome of FER-IMMU cluster 1 was the worst (P = 0.467) ( Figure 2E ).

[0057] The comparison of the expression of FER-IMMU genes in tumor tissues and normal tissues in these clusters showed that the expression of these genes in FER-IMMU cluster 1 was downregulated compared with that in FER-IMMU cluster 2 ( Figure 3A -B). However, the heatmap analysis generated from the clinical data of the two clusters showed no significant differences in the clinical characteristics of the patients ( Figure 3C ).

[0058] We found no obvious differences in tumor mutation burden (TMB) and microsatellite instability (MSI) between FER-IMMU cluster 1 and FER-IMMU cluster 2.Figure 4A -C).

[0059] In our study, gene ontology (GO) enrichment analysis was performed on two clusters. In the biological process (BP) category, both clusters were enriched in genes related to the positive regulation of cytokine production and the regulation of smooth muscle cell proliferation. In the cellular component (CC) category, the clusters were mainly enriched in the NADPH oxidase complex and the oxidoreductase complex. In the molecular function (MF) category, both clusters were mostly enriched in heme binding and tetrapyrrole binding (Figure 4D).

[0060] We evaluated the expression of immune checkpoint genes in different FER-IMMU clusters, including BTNL3, CD86, CD160, KIR2DL1, and KIR2DL4. An increased expression level of these immune checkpoint genes was observed in FER-IMMU cluster 1 ( Figure 4E ).

[0061] 3. Survival analysis and construction of the FER-IMMU score

[0062] Kaplan-Meier survival analysis and log-rank test were performed using the R package "survminer". Prognostic genes and independent predictors were identified using the "survival" package through univariate and multivariate Cox proportional hazards regression analysis. Prognostic central genes were analyzed by least absolute shrinkage and selection operator (LASSO) regression modeling using the "glmnet" and "survival" packages. The coefficient of each gene was calculated by multivariate Cox analysis, and the FER-IMMU gene signature was constructed using the selected genes. The FER-IMMU score was calculated based on the sum of the product of gene expression and the corresponding coefficients. Patients were divided into high FER-IMMU score and low FER-IMMU score groups according to the optimal cut-off value determined by the "survminer" package.

[0063] The FER-IMMU model from retrospective genomic analysis of a large cohort has limitations in quantitatively predicting individual patient risk. Therefore, we developed a risk scoring system, named FER-IMMU score, to quantify the FER-IMMU clusters in patients with lung adenocarcinoma (LUAD). The TCGA-LUAD dataset, known for its high-quality data and large sample size, was selected as the training set for constructing the FER-IMMU score. Initially, univariate Cox analysis identified 11 FER-IMMU genes as prognostic indicators, and these 11 FER-IMMU genes were NOX4, DUOK1, HMOK1, IFNG, IL6, LIFR, NR4A1, CDH1, NR5A2, LCN2, and ANGPTL7 ( Figure 5A ).

[0064] Subsequent Kaplan-Meier analysis showed that high expression of NOX4, HMOK1, and IL6 was associated with poor prognosis, while high expression of LIFR was associated with better prognosis ( Figure 5B -E). Notably, the total mutation rate of these 11 genes was 11.13%, and NOX4, NR5A2, and LIFR had particularly high mutation frequencies ( Figure 5F ).

[0065] First, we performed LASSO regression analysis on 11 prognostic FER-IMMU genes using the TCGA dataset trained by LASSO regression analysis and the GEO dataset for prognostic validation by Kaplan-Meier analysis ( Figure 6A -B) to reduce the dimension of the survival data using the TCGA-LUAD training cohort.

[0066] Second, a two-gene model was developed through multivariable Cox analysis, and the FER-IMMUscore was calculated as follows: FER-IMMU score = (NR5A2 × 0.04531979) + (LIFR × -0.007303573). The "survminer" software package was used to determine the optimal cutoff value to automatically classify each dataset into a low-risk group (low FER-IMMU score) and a high-risk group (high FER-IMMU score).

[0067] Finally, Kaplan-Meier analysis of the training set showed that patients with high FER-IMMU score had significantly worse prognosis than those with low FER-IMMU score (P = 0.046, ). Validation set analyses including GSE68465 (P < 0.001, ) and GSE30219 (P < 0.027, ) confirmed the poor prognosis of patients with high FER-IMMU score.

[0068] The FER-IMMU score was validated through multiple datasets and is a reliable tool for classifying lung adenocarcinoma patients into different prognostic groups based on their genomic characteristics.

[0069] 4. Somatic Mutation Analysis

[0070] Somatic mutation data downloaded from the TCGA database (link: https: / / portal.gdc.cancer.gov / projects / TCGA-STAD) were analyzed using the R package "maftools". Tumor mutation burden (TMB) was calculated as the number of nonsynonymous somatic mutations per megabase of the cancer sample.

[0071] We investigated the somatic mutations in different FER-IMMU score groups in the TCGA cohort and found that the overall mutation rate in the high FER-IMMU score group (91.53%, Figure 7A ) was significantly higher than that in the low FER-IMMU score group (90.04%, Figure 7B ). Figure 7A And Figure 7B The waterfall plot was used to show the distribution of somatic mutations in different FER-IMMU score groups in the TCGA cohort. Through univariate and multivariate Cox regression analyses, we found that FER-IMMU score stratification had independent prognostic significance compared with clinical parameters ( Figure 7C ). Figure 7C The forest plot showed the results of univariate and multivariate Cox proportional hazards regression analyses, illustrating the independent prognostic significance of different FER-IMMU score stratifications compared with clinical parameters.

[0072] Subsequently, we performed gene set enrichment analysis (GSEA) to determine the enrichment of KEGG pathways and GO functions in the high FER-IMMU score and low FER-IMMU score groups. The high FER-IMMU score group showed enrichment in several pathways including complement and coagulation cascades, cytokine-cytokine receptor interaction, and cytochrome P450-mediated xenobiotic metabolism, as well as porphyrin and chlorophyll metabolism and primary immunodeficiency ( Figure 7D ).

[0073] This group was also enriched in biological processes such as keratinization and cellular components such as blood microparticles, cornified envelope, immunoglobulin complex, and T cell receptor complex ( Figure 7E ). In contrast, the low FER-IMMU score group was enriched in pathways such as SLE and taste transduction ( Figure 7F ).

[0074] In addition, this group was enriched in biological processes such as nucleosome organization and protein localization to CENP-A-containing chromosomes, cellular components such as chromosomal centromere core domain, and molecular functions such as structural components of chromatin ( Figure 7G ).

[0075] 5. Immunotherapy response prediction and validation

[0076] Tumor Immune Dysfunction and Rejection (TIDE) is a computational framework for assessing the likelihood of tumor immune escape from gene expression profiles in cancer samples. The TIDE score consists of two components: immune dysfunction and immune rejection, which can serve as surrogate biomarkers for predicting the response of various cancer types to immune checkpoint blockade. For more information about TIDE, please visit: http: / / tide.dfci.harvard.edu / .

[0077] In this study, we analyzed the clinicopathological characteristics of the FER-IMMU score groups in the meta lung adenocarcinoma cohort and found that high FER-IMMU score was associated with advanced TNM stage ( Figure 8 A-D). Moreover, there were no significant differences in gender and age between the two groups ( Figure 8 E-F). In summary, FER-IMMU score is correlated with the clinical characteristics of lung adenocarcinoma and can be considered an independent prognostic factor.

[0078] 6. Immune cell infiltration analysis

[0079] The CIBERSORT algorithm was used to calculate immune cell infiltration in multiple LUAD datasets.

[0080] We analyzed the differences in the expression of immune checkpoints between the high FER-IMMU score and low FER-IMMU score groups. Notably, compared with the low FER-IMMU score group, the expression levels of TDO2, NFRSF9, TNFSF4, CD27, VTCN1, CD70, and TNFRSF18 were increased in the high FER-IMMU score group. In contrast, the expressions of CEACAM1, LAG3, and TGIT were significantly downregulated in the high FER-IMMU score group ( Figure 9A ). Figure 9A The differential expressions of immune checkpoints and regulators between the high and low FER-IMMU score groups were analyzed. There were no significant differences in tumor mutation burden (TMB) and microsatellite instability (MSI) between the two groups ( Figure 9B -D). Figure 9B , Figure 9C Box plots and scatter plots illustrate the differential analysis of tumor mutation burden (TMB) between the high and low FER-IMMU score groups; Figure 9D For violin plots describe the comparative analysis of microsatellite instability (MSI) between the high and low FER-IMMU score groups.

[0081] Subsequently, we estimated the efficacy of immunotherapy in lung adenocarcinoma. We used the TIDE online tool to predict the potential benefits of immunotherapy in the TCGA-LUAD cohort. It was found that the TIDE score was positively correlated with immune escape, indicating a reduced benefit of immunotherapy and being associated with adverse survival outcomes. The data showed that compared with the low FER-IMMU score group, the TIDE and immune escape scores were higher in the high FER-IMMU score group ( Figure 9D ), suggesting that patients with a high FER-IMMU score may benefit less from immunotherapy. However, no obvious correlation was observed between the immune exclusion score and the high FER-IMMU score group ( Figure 9D ).

[0082] These results highlight the complex interactions between FER-IMMU and the tumor microenvironment, which affect the response to immunotherapy. We also examined the immune profiles of different FER-IMMU score groups, focusing on the composition of immune cell populations and their functional capabilities. We observed differences in the infiltration levels of various immune cell subsets, including plasma cells, T cell CD4 memory resting, T cell CD4 memory activated, T regulatory cells (Tregs), monocytes, and eosinophils, between the high-risk and low-risk FER-IMMU score groups. Kaplan-Meier analysis showed that in lung adenocarcinoma (LUAD), the infiltration of immune cells such as T follicular helper cells, resting CD4+ T memory cells, plasma cells, resting natural killer (NK) cells, monocytes, and resting mast cells was positively correlated with prognosis. However, a large infiltration of activated dendritic cells, neutrophils, and M0 macrophages was negatively correlated with prognosis ( Figure 10A ).

[0083] In addition, our findings also elucidated that there were significant differences in immune functions such as antigen-presenting cell co-inhibition (APC_co_inhibition), antigen-presenting cell co-stimulation (APC_co_stimulation), B cells, CCR, CD8+ T cells, checkpoint inhibition, and cytotoxic activity between the high-risk and low-risk FER-IMMU groups. In particular, the enhancement of immune functions such as B cells, CD8+ T cells, checkpoint regulation, cytotoxic activity, dendritic cells (DCs), human leukocyte antigen (HLA), immature dendritic cells (iDCs), pro-inflammatory factors, mast cells, and plasmacytoid dendritic cells (pDCs) T cell co-inhibition, T helper cells, T follicular helper cells (Tfh), tumor-infiltrating lymphocytes (TIL), and type II interferon (IFN) was positively correlated with a better prognosis in LUAD. Conversely, an increase in MHC class I expression and paracrine inflammation were associated with a poorer prognosis.Figure 10B )。

[0084] 7. Statistical analysis

[0085] Statistical analysis was performed using R software (version 4.1.0) and the Sangerbox tool (link: http: / / www.sangerbox.com / tool). Continuous variables were expressed as the standard error of the mean and compared using Student's t-test or the Wilcoxon rank-sum test. Categorical data were compared using the chi-square test. Statistical significance was defined as P < 0.05 and expressed as P < 0.05, P < 0.01, and P < 0.001.

[0086] The principle of the present invention is as follows:

[0087] First, we identified a group of FER-IMMU-related genes that are closely related to patient survival and prognosis. Then, we divided them into two groups, a high-risk group and a low-risk group. We found that patients in the high-risk group had the worst prognosis. Moreover, in the high-risk group, the expression levels of most FER-IMMU genes were relatively low. This indicates that patients in the high-risk group may have a poor response to certain treatments. However, the expression levels of some immune checkpoints were elevated in patients in the high-risk group, such as BTNL3, CD86, CD160, KIR2DL1, and KIR2DL4. The disorder of these immune checkpoints may be related to changes in the immune microenvironment. These immune checkpoints may play an important role in tumor immune surveillance and escape by regulating the activities of T cells and NK cells and affecting the activation and inhibition of immune responses, but the specific mechanism still needs further study. This may provide a new way to explore the immune mechanism and develop immunotherapy. As for the low-risk group, the expression of its FER-IMMU genes was higher and the prognosis was better than that of the high-risk group, which may be closely related to the role of ferroptosis-immunity. It is possible that the high expression of these genes promotes the ferroptosis of LUAD by affecting the immune microenvironment, thereby affecting the prognosis of patients. Therefore, dividing patients into different FER-IMMU clusters may help to formulate personalized treatment strategies and improve the efficacy and survival rate. In addition, our study revealed the molecular characteristics and biological functions of FER-IMMU-related genes, and both FER-IMMU patterns were enriched in immune-related signaling pathways. This indicates that immune-based therapies may have important significance in the treatment of LUAD. However, consensus clustering analysis usually uses fixed clustering algorithms and parameter settings, which may limit its ability to adapt to data diversity. Different datasets may require different clustering strategies, and consensus clustering analysis is not flexible enough to meet these needs. Therefore, this may lead to the conclusion drawn by consensus clustering analysis being deviated from the actual situation.

[0088] To address this issue, we developed and validated an individual risk assessment tool for LUAD, the FER-IMMU score, through machine learning methods.

[0089] First, we identified 11 FER-IMMU genes associated with prognosis (NOX4, DUOK1, HMOK1, IFNG, IL6, LIFR, NR4A1, CDH1, NR5A2, LCN2, and ANGPTL7) through univariate Cox analysis.

[0090] Subsequently, we performed dimensionality reduction on these genes using LASSO regression analysis and constructed a signature consisting of two genes (LIFR and NR5A2). We developed the following scoring formula: FER-IMMU score = NR5A2 × 0.04531979 + (LIFR × -0.007303573). The calculation of the FER-IMMU score can quickly and easily identify the risk level of patients. This is consistent with previous literature. The orphan nuclear receptor NR5A2 plays a novel regulatory role as a transcription factor and stemness regulator in lung cancer stem cells (CSCs) and lung tumorigenesis, and the elevation of its transcriptional activation and Nanog expression levels shows diagnostic and prognostic value in human lung cancer.

[0091] However, Quan et al. demonstrated that NR5A2 plays dual roles in pancreatic ductal adenocarcinoma (PDAC). On the one hand, in differentiated cancer cells, NR5A2 promotes cell proliferation by inhibiting CDKN1A. On the other hand, in the CSC population, NR5A2 enhances stemness by directly binding to the promoter / enhancer region of SOX2, thereby upregulating SOX2. These findings are consistent with ours, indicating that NR5A2 expression promotes tumorigenesis and progression, but the potential biphasic role of NR5A2 in LUAD still requires further investigation.

[0092] LIFR (leukemia inhibitory factor receptor) is a molecule that plays a key role in signal transduction. A decrease in LIFR expression has been observed to be associated with a poor prognosis in pancreatic cancer patients with KRAS gene mutations. A correlation between a decrease in LIFR expression and a shorter survival time has also been observed in KRAS-mutated non-small cell lung cancer. These findings suggest that the silencing of LIFR is a common mechanism of KRAS-mediated cell transformation.

[0093] In addition, the LIFR / STAT3 signaling pathway appears to have a dual role, potentially mediating tumor-promoting and tumor-suppressing signals depending on the genetic makeup of tumor cells. This pathway may play different roles in different cells within the tumor microenvironment. These findings suggest that LIFR may inhibit tumor development, which is consistent with our findings, but the exact mechanism still requires further investigation.

[0094] Next, our analysis of FER-IMMU gene mutations may help identify potential therapeutic targets and drug resistance mechanisms. We observed a relatively high mutation rate of the TP53, TTN, and MUC16 genes in two FER-IMMU groups, indicating that mutations in these genes may lead to various phenotypic changes, including tumorigenesis, muscle diseases, and changes related to immune therapy response and prognosis. These mutations are of great significance for disease progression and the selection of treatment strategies. For reference, the mutant groups in LUAD patients showed differences in somatic mutations, mRNA-seq, miRNA-seq, immune infiltration, and immune regulators, indicating that TP53 gene mutations play a key role in the occurrence and progression of LUAD. Similarly, gene expression heterogeneity driven by TTN mutations prolonged the survival of patients with lung adenocarcinoma and provided valuable clues for the prognosis of TTN gene mutations in lung adenocarcinoma. The abnormal expression of specific MUC proteins, especially mucin 16 (MUC16) in tumor cells, is closely related to tumorigenesis, proliferation, and metastasis. The high mutation rate of these genes indicates that they play important roles in the pathophysiological process of LUAD, making them potential candidate genes for immune-based targeted therapy or key factors explaining drug resistance mechanisms.

[0095] Therefore, our study also revealed the molecular characteristics and biological functions of FER-IMMU-related genes, and found that both FER-IMMU groups were enriched in immune-related signaling pathways, indicating that immune-based therapies may have potential relevance in LUAD treatment. Based on the calculated FER-IMMU scores, we divided the patients into low-risk and high-risk groups. Kaplan-Meier analysis showed that patients in the high-risk group had a poor prognosis, which was closely related to the clinicopathological characteristics of LUAD (e.g., advanced TNM stage). More importantly, compared with the low-risk group, the high-risk group showed higher TIDE and immune dysfunction, while the increase in immune cell infiltration and improvement of immune function may be related to the complexity of the tumor microenvironment (TME) and the phenotype and differentiation status of immune cells. Studies have shown that an immunosuppressive microenvironment rich in regulatory CD4+ T lymphocytes (Tregs) promotes the progression of LUAD. TAM-FOLR2 (tumor-associated macrophage-folate receptor 2) and CD4+NR4A3 (the third member of the CD4+ nuclear receptor subfamily 4A) were significantly increased in invasive adenocarcinoma, which may affect the recruitment and function of immune cells through the chemokine ligand / receptor of the time trajectory and the differential expression profile. There was infiltration of a large number of B cells and plasma cells (TIB) in the early LUAD tumor tissue, and they showed different phenotypes and differentiation states during tumor progression. However, according to the findings of the current study, the exact mechanism remains unclear and further research is needed. The increased expression of immune checkpoint molecules (TDO2, NFRSF9, TNFSF4, CD27, VTCN1, CD70, and TNFRSF18) in the low-risk subgroup indicates that immune checkpoint inhibitors are more likely to exert anti-tumor effects in the low-risk group. As we can see, the final results of immune checkpoints grouped by two methods, namely consensus clustering analysis and machine learning methods, are inconsistent, but the results from the machine learning method are more persuasive.

[0096] Therefore, it is reasonable to believe that grouping samples by machine learning methods is more accurate than consensus clustering analysis. Taken together, FER-IMMU responds to the complex and subtle interactions of the tumor microenvironment in regulating immune therapy responses and reveals the potential relevance of FER-IMMU in tumor immunity and immunotherapy, which will provide new ideas for subsequent research.

Claims

1. Application of LIFR and NR5A2 as biomarkers in LUAD.

2. A lung adenocarcinoma survival prognosis model based on machine learning to construct an iron death-immune feature score, characterized in that, By detecting the expression levels of the two genes NR5A2 and LIFR, it is established according to the following scoring formula: FER-IMMU score = (NR5A2 × 0.04531979) + (LIFR × -0.007303573), where FER-IMMU score is the scoring result, NR5A2 is the expression level of the NR5A2 gene, and LIFR is the expression level of the LIFR gene.

3. The lung cancer prognosis assessment system based on ferroptosis-immune related genes according to claim 2, wherein According to the calculation result of the scoring formula, the model divides patients into a low-risk group and a high-risk group, and the division standard line is determined by the median after scoring to determine the high and low expression groups.

4. A computer software product, comprising the lung cancer prognosis assessment system based on ferroptosis-immune related genes described in claim 2 or 3.

5. A method for the prognosis of patients with lung adenocarcinoma, using the lung cancer prognosis assessment system based on ferroptosis-immune related genes described in claim 2 or 3, characterized in that, The method comprises the following steps: (1) Obtain the gene expression levels of NR5A2 and LIFR; (2) Input the obtained gene expression levels of NR5A2 and LIFR into the model for predicting the prognosis of lung adenocarcinoma through ferroptosis-immunity; (3) Evaluate the prognosis of the patient according to the scoring result obtained from the model for predicting the prognosis of lung adenocarcinoma through ferroptosis-immunity or the high or low risk of the patient.

6. The method for constructing a lung cancer prognosis evaluation system based on ferroptosis-immune related genes according to claim 2 or 3, characterized in that Comprises the following steps: (1) Data acquisition and preprocessing: Download the RNA sequencing data and clinicopathological information of lung adenocarcinoma samples from The Cancer Genome Atlas and Gene Expression Omnibus databases; use the "ComBat" tool to correct the systematic batch effects between datasets; exclude patient data with follow-up or survival time less than 30 days; (2) Creation of the FER-IMMU gene list: Combine the ferroptosis gene list and the immune gene list to create the FER-IMMU gene list; download the ferroptosis gene list from the FerrDb database and the immune gene list from the ImmPort database; remove the overlapping genes in the two lists; (3) Survival analysis and construction of the FER-IMMU score: Perform Kaplan-Meier survival analysis and log-rank test using "survminer"; Identify prognostic genes and independent predictors through univariate and multivariate Cox proportional hazards regression analysis; perform LASSO regression modeling analysis using the "glmnet" and "survival" software packages to determine the prognostic central genes; Calculate the coefficient of each gene through multivariate Cox analysis to construct the FER-IMMU gene signature; Calculate the FER-IMMU score according to the sum of the gene expression and the corresponding coefficients; (4) Somatic mutation analysis: Analyze the somatic mutation data downloaded from the TCGA database using "maftools"; calculate the tumor mutation burden TMB; (5) Prediction and verification of immunotherapy response: Use the TIDE scoring system to evaluate the possibility of tumor immune escape; predict the potential benefits of various cancer types in response to immune checkpoint blockade; (6) Model validation: Validate the FER-IMMU score as a reliable scoring tool on the GEO dataset; compare the survival outcomes of different risk groups through Kaplan-Meier analysis and survival curves.

Images