Application of Machine Learning Models Based on H&E Staining for Mitochondrial-Related Genes Classification and Prognosis of Lung Adenocarcinoma

Application of Machine Learning Models Based on H&E Staining for Mitochondrial-Related Genes Classification and Prognosis of Lung Adenocarcinoma | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Application of Machine Learning Models Based on H&E Staining for Mitochondrial-Related Genes Classification and Prognosis of Lung Adenocarcinoma Xu Sijuan, Lu Jianzhong This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7252366/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Lung adenocarcinoma (LUAD) is a leading cause of cancer-related mortality worldwide, necessitating the identification of reliable prognostic biomarkers. This study innovatively combines mitochondrial-related genes classification with pathological histology. We developed machine learning models to predict LUAD prognosis with improved accuracy.We utilized RNA sequencing data from 443 LUAD patients and paired H&E-stained pathological images from 327 patients, both sourced from The Cancer Genome Atlas (TCGA) database. Using a non-negative matrix factorization (NMF) clustering algorithm, we clustered 213 mitochondrial-related genes into high-risk (Cluster 1) and low-risk (Cluster 2) categories.Survival analysis confirmed that the high-risk group is an independent risk factor for overall survival (OS) (HR = 1.463, p = 0.031). In parallel, we constructed an eight-feature pathomics model using various machine learning techniques, achieving a strong predictive performance with an area under the curve (AUC) of 0.836. The pathomics score (PS) derived from this model was identified as an independent prognostic factor (HR = 1.686, p = 0.013). Moreover, functional enrichment analysis revealed that the high-PS group is associated with several critical pathways and alterations, including activation of the G2M checkpoint pathway, upregulated lysine degradation, reduced resting dendritic cell infiltration, TP53/TTN mutations, and increased tumor mutation burden.This study represents a novel integration of mitochondrial metabolic genes classification with pathological histology, demonstrating that pathomics features indicative of mitochondrial subtypes serve as potential prognostic markers for LUAD and might suggest promising avenues for personalized treatment strategies. Health sciences/Biomarkers Biological sciences/Cancer Biological sciences/Computational biology and bioinformatics Health sciences/Oncology Prognostic model Lung adenocarcinoma Mitochondrial-related genes Pathomic Machine learning algorithm Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 1. Introduction Lung cancer, primarily known as primary bronchogenic carcinoma, originates in the bronchial or alveolar epithelium [ 1 ][ 2 ] . Among the various types of lung cancer, lung adenocarcinoma (LUAD) is categorized as a type of non-small cell lung cancer [ 3 ] .The primary treatment approach for LUAD involves surgical resection; however, the prognosis for patients diagnosed at advanced stages remains unfavorable [ 4 ] .For those with mid to late-stage lung cancer who are no longer candidates for surgery, treatment objectives shift towards prolonging life, improving quality of life, and striving for long-term survival despite the presence of tumors [ 2 ] .Traditional markers used for LUAD diagnosis include clinical pathological features, carcinoembryonic antigen, carbohydrate antigen 125, and CT pathological examination methods. Unfortunately, these markers are increasingly inadequate for the demands of precision medicine [ 5 ] , highlighting the urgent need to identify novel prognostic markers that can offer new basis for personalized precision therapy. Mitochondrial-related genes play a crucial role in essential metabolic processes such as oxidative phosphorylation, the tricarboxylic acid cycle, and fatty acid β-oxidation [ 6 ][ 7 ] , thereby maintaining the cell's energy supply and balance. Previous studies have shown that alterations in mitochondrial are intricately linked to LUAD cell growth, evasion of apoptosis, and the development of drug resistance [ 8 ][ 9 ] .Mitochondrial related genes like SLC25A4 and ALDOA exhibit high expression levels in LUAD tissues and correlate with poor patient prognosis [ 10 ] .Subtype clustering based on the expression characteristics of mitochondrial-related genes has emerged as a promising approach to understanding the molecular heterogeneity of LUAD. Recent studies have demonstrated that mitochondrial-related genes can aid in the classification of LUAD, revealing notable differences in immune status, metabolic features, and survival rates among distinct subtypes [ 11 ][ 12 ][ 13 ] .In a multi-omics integrated analysis, Zhang et al. [ 11 ] developed a Mitochondrial Pathway Signature (MitoPS) based on 149 mitochondria-related pathways. This signature stratified LUAD patients into distinct subtypes. The high-MitoPS subtype exhibited mitochondrial metabolic dysfunction and an immunosuppressive tumor microenvironment. This included reduced CD8⁺ T cell infiltration and an "immune desert" phenotype, along with upregulation of immune checkpoint molecules and poorer survival outcomes.By contrast, the low-MitoPS subtype was associated with an“ immune-activated ”phenotype and better treatment response.Furthermore, the core gene NDUFB10 was identified as a key immune regulator, whose expression level influences CD8⁺ T cell infiltration and immune checkpoint protein expression, thereby further highlighting the critical role of mitochondria in regulating the tumor immune microenvironment. [ 11 ] .Mennuni M et al. revealed that elevated mitochondrial DNA (mtDNA) levels fuel tumor proliferation—a characteristic feature aligning with highly mitochondrial subtypes and tumor growth phenotypes [ 12 ] .Notably, this transcriptional signature extends its influence to the tumor immune microenvironment. Zhang Y et al. identified the mitochondrial-related circular RNA cEMSY as enhancing dendritic cell and CD8⁺ T cell infiltration, subsequently improving immunotherapy response in lung cancer [ 13 ] . In conclusion, mitochondrial-related genes are not only crucial in the onset and progression of LUAD, but their expression patterns and subtype classifications may also serve as valuable foundations for diagnostic and therapeutic strategies. Although current research has advanced our understanding of mitochondrial typing in LUAD, no studies have precisely predicted mitochondrial genotyping based on pathomics. Furthermore, the prognostic value of such predictions in LUAD patients remains unreported. Major methods for detecting mitochondrial-related genes (e.g., qPCR, RNA-seq, mass spectrometry-based metabolomics, and mtDNA sequencing) are used to analyze genes expression, mutations, and metabolic alterations [ 14 ][ 15 ] . However, these methods have limitations such as strict sample quality demands, mtDNA degradation or contamination risks, high costs, and difficulties in data integration [ 15 ] . Deciphering the synergistic regulatory mechanisms between mitochondrial and nuclear genes presents significant challenges for biological interpretation. H&E-stained sections are core imaging data used for clinical diagnosis and form the basis of pathomics analysis [ 16 ] . Pathomics applies artificial intelligence to convert pathological images into high-throughput quantitative features, such as texture, morphology, and biological characteristics [ 17 ] . This enables tumor diagnosis, molecular expression assessment, and prognostic evaluation [ 18 ][ 19 ] .Nevertheless, its application in predicting mitochondrial subtypes remains exploratory. The potential of pathomics for accurate prediction of this subtyping lacks robust validation.The key bottleneck lies in biological interpretability: Integration of genomic, transcriptomic, and mutational data is essential to decode pathological features, making this cross-omics research critically valuable. This study proposes a novel approach to classify genes associated with mitochondrial through clustering analysis, aiming to investigate the prognostic significance of mitochondrial typing. Following this, we plan to develop a pathological model that predicts mitochondrial typing by utilizing paired H&E images alongside RNA transcriptome samples, and we will assess its prognostic value in relation to overall survival (OS). Ultimately, we will delve into the biological interpretability of the pathological model by integrating transcriptomic and mutation data. 2. Results 2.1 Construction of Prognostic Risk Model for Mitochondrial-Related Genes 2.1.1 NMF typing In the training set, the optimal number of cluster was determined as k = 2 based on the steepest decline in the cophenetic correlation coefficient. Subsequently, samples were divided into two major molecular subtypes using the unsupervised clustering method of NMF and labeled Cluster 1 and Cluster 2. (Fig. 3) 2.1.2 Prognostic Value of Mitochondrial Genotyping in LUAD A total of 443 patients with LUAD from the TCGA database were included in the survival analysis, with patients divided into the high-risk group (n = 148) and the low-risk group (n = 295). Patient clinical information is shown in Table 1.Gender, pathologic_stage, smoking_status, and radiotherapy were significantly different between the two groups ( p < 0.05). According to survival data from the Kaplan-Meier (KM) curve analysis, cluster 1 was defined as the high-risk group, and cluster 2 was defined as the low-risk group. The median overall survival for the high-risk group was 39.03 months, while that for the low-risk group was 51.03 months. The Log-rank test indicated that the difference in overall survival rates between the high and low-risk groups was statistically significant ( p < 0.05)(Fig. 4A). Table 1 The clinical characteristics of cluster Variables Total (n = 443) High-risk group (n = 148) Low-risk group (n = 295) p Age, n (%) 0.346 ~65 215 (49) 77 (52) 138 (47) 66~ 228 (51) 71 (48) 157 (53) Gender, n (%) < 0.001 Female 244 (55) 60 (41) 184 (62) Male 199 (45) 88 (59) 111 (38) Residual_tumor, n (%) 0.905 R0 297 (67) 100 (68) 197 (67) R1/R2 16 (4) 6 (4) 10 (3) RX/Unknown 130 (29) 42 (28) 88 (30) Histologic_type, n (%) 0.465 Mixed Subtype 97 (22) 32 (22) 65 (22) NOS 270 (61) 95 (64) 175 (59) Others 76 (17) 21 (14) 55 (19) Pathologic_stage, n (%) 0.018 I/II 353 (80) 108 (73) 245 (83) III/IV 90 (20) 40 (27) 50 (17) Tumor_location, n (%) 0.935 L-Lower 69 (16) 23 (16) 46 (16) L-Upper 111 (25) 39 (26) 72 (24) R-Lower 85 (19) 25 (17) 60 (20) R-Middle 20 (5) 7 (5) 13 (4) R-Upper 158 (36) 54 (36) 104 (35) Smoking_status, n (%) < 0.001 Current 105 (24) 50 (34) 55 (19) Former 271 (61) 89 (60) 182 (62) Nonsmoker 67 (15) 9 (6) 58 (20) Radiotherapy, n (%) 0.006 NO 395 (89) 123 (83) 272 (92) YES 48 (11) 25 (17) 23 (8) Chemotherapy, n (%) 0.664 NO 289 (65) 94 (64) 195 (66) YES 154 (35) 54 (36) 100 (34) OS, n (%) 0.034 Alive 289 (65) 86 (58) 203 (69) Dead 154 (35) 62 (42) 92 (31) OS.time, Median (Q1,Q3) 21.9 (14.47, 35.87) 20.72 (12.38, 35.08) 22.3 (15.05, 37.4) 0.192 The univariate COX analysis showed that the high-risk group was significant associated with worse OS (HR = 1.514, 95% CI: 1.097–2.092, p = 0.012). Additionally, Residual Tumor: R1/R2 vs R0 (HR = 4.058, 95% CI: 2.260–7.287, p < 0.001), Histologic Type: Others vs. Mixed Subtype (HR = 0.427, 95% CI: 0.224–0.814, p = 0.010), Pathologic Stage (HR = 2.560, 95% CI: 1.822–3.597, p < 0.001), and Radiotherapy (HR = 1.67, 95% CI: 1.077–2.589, p = 0.022) were also significant associated with OS. In multivariate analysis, after adjustment, the high-risk group (HR = 1.463, 95% CI: 1.036–2.065, p = 0.031) remained a significant risk factor for OS. Moreover, Residual Tumor: R1/R2 vs R0 (HR = 4.343, 95% CI: 2.182–8.643, p < 0.001) and Pathologic Stage (HR = 2.472, 95% CI: 1.682–3.632, p < 0.001) were identified as independent risk factors for OS( Fig. 4B). In the subgroup analysis, interaction tests revealed no significant interactions between the risk group and various subgroups including age, sex, postoperative pathological stage of residual tumor, radiotherapy, chemotherapy, smoking status, residual tumor, pathological subtypes, and tumor location subgroups (all interaction p- values > 0.05). This indicates that the impact of the high-risk group on overall survival did not significantly differ across these subgroups (Fig. 4C). These results indicate the potential clinical significance of mitochondrial subtyping in the prognosis of LUAD patients. 2.2 Developing a Mitochondrial Genotyping-Associated Pathomics Model 2.2.1 Construction and Evaluation of a Mitochondrial Genotyping-Integrated Pathomics Model A total of 327 samples with complete pathological images were divided into a training set (n = 229) and a validation set (n = 98). Patients' baseline clinicopathological characteristics of the patients are presented in Table 2. Feature selection was first performed using the mRMR method,which selected the top 20 most relevant features. Then, recursive feature elimination (RFE)was performed, as shown in the feature selection diagram, resulting in 8 selected features (Fig. 5A).The relative importance of these selected features in the XGBoost algorithm is demonstrated in Fig. 5B, with the three most significant features beening square GLDM Dependence Variance, square First Order Minimum, and wavelet HH GLCM MCC(Fig. 5B).The pathological model showed good predictive performance. The ROC curve indicated that the AUC was 0.836 in the training set and 0.745 in the validation set(Fig. 5C, Fig. 5D).The calibration curve and Hosmer-Lemeshow goodness-of-fit test showed good agreement between the model’s predicted probabilities and actual outcomes ( p > 0.05).The decision curve analysis (DCA) results indicate the model’s clinical usefulness across a wide range of scenarios.In the training set, the threshold was 0.337 with an accuracy of 0.769, sensitivity of 0.844, specificity of 0.73, Brier score of 0.132, and 95% CI of 0.782–0.891. In the validation set, accuracy was 0.704, sensitivity of 0.697, specificity 0.708, Brier score 0.119, and 95% CI of 0.638–0.851.Details are shown in Fig. 5E-H. The inter-group difference analysis revealed significantly higher PS in the high-risk group than in the low-risk group for both the training set (Fig. 5I) and validation set (Fig. 5J) ( p < 0.001). Table 2 The clinical characteristics of cluster in cases of TCGA Variables Total (n = 327) Train (n = 229) Validation (n = 98) p Risk_group, n (%) 1 High 110 (33.6) 77 (33.6) 33 (33.7) Low 217 (66.4) 152 (66.4) 65 (66.3) Age, n (%) 0.875 ~65 164 (50.2) 116 (50.7) 48 (49) 66~ 163 (49.8) 113 (49.3) 50 (51) Gender, n (%) 0.744 Female 183 (56) 130 (56.8) 53 (54.1) Male 144 (44) 99 (43.2) 45 (45.9) Residual_tumor, n (%) 0.446 R0 220 (67.3) 150 (65.5) 70 (71.4) R1/R2 13 (4) 11 (4.8) 2 (2) RX/Unknown 94 (28.7) 68 (29.7) 26 (26.5) Histologic_type, n (%) 0.196 Mixed Subtype 67 (20.5) 46 (20.1) 21 (21.4) NOS 204 (62.4) 149 (65.1) 55 (56.1) Others 56 (17.1) 34 (14.8) 22 (22.4) Pathologic_stage, n (%) 1 I/II 265 (81) 186 (81.2) 79 (80.6) III/IV 62 (19) 43 (18.8) 19 (19.4) Tumor_location, n (%) 0.694 L-Lower 56 (17.1) 39 (17) 17 (17.3) L-Upper 76 (23.2) 56 (24.5) 20 (20.4) R-Lower 63 (19.3) 46 (20.1) 17 (17.3) R-Middle 14 (4.3) 8 (3.5) 6 (6.1) R-Upper 118 (36.1) 80 (34.9) 38 (38.8) Smoking_status, n (%) 0.864 Current 82 (25.1) 58 (25.3) 24 (24.5) Former 204 (62.4) 141 (61.6) 63 (64.3) Nonsmoker 41 (12.5) 30 (13.1) 11 (11.2) Radiotherapy, n (%) 0.19 NO 293 (89.6) 209 (91.3) 84 (85.7) YES 34 (10.4) 20 (8.7) 14 (14.3) Chemotherapy, n (%) 0.824 NO 219 (67) 152 (66.4) 67 (68.4) YES 108 (33) 77 (33.6) 31 (31.6) OS, n (%) 0.5 Alive 213 (65.1) 146 (63.8) 67 (68.4) Dead 114 (34.9) 83 (36.2) 31 (31.6) OS.time, Median (Q1,Q3) 21.73 (14.48, 35.1) 22.13 (14.93, 34.87) 20.32 (13.65, 35.17) 0.293 2.2.2 Prognostic Significance of Pathomics Features in LUAD Using the R package "survminer," the cutoff value of 0.3047 for the pathomics score (PS) predicted by the XG-Boost model was selected to divide 327 patients into a PS high-risk group (183) and a low-risk group (144).There was no significant difference in age distribution, residual tumor, pathologic stage, tumor location, smoking status, radiotherapy, and chemotherapy between the high and low groups (all interaction p-values > 0.05). However, there were significant differences in gender and histologic type between the high and low PS groups ( p < 0.05).Details are shown in Table 3.The median overall survival time for the high PS group was 40.967 months, while it was 59.933 months for the low PS group. High PS was significantly correlated with worse OS ( p < 0.05) (Fig. 6A). In univariate analysis, we found that the high PS group was a statistically significant risk factor for OS (HR = 1.933, 95% CI: 1.298–2.879, p = 0.001).In multivariate analysis, after adjustment for confounders, the high PS group (HR = 1.686, 95% CI: 1.118–2.542, p = 0.013) remained a significant risk factor for OS. Datails were shown in Fig. 6B.In the subgroup analysis, we found that high PS was a significant risk factor for OS in both the 65 subgroup (HR = 2.062, 95% CI: 1.169–3.637, p = 0.012).The interaction test indicated that there was no significant interaction between PS and different subgroups, including age, gender, residual tumor, pathological stage, radiotherapy, chemotherapy, smoking status, and tumor location (all p for interaction > 0.05). In other words, the effect of PS on OS was consistent across the age subgroups and other analyzed groups. This suggests that PS generated by the model serves as an independent prognostic factor for LUAD. Details are shown in Fig. 6C. Table 3 The clinical characteristics of PS subgroups Variables Total (n = 327) PS Low (n = 144) PS High (n = 183) p Age, n (%) 0.95 ~65 164 (50) 73 (51) 91 (50) 66~ 163 (50) 71 (49) 92 (50) Gender, n (%) 0.046 Female 183 (56) 90 (62) 93 (51) Male 144 (44) 54 (38) 90 (49) Residual_tumor, n (%) 0.271 R0 220 (67) 97 (67) 123 (67) R1/R2 13 (4) 3 (2) 10 (5) RX/Unknown 94 (29) 44 (31) 50 (27) Histologic_type, n (%) 0.026 Mixed Subtype 67 (20) 22 (15) 45 (25) NOS 204 (62) 90 (62) 114 (62) Others 56 (17) 32 (22) 24 (13) Pathologic_stage, n (%) 0.099 I/II 265 (81) 123 (85) 142 (78) III/IV 62 (19) 21 (15) 41 (22) Tumor_location, n (%) 0.938 L-Lower 56 (17) 24 (17) 32 (17) L-Upper 76 (23) 36 (25) 40 (22) R-Lower 63 (19) 28 (19) 35 (19) R-Middle 14 (4) 5 (3) 9 (5) R-Upper 118 (36) 51 (35) 67 (37) Smoking_status, n (%) 0.156 Current 82 (25) 31 (22) 51 (28) Former 204 (62) 90 (62) 114 (62) Nonsmoker 41 (13) 23 (16) 18 (10) Radiotherapy, n (%) 0.205 NO 293 (90) 133 (92) 160 (87) YES 34 (10) 11 (8) 23 (13) Chemotherapy, n (%) 0.824 NO 219 (67) 95 (66) 124 (68) YES 108 (33) 49 (34) 59 (32) 2.3 Potential Biological Mechanisms of Mitochondrial Genotyping in Pathological Features We analyzed GSVA differences between high and low PS groups in LUAD.The results showed that in the Hallmark genes set, the high PS group was significantly enriched in the G2M_CHECKPOINT, SPERMATOGENESIS, E2F-TARGETS, and other signaling pathways (Fig. 7A, Note:Pathway information are derived from the KEGG database [30] ).The top 30 pathways visualized in KEGG indicated that in the KEGG genes set, the high PS group was significantly enriched in the LYSINE_DEGRADATION, GLYOXYLATE-AND- DICARBOXYLATE, BASE-EXCISION-REPAIR, and other signaling pathways( Fig. 7B, Note:Pathway information are derived from the KEGG database [30] ).We analyzed immune cell infiltration between high and low PS groups. Among 22 immune cell types, only resting dendritic cells, resting mast cells, and activated CD4 memory T cells showed significant differences between the groups (Fig. 7C).genes mutation profiling showed that missense mutations were the most common, followed by frame shift deletions and nonsense mutations.Among them, the high PS group had a higher overall tumor mutational burden (TMB), indicating that patients in this group had more tumor mutations. Mutation rates of TP53 and TTN genes in both high and low PS groups exceeded 40%.Mutation rates of TP53 and TTN were higher in the high PS group than in the low PS group.We speculated that the high MT was associated with a poor prognosis in LUAD, possibly due to common mutations in genes such as TP53 and TTN( Fig. 7D, Fig. 7E). 3. Discussion We developed a mitochondria subtyping-based pathomics model for predicting the prognosis of LUAD patients through PS. Additionally, we explored the model’s underlying biological interpretability.The main findings of this study were: (1) Mitochondrial genotyping based on NMF clustering demonstrated independent prognostic value in LUAD (HR = 1.463, 95% CI: 1.036–2.065, p = 0.031); (2) The HE pathology-based pathomics model with 8 features achieved accurate prediction of mitochondrial subtypes with an AUC of 0.836 in the training set; (3) The biological interpretability analysis of the pathomics model showed that the PS, which reflects mitochondrial genotypes, was significantly associated with several factors. These included pathways such as the G2/M checkpoint, immune cell infiltration (including resting dendritic cells, resting mast cells, and CD4 memory T cells), genes mutations like TP53 and TTN. Mitochondria, as central organelles in cellular energy metabolism, exhibit functional abnormalities that are strongly correlated with cancer progression and poor prognosis across multiple malignancies.Extensive research on mitochondrial DNA, metabolic pathways, and genes expression in LUAD has revealed associations with patient survival rates and recurrence risk. For instance, a study by Hertweck KL et al. [ 31 ] found that LUAD subtypes characterized by high mitochondrial-related genes expression exhibited significant enrichment in oxidative phosphorylation (OXPHOS), the tricarboxylic acid (TCA) cycle, and metabolic reprogramming pathways. Patients in this group experienced significantly poorer outcomes. Similarly, Han et al. [ 32 ] , in a study published in Nature, showed that mitochondria-specific network architectures in non-small cell lung cancer (NSCLC) are linked to metabolic dependencies. Their findings suggest that structural metabolic reorganization of mitochondria is positively associated with tumor progression.Extending beyond LUAD, Bezwada et al. [ 33 ] revealed in another Nature study that metastatic sites in clear cell renal cell carcinoma (ccRCC) exhibit distinct mitochondrial metabolic reprogramming. Notably, TCA cycle activity was more prominent in metastatic lesions than in primary tumors, suggesting a fundamental shift in metabolic strategy during metastasis.Together, these findings underscore the crucial role of mitochondria in cancer metabolism and support the exploration of mitochondria-based pathomic features for prognostic modeling in LUAD. In summary, existing research has established the prognostic foundation of mitochondria in LUAD and other cancers.This study further confirmed the independent predictive value of mitochondrial-related genes subtyping in LUAD(HR = 1.463).This study further substantiates the independent predictive value of mitochondrial-related genes subtyping (HR = 1.463) through NMF clustering. This underscores the potential of mitochondrial dynamics, such as metabolic or genomic alterations, as potential biomarkers to advance precision medicine, including early intervention for high-risk patients.Future studies should validate this subtyping approach in large prospective cohorts and integrate treatment response data to achieve clinical translation. In recent years, pathomics models based on HE-stained pathology images have emerged as vital tools for predicting molecular features of lung cancer, owing to their non-invasive nature, cost-effectiveness, and high-throughput advantages. Current research primarily focuses on driver genes and immune markers. For instance, Yuan L et al. [ 34 ] reviewed 109 relevant publications published between 2018 and 2023. Their analysis revealed that integrating pathomics with multi-omics studies was crucial for in-depth assessment and characterization of the tumor microenvironment, enabling the identification of broader tumor feature spectra.This approach ultimately facilitates the development of multimodal fusion models to precisely evaluate personalized immunotherapy efficacy and prognosis in lung cancer patients.For example, MA L et al. [ 35 ] employed multi-omics modeling to analyze progressive supranuclear palsy (PSP), identifying AKT1 driver mutations and impaired TP53 tumor suppressor pathways, thereby demonstrating a link between pathological morphology and genomic instability.Similarly, the pathomics model developed in this study to predict mitochondrial-related genes subtypes shows potential clinical significance in the prognostic assessment of LUAD.This study reveals significant biological associations of the PS.Specifically, in the LUAD model, PS reflects activation states of pathways such as the G2/M checkpoint. This finding is consistent with established research by Zhang L et al. [36] , which shows that mitochondrial-related genes promote tumor progression by activating this pathway. Regarding immune cell infiltration, this study found that the high PS group had lower infiltration levels of resting dendritic cells(DCs), resting mast cells, and CD4⁺ memory T cells ,which was associated with poorer OS (p < 0.05).These findings align with the established understanding that immune cell infiltration shapes the tumor immune microenvironment (TME) and is closely associated with tumor progression, patient survival, and treatment response [ 11 ][37][38][39][40] .Specifically ,resting mast cells have been shown to be strongly associated with better OS and disease-free survival (DFS),while activated mast cells correlate with adverse prognosis [40][ 41 ] . This founding aligns with the observed enrichment of resting mast cells in the low-risk PS group. another study developed an immune prognostic signature based on ferroptosis-related genes. It demonstrated that resting dendritic cells (DCs) and resting mast cells are closely associated with better OS [42] , further supporting the link between PS and immune characteristics in LUAD. From a mechanistic perspective, mitochondria, as the core of cellular energy metabolism, regulate immune activity within the TME, ether directly or indirectly through metabolic reprogramming when they become dysfunctional. Mitochondrial metabolic subtypes reshape the tumor immune microenvironment through several mechanisms. These include influencing antigen presentation, immune cell activation, and immune checkpoint expression. Together, these changes modulate patient prognosis and treatment response [ 11 ][37][38][39][40] . Specific mechanisms include: (1)Metabolic reprogramming promotes an immunosuppressive microenvironment [ 11 ][37] . Enhanced mitochondrial metabolism and glycolysis suppress antigen-presenting cell function. Tumor cells consume glucose and amino acids. This impairs DC antigen presentation and inhibits CD8⁺ T lymphocytes. This leads to reduced immune cell infiltration. (2)Dendritic cell dysfunction. Multi-omics analyses indicate that in high-risk subtypes, the expression of activation markers (e.g., CD80/CD86) in dendritic cells is reduced, while the secretion of immunosuppressive cytokines (e.g., IL-10, TGF-β) is increased [37] . These changes are closely associated with mitochondrial metabolic reprogramming. Moreover tumor-derived lactate and ketone bodies inhibit DC maturation through epigenetic mechanisms such as histone modification. Moreover, efferocytosis and the TGF-β and WNT/β-catenin signaling pathways have been reported to suppress DC maturation [38][39] . (3)Immune checkpoint dysregulation and T cell exhaustion [37][40] . Abnormal mitochondrial metabolism can also lead to dysregulated expression of immune checkpoints such as PD-L1, CTLA-4, and LAG-3. Certain checkpoint molecules are upregulated, however, the overall signaling pathway suppression leads to T cell exhaustion. This contributes to an immune "desert" phenotype, which is associated with poor prognosis and resistance to immune checkpoint inhibitors. Furthermore, our mutation analysis showed that patients in the high-PS group had significantly higher TMB and increased mutation frequencies of TP53 and TTN. These findings are visually summarized in the waterfall plot.TP53 is one of the most frequently mutated tumor suppressor genes in LUAD (mutation rate ~ 40–50%) and plays a critical role in DNA repair, apoptosis, and cell cycle regulation. Mutations in TP53 may enable cells to evade normal checkpoint controls, which promotes tumorigenesis and progression [ 43 ] .TTN alterations are also common in LUAD and are linked to tumor heterogeneity and patient prognosis [ 44 ] . Our findings align with existing research on TP53 and TTN, further validating mutation analysis based on PS stratification. Meanwhile, TP53 and TTN mutations have been considered potential biomarkers for predicting LUAD patients’ response to immune checkpoint inhibitor (ICI) treatment [40] . This study has some limitations often found in similar bioinformatics mining and model construction studies. (1)Retrospective data and validation cohorts: The data used in this study were entirely from public databases such as TCGA and MitoCarta 3.0, which are retrospective. The research was based on a single TCGA cohort with a cross-sectional design, which to some extent limits the generalizability of the results and makes it difficult to assess dynamic changes related to longitudinal follow-up or treatment response. Nevertheless, we confirmed the reliability of mitochondrial subtyping and the pathomics model using multiple independent cohorts;however,their clinical translation still requires validation through large-scale, multi-center prospective studies. Therefore, future efforts should focus on improving the model’s reliability in longitudinal study designs and promoting its clinical application. Depth of mechanistic exploration: The study explored the biological mechanisms of mitochondrial subtyping through pathway enrichment, immune infiltration, and genes mutation analysis. However, the specific molecular pathways through which mitochondrial-related genes influence immune infiltration, genes mutations, and other processes remain unclear and require further in-depth in vivo and in vitro experiments for clarification. Additionally, somatic mutation analysis was conducted in TCGA-LUAD samples. The samples were stratified by high and low pathomics score (PS) and visualized using the maftools package. However ,these results remain primarily descriptive and visual, lacking advanced statistical inference or interaction testing. More systematic analyses of statistical associations and interactions are needed to fully evaluate the relationship between PS and mutational profiles. (3)Generalizability of the model: Although the model performed well in independent datasets, there is inherent selection bias in public database populations. Whether the predictive efficacy of the model remains consistent across different ethnic groups, clinical subtypes, or LUAD patient populations, receiving different treatment regimens requires further evaluation. In summary, this study demonstrates that pathomics score, derived from machine learning based image analysis ,can serve as an independent prognostic factor. It allows for low-cost and accurate assessments based on routine clinical H&E slides. This approach can help advance precision diagnosis and treatment for LUAD. The study also reveals an important association between mitochondrial-related genes subtyping and LUAD prognosis, providing a new perspective for personalized treatment.Future research should incorporate multi-center clinical data and experimental validation to strenghten generalizability. This approach will also enhance machanistic understanding and advance precision medicine in LUAD. These findings not only help optimize treatment plans but also provide important insights for early identification of high-risk individuals and improved clinical management. 4. Methods 4.1. Data Sources and Acquisition This study used data from several public databases.Genomic and pathological data for LUAD were obtained from the TCGA database (https://portal.gdc.cancer.gov/), with data usage following DBPC policies. Mitochondrial-related genes were sourced from the MitoCarta 3.0 database, which includes 1,136 human mitochondrial-related genes. Among these, 213 genes have transcriptomic data in TCGA, presented as RNA-seq log2(FPKM + 1) format.Clinical data related to LUAD were also collected. This included survival status, survival time, age, gender, tumor stage, pathological stage, history of radiotherapy, smoking history, pathological type, tumor location, and chemotherapy history. Primary, treatment-naive LUAD cases with RNA-seq data were included. Samples with missing follow-up data, survival time less than 30 days, or incomplete clinical data were excluded. This process left 443 cases for survival analysis.Additionally, among the 478 samples with available pathological images, those failing quality control were removed.After quality control, 327 cases with complete clinical data, RNA-seq, and qualified pathological images were selected to construct the pathology-based omics model.Specific inclusion and exclusion criteria are shown in Fig. 1, The overall framework was summarized in Fig. 2. Because the TCGA database is publicly accessible and all patient information is anonymized, this study did not require ethics approval. 4.2. Construction of Prognostic Risk Model for Mitochondrial-Related Genes 4.2.1 NMF Clustering We used the "NMF" package in R for typing analysis. NMF (non-negative matrix factorization) unsupervised clustering method was used to cluster mitochondrial-related genes, obtaining samples of different pathological types. During clustering, we applied the Brunet algorithm with the rank parameter set between 2 and 5.We selected the optimal rank value based on the point at which the cophenetic correlation coefficient showed the steepest decline.For mitochondrial typing, RNA-seq data in log2(FPKM + 1) format were used. 4.2.2 Prognostic Analysis Patients were divided into high-risk and low-risk groups based on model-predicted risk scores. The impact of MRGs on patient prognosis was systematically evaluated using survival analysis, Cox regression models, and subgroup analysis. The detailed methodologies are described as follows.(1) Survival analysis for the classification was conducted using the R package "survival," with "survminer" applied to summarize and visualize the results. Kaplan-Meier survival curves were used to illustrate the survival probabilities of different risk groups over time, with the median survival time defined as the time when survival probability reaches 50%.Log-rank tests were used to assess differences in survival between groups.(2)Univariate-multivariate analysis was performed using the "survival" package in R, and tables and forest plots for univariate and multivariate analyses were created. The Cox proportional hazards model can study the relationship between one or more risk factors and the occurrence of survival outcomes. Univariate Cox regression was used to identify factors associated with OS, while multivariate Cox regression was applied to determine independent prognostic factors and examine the combined effects of multiple variables. When HR > 1, the independent variable is considered a risk factor; when HR < 1, the independent variable is considered a protective factor.(3)Subgroup Analysis and Interaction Test:The "cmprsk," "survival," and "forestplot" packages in R were used to conduct exploratory subgroup analyses to evaluate the impact of risk groups (high-risk vs low-risk) on patient prognosis across different covariate subgroups. Likelihood ratio tests were employed to assess the interaction between risk groups and other covariates. 4.3. Pathological Methodology 4.3.1 Image Acquisition, Segmentation, and Feature Extraction Image acquisition : Pathological images were downloaded from the TCGA (https://tcga-data.nci.nih.gov/tcga/) database. These images consist of formalin-fixed and paraffin-embedded pathological tissue slices in svs format, with a maximum magnification of 20× or 40× (H&E-stained histopathological images) [20][21] . Train/validation split : Patients were randomly allocated into a training set (n = 229) and a validation set (n = 98) in a 7:3 ratio [22][23] . A fixed random seed was used to ensure reproducibility. Pathological image processing and segmentation: The OTSU algorithm (https://opencv.org/) was used to segment the tissue area of the pathological slice. Also called the maximum inter-class variance method, this thresholding algorithm binarizes the image by separating the background from the tissue region of interest [24] .The 40× images were segmented into multiple 1024×1024-pixel sub-images. The 20× images were segmented into multiple 512×512-pixel sub-images and then upsampled to 1024×1024 pixels.Pathologists reviewed the images and excluded those with poor quality (contamination, blurriness, or more than 50% blank areas) sub-images. Ten sub-images were randomly selected from each pathological image for subsequent analysis [20][21] . Feature extraction: The PyRadiomics open-source package (https://pyradiomics.readthedocs.io/en/latest/) was used to standardize each sub-image and extract 93 original features, including first-order and second-order features. Additionally, higher-order features were extracted, such as wavelet (LL, LH, HL, HH), square, square root, logarithm, exponential, gradient, and LBP2D, resulting in a total of 1,023 features.After extracting features from the 10 sub-images of each patient's pathological image, their average was calculated and used as the representative pathological feature for subsequent data analysis [19][25][26] . 4.3.2 mRMR_RFE Feature Selection The best feature subset was selected using the mRMR(maximum relevance minimum redundancy) algorithm and the RFE(recursive feature elimination) algorithm. The mRMR algorithm selects features by considering both the correlation between features and the target variable, as well as the correlation among features themselves. The method uses mutual information to measure relevance and redundancy. Relevance is calculated by the information gain of each feature with the category, while redundancy is measured by summing mutual information among features and dividing by the square of the feature subset size. RFE feature selection ranks predictors before modeling and removes the least important features sequentially.The goal is to find a subset of predictors that can be used to generate an accurate model. The model is trained iteratively. After each training, n least important features are removed. The model is retrained with the remaining features, and their importance is reassessed. This process repeats until the optimal feature subset is obtained.The mRMR method selects the top 20 features, which are then refined using RFE feature selection.RFE with 10-fold cross-validation was used to further refine the 20 features. Based on the performance shown in Fig. 5A, the accuracy peaked at approximately 0.62 when the number of features was dropped to 8.Consequently, 8 features were retained for modeling [27][28][29] . 4.3.3 XGBoost model establishment Using the "caret" package in R, we built a model on the training set with features selected by mRMR_RFE. The model was developed using the XGBoost algorithm, which stands for Extreme Gradient Boosting.XGBoost is a boosting algorithm based on the gradient boosting framework that iteratively adds decision trees to reduce prediction errors. Compared to traditional gradient boosting machines (GBM), XGBoost offers significant improvements in computational efficiency, parallel processing, and model flexibility, resulting in more robust performance in classification and regression tasks.XGBoost improves performance through several key features: (1) introducing a regularization term to prevent overfitting; (2) using an approximate greedy algorithm during tree construction that evaluates gains of all features at each split and selects the one with the highest gain; (3) considering weighted quantiles of feature values when splitting nodes to enhance robustness; and (4) automatically handling missing values by assigning them to child nodes that increase gain after splitting.Using the XGBoost algorithm, we modeled the selected pathological features to predict risk group classification. Pathological model outputs included risk probabilities and pathomics scores (PS). 4.3.4 XGBoost model evaluation The R packages "pROC," "measures," "ResourceSelection," "rms," and "rmda" were used to evaluate the model's performance. Evaluation metrics included accuracy (ACC), specificity (SPE), sensitivity (SEN), positive predictive value (PPV), and negative predictive value (NPV).The x-axis of the receiver operating characteristic curve (ROC) represents the false positive rate (1 - specificity), and the y-axis represents the true positive rate (sensitivity).A larger receiver operating characteristic(ROC)-area under the curve(AUC) means a greater area under the curve. When the curve bulges more towards the upper left corner, it indicates better model performance.Calibration curves were drawn and the Hosmer-Lemeshow goodness-of-fit test was conducted to evaluate the calibration of the pathological prediction model; the Brier score was used to quantify the overall performance of the pathological prediction model, with smaller values indicating better consistency in model predictions; decision curves (DCA) were plotted to show the clinical benefit of the pathological prediction model. 4.4. Mechanism Analysis of Pathomics 4.4.1 GSVA Enrichment Analysis between High and Low PS Groups The expression matrix of 327 LUAD patients from TCGA was used to calculate pathway enrichment scores for KEGG and Hallmark genes sets in each sample. This calculation was performed using GSVA.The R package "limma" was used for differential analysis between high and low PS groups. A total of 184 KEGG pathways and 50 Hallmark genes sets were analyzed. 4.4.2 Analysis of Differences in Immune Cell Abundance The 'CIBERSORT' package was used to analyze immune infiltration, and the Wilcoxon rank-sum test was applied to assess differences in immune cell infiltration between high and low PS groups. 4.4.3 Genes Mutation Analysis between High and Low PS Groups The relationship between the mutation profile of TCGA-LUAD patients and model-predicted PS groups was assessed using available somatic mutation data. Mutation data were downloaded from the TCGA data portal, with 322 samples overlapping with pathological data. Somatic variants were stored in mutation annotation format (MAF). The R package maftools was employed to analyze the mutation data and visualize the top 15 genes with the highest mutation frequencies. 4.5. Statistical analysis Data processing, statistical analyses, and visualization were performed using R software (v4.1.0). Categorical variables were compared using chi-square tests, while continuous variables were analyzed using Wilcoxon rank-sum tests or t-tests. For the pathological intersection sample set from TCGA, samples with complete pathological images, genes expression matrices, and clinical data were screened using the caret (v6.0-93) and survminer (v0.4.9) packages. Dataset partitioning was performed using the caret package with a fixed random seed (set.seed(123)); zero-variance features were removed, and samples were randomly split into training and validation sets at a 7:3 ratio. The 1,023 radiomic features extracted from the validation set using pyradiomics (v3.0.1) were standardized by Z-score normalization based on the (mean ± standard) deviation of the training set.Inter-group differences in clinical characteristics were assessed using independent t-tests (continuous variables) or chi-square tests (categorical variables).Inter-group differences in XGBoost predictions were analyzed using the ggpubr package.Optimal cutoff values for PS were computed using survminer to create high/low binary groups. Statistical significance was defined as p < 0.05, with notation as follows: ns ( p ≥ 0.05), * ( p < 0.05), ** ( p < 0.01), *** ( p < 0.001), and **** ( p < 0.0001). Declarations Author Contribution Sijuan Xu :conceptuatization,experimental design(lead),writing-original draft(lead),and writing -review and deiting (equal).Jianzhong Lu:conceptualization (equal), data collection and analysis (lead), writing – original draft (equal), and writing – review and editing (equal).All authors read and approved the final manuscript. Acknowledgement We thank the authors of TCGA and MitoCarta 3.0 databas for making their data public for analysis. Data Availability The data used in this study were sourced from two publicly accessible databases: TCGA (https://portal.gdc.cancer.gov/) and MitoCarta 3.0 (https://www.broadinstitute.org/files/shared/metabolism/mitocarta/human.mitocarta3.0.html).The data used in this study came from the publicly accessible databases TCGA and MitoCarta 3.0.We confirm that other researchers can obtain the same datasets from these repositories for further analysis. References Siegel, R. L., Miller, K. D., Fuchs, H. E., Jemal, A. & Cancer statistics CA: A Cancer Journal for Clinicians, 72(1), 7–33. (2022). https://doi.org/10.3322/caac.21708 (2022). Leiter, A., Veluswamy, R. R. & Wisnivesky, J. P. The global burden of lung cancer: current status and future trends. Nat. Reviews Clin. Oncol. , 20 (9), 624–639. https://doi.org/10.1038/s41571-023-00798-3(2023 ). Chen, H. Correlation of LOXL2 expression in non-small cell lung cancer with immunotherapy. Int. J. Clin. Exp. Pathol. 17 (9), 268–286. https://doi.org/10.62347/zieg9007 (2024). Zhao, H., Li, W., Li, X., Ding, Z. & Zhao, S. Survival benefit of resection surgery for lung adenocarcinoma with bone metastases and a post-operative prognosis nomogram establishment and validation. J. Thorac. Disease , 14 (12), 4877–4893. https://doi.org/10.21037/jtd-22-1514(2022 ). Zhang, Z. et al. Dynamics of Serum Tumor Markers Can Serve as a Prognostic Biomarker for Chinese Advanced Non-small Cell Lung Cancer Patients Treated With Immune Checkpoint Inhibitors. Front. Immunol. 11 https://doi.org/10.3389/fimmu.2020.01173 (2020). Wu, P. et al. FOXK2 in skeletal muscle development: a new pathogenic gene for congenital myopathy with ptosis. EMBO Mol. Med. , 17 (7), 1599–1630. https://doi.org/10.1038/s44321-025-00247-x(2025 ). BALLARD, J. W. O. & MELVIN, R. G. Linking the mitochondrial genotype to the organismal phenotype. Mol. Ecol. , 19 (8), 1523–1539. https://doi.org/10.1111/j.1365-294x.2010.04594.x(2010 ). Dong, L., Gopalan, V., Holland, O. & Neuzil, J. Mitocans Revisited: Mitochondrial Targeting as Efficient Anti-Cancer Therapy. Int. J. Mol. Sci. 21 (21), 7941. https://doi.org/10.3390/ijms21217941( (2020). Jeon, J. H. et al. Migration and invasion of drug-resistant lung adenocarcinoma cells are dependent on mitochondrial activity. Exp. Mol. Med. 48 (12), e277–e277. https://doi.org/10.1038/emm.2016.129 (2016). Zhang, X. et al. Multi-omic and machine learning analysis of mitochondrial RNA modification genes in lung adenocarcinoma for prognostic and therapeutic implications. Transl Oncol. 53 , 102306. http://doi.org/10.1016/j.tranon.2025.102306 (2025). Zhang, P. et al. Mitochondrial Pathway Signature (MitoPS) predicts immunotherapy response and reveals NDUFB10 as a key immune regulator in lung adenocarcinoma. J. Immunother. Cancer . 13 (7), e012069. https://doi.org/10.1136/jitc-2025-012069 (2025). Mennuni, M. et al. High mitochondrial DNA levels accelerate lung adenocarcinoma progression. Sci. Adv. 10 (44). https://doi.org/10.1126/sciadv.adp3481( (2024). Zhang, Y. et al. The circRNA cEMSY Induces Immunogenic Cell Death and Boosts Immunotherapy Efficacy in Lung Adenocarcinoma. Cancer Res. 85 (3), 497–514. https://doi.org/10.1158/0008-5472.can-24-1484( (2024). Wen, J. J., Cummins, C. B., Williams, T. P. & Radhakrishnan, R. S. The Genetic Evidence of Burn-Induced Cardiac Mitochondrial Metabolism Dysfunction. Biomedicines 8 (12), 566. https://doi.org/10.3390/biomedicines8120566( (2020). Jarrett, S. G., Rohrer, B., Perron, N. R., Beeson, C. & Boulton, M. E. Assessment of mitochondrial damage in retinal cells and tissues using quantitative polymerase chain reaction for mitochondrial DNA damage and extracellular flux assay for mitochondrial respiration activity. Methods Mol. Biol. , 935 :227–243 http://doi.org/10.1007/978-1-62703-080-9_16(2013 ). Chen, Y. et al. Multiphoton microscopy as a diagnostic imaging modality for pancreatic neoplasms without hematoxylin and eosin stains. J. Biomed. Opt. , 19 (9):96008. http://doi.org/10.1117/1.JBO.19.9.096008(2014 ). Chang, L. et al. Advancing precision medicine: the transformative role of artificial intelligence in immunogenomics, radiomics, and pathomics for biomarker discovery and immunotherapy optimization. Cancer Biology Med. , 1–15 https://doi.org/10.20892/j.issn.2095-3941.2024.0376(2025 ). Kunchur, N. N. & Mostaço-Guidolin, L. B. Development of an image classification pipeline for atherosclerotic plaques assessment using supervised machine learning. BMC Bioinform. , 23 (1). https://doi.org/10.1186/s12859-022-05059-1(2022 ). Liu, K. & Hu, J. Classification of acute myeloid leukemia M1 and M2 subtypes using machine learning. Comput. Biol. Med. , 147 :105741. http://doi.org/10.1016/j.compbiomed.2022.105741(2022 ). Chen, L., Zeng, H., Zhang, M., Luo, Y. & Ma, X. Histopathological image and gene expression pattern analysis for predicting molecular features and prognosis of head and neck squamous cell carcinoma. Cancer Med. 10 (13), 4615–4628. http://doi.org/10.1002/cam4.3965 (2021). Zeng, H., Chen, L., Zhang, M., Luo, Y. & Ma X.Integration of histopathological images and multi-dimensional omics analyses predicts molecular features and prognosis in high-grade serous ovarian cancer. Gynecol. Oncol. 163 (1), 171–180. https://doi.org/10.1016/j.ygyno.2021.07.015 (2021). Qu, W. F. et al. Development of a deep pathomics score for predicting hepatocellular carcinoma recurrence after liver transplantation. Hep. Intl. 17 (4), 927–941. https://doi.org/10.1007/s12072-023-10511-2 (2023). Cai, X. et al. Development of a Pathomics-Based Model for the Prediction of Malignant Transformation in Oral Leukoplakia. Laboratory investigation; a journal of technical methods and pathology, 103(8), 100173. https://doi.org/10.1016/j.labinv.2023.100173, LI FP,et al.MRI–based radiomics nomogram for predicting temporal lobe injury after radiotherapy in nasopharyngeal carcinoma. European Radiology; 32:1106–1114. https://doi.org/10.1007/s00330-021-08254-5 (2022) (2023). Li, M. L. et al. A clinical-radiomics nomogram for the preoperative prediction of lymph node metastasis in colorectal cancer. J. Transl Med. 18 , 46. https://doi.org/10.1186/s12967-020-02215-0 (2020). Li, Y., Chen, C. M., Li, W. W., Shao, M. T. & Dong, Y., Zhang QC.Radiomic features based on pyradiomics predict CD276 expression associated with breast cancer prognosis.Heliyon 10, e37345. (2024). https://doi.org/10.1016/j.heliyon.2024.e37345 Wang, X. et al. Weakly Supervised Deep Learning for Whole Slide Lung Cancer Image Analysis. IEEE Trans. Cybernetics . 50 (9), 3950–3962. https://doi.org/10.1109/tcyb.2019.2935141 (2020). Saednia, K. et al. Quantitative digital histopathology and machine learning to predict pathological complete response to chemotherapy in breast cancer patients using pre-treatment tumor biopsies. Sci. Rep. , 12 (1). https://doi.org/10.1038/s41598-022-13917-4(2022 ). Nishio, M., Nishio, M., Jimbo, N. & Nakane, K. Homology-Based Image Processing for Automatic Classification of Histopathological Images of Lung Tissue. Cancers 13 (6), 1192. https://doi.org/10.3390/cancers13061192 (2021). Kanehisa, M., Furumichi, M., Sato, Y., Matsuura, Y. & Ishiguro-Watanabe M.KEGG: biological systems database as a model of the real world. Nucleic Acids Res. 53 . D672–D677. http://doi.org/10.1093/nar/gkae909 (2025). Hertweck, K. L. et al. Clinicopathological significance of unraveling mitochondrial pathway alterations in non-small‐cell lung cancer. FASEB J. , 37 (7). https://doi.org/10.1096/fj.202201724rr(2023 ). Han, M. et al. Spatial mapping of mitochondrial networks and bioenergetics in lung cancer. Nature , 615 (7953), 712–719. https://doi.org/10.1038/s41586-023-05793-3(2023 ). Bezwada, D. et al. et,al. Mitochondrial complex I promotes kidney cancer metastasis. Nature , 633 (8031), 923–931. http://doi.org/10.1038/s41586-024-07812-3(2024 ). Yuan, L. et al. Unveiling the landscape of pathomics in personalized immunotherapy for lung cancer: a bibliometric analysis. Front. Oncol. 14 https://doi.org/10.3389/fonc.2024.1432212 (2024). Ma, L. et al. An advanced molecular medicine case report of a rare human tumor using genomics, pathomics, and radiomics. Front. Genet. 13. https://doi.org/10.3389/fgene.2022.987175 (2023). Zhang, L. et al. Xanthatin Induces Cell Cycle Arrest at G2/M Checkpoint and Apoptosis via Disrupting NF-κB Pathway in A549 Non-Small-Cell Lung Cancer Cells. Molecules, 17(4), 3736–3750. https://doi.org/10.3390/molecules17043736, Optimized Dynamic Network Biomarker Deciphers a High-Resolution Heterogeneity Within Thyroid Cancer Molecular Subtypes. Med Research, https://doi.org/10.1002/mdr2.70004(2025),,&Clinical prognostication and immunotherapy response prediction in esophageal squamous cell carcinoma using the DNA damage repair-associated signature. Environmental toxicology, 39(5), 2803–2816. https://doi.org/10.1002/tox.24155(2024),, Integration of the bulk transcriptome and single-cell transcriptome reveals efferocytosis features in lung adenocarcinoma prognosis and immunotherapy by combining deep learning. Cancer cell international, 24(1), 388. https://doi.org/10.1186/s12935-024-03571-3(2024),,& Integrating multi-omics and machine learning survival frameworks to build a prognostic model based on immune function and cell death patterns in a lung adenocarcinoma cohort. Frontiers in immunology, 15, 1460547. https://doi.org/10.3389/fimmu.2024.1460547(2024) (2012). Wang, C. et al. Patterns of immune infiltration in lung adenocarcinoma revealed a prognosis-associated microRNA-mast cells network. Human cell, 33(1), 205–219. https://doi.org/10.1007/s13577-019-00300-1, Development and Validation of a Robust Ferroptosis-Related Prognostic Signature in Lung Adenocarcinoma. Frontiers in cell and developmental biology, 9, 616271. https://doi.org/10.3389/fcell.2021.616271 (2021) (2020). Frille, A. et al. TP53 co-mutations in advanced lung adenocarcinoma: comparative bioinformatic analyses suggest ambivalent character on overall survival alongside KRAS, STK11 and KEAP1 mutations. Front. Oncol. 14 , 1357583. https://doi.org/10.3389/fonc.2024.1357583 (2024). Liu, Z. et al. Heterogeneous pattern of gene expression driven by TTN mutation is involved in the construction of a prognosis model of lung squamous cell carcinoma. Front. Oncol. 13 , 916568. https://doi.org/10.3389/fonc.2023.916568 (2023). Additional Declarations No competing interests reported. Supplementary Files supplementaryimformation.zip Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7252366","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":548025377,"identity":"a8570f1c-f962-4889-b959-4750b011f79c","order_by":0,"name":"Xu Sijuan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABAElEQVRIiWNgGAWjYJCCAxJgEsy2AWLmBhiPoBZGoNI0BgY2RsJaoBrBWg4T1mJwI/3hAcs2uzy+483PH3zcc17O4H5j44MPZxjk+cWw65OckWNwQLItuVjyzDHDxhnPbhsbHGNsNpxxg8Fw5uwErFr4JXIYgFqYEzfcyGFs5jlwO3HDMcY2aZ4PDAkGt7FrYZNIfwDUUg/Tco6wFn6JBJDDDsO0HIBquYFbi2TPG4MDEueOJ84E+mXmjAPJxpLHEoF+OSOB0y8Gx9Mff5Yoq07sO9784MOHA3ZyfIcPH3zw4ZiNPL80di0gwCyBRRCbGAIwfsArPQpGwSgYBSMeAAB+d2xXpckGugAAAABJRU5ErkJggg==","orcid":"","institution":"Affiliated Hospital of Xiangnan University","correspondingAuthor":true,"prefix":"","firstName":"Xu","middleName":"","lastName":"Sijuan","suffix":""},{"id":548025378,"identity":"ac1634f6-b4ac-461a-8809-ab317c865bcb","order_by":1,"name":"Lu Jianzhong","email":"","orcid":"","institution":"The Fourth People’s Hospital of Chenzhou","correspondingAuthor":false,"prefix":"","firstName":"Lu","middleName":"","lastName":"Jianzhong","suffix":""}],"badges":[],"createdAt":"2025-07-30 11:38:17","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7252366/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7252366/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":96844759,"identity":"1467b134-b028-4acf-85c1-d4531a3487de","added_by":"auto","created_at":"2025-11-26 16:29:49","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12377,"visible":true,"origin":"","legend":"","description":"","filename":"Figurelegends.docx","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/e419e31edd0d4cd91345561d.docx"},{"id":96918984,"identity":"11a29faf-560d-4cfe-ad96-9eb2ecf819b7","added_by":"auto","created_at":"2025-11-27 14:12:57","extension":"doc","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":431616,"visible":true,"origin":"","legend":"","description":"","filename":"manuscriptcleanversion.doc","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/fa59f3f80b4edb726254e039.doc"},{"id":96844768,"identity":"38d641f3-fbaf-491c-8f97-7c16cbfc9db4","added_by":"auto","created_at":"2025-11-26 16:29:49","extension":"tiff","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3146678,"visible":true,"origin":"","legend":"","description":"","filename":"figure1.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/c927243b883101e605026657.tiff"},{"id":96918275,"identity":"0f4764cc-2b99-4f25-a92f-89ae3dd910c3","added_by":"auto","created_at":"2025-11-27 14:11:34","extension":"tiff","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3146678,"visible":true,"origin":"","legend":"","description":"","filename":"figure2.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/ff34e575ad3d3deaa5714609.tiff"},{"id":96917587,"identity":"91ac5dac-0a08-4cba-baf0-8a7e551ee552","added_by":"auto","created_at":"2025-11-27 14:10:10","extension":"tiff","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3146678,"visible":true,"origin":"","legend":"","description":"","filename":"figure3.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/0449fde6c72a653a3a4b89af.tiff"},{"id":96844765,"identity":"e6947830-47bc-48c8-807a-bcb38b5a8ac4","added_by":"auto","created_at":"2025-11-26 16:29:49","extension":"tiff","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3146678,"visible":true,"origin":"","legend":"","description":"","filename":"figure4.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/5714e9dc98ac3850721abafb.tiff"},{"id":96844771,"identity":"dd8f7259-c8bc-4e4a-91ee-ef73a24fc99d","added_by":"auto","created_at":"2025-11-26 16:29:49","extension":"tiff","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3146678,"visible":true,"origin":"","legend":"","description":"","filename":"figure5.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/6b2a9192c64adf2f0d26a5ab.tiff"},{"id":96844775,"identity":"1b6e84db-5fe5-4a4e-8259-92d06be6c1db","added_by":"auto","created_at":"2025-11-26 16:29:49","extension":"tiff","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3146678,"visible":true,"origin":"","legend":"","description":"","filename":"figure6.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/21554d1d3e8faf99dbb93732.tiff"},{"id":96844782,"identity":"5e867973-17a3-43b6-ae8a-b6b3b7b1df66","added_by":"auto","created_at":"2025-11-26 16:29:49","extension":"tiff","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3146678,"visible":true,"origin":"","legend":"","description":"","filename":"figure7.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/c97e7e15860ba8830a560282.tiff"},{"id":96920216,"identity":"4ffdd9c9-6409-437c-bac4-c3c80440656d","added_by":"auto","created_at":"2025-11-27 14:14:56","extension":"json","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5136,"visible":true,"origin":"","legend":"","description":"","filename":"2cf58a05b1854cee9de756746b46525f.json","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/60d62238aa2a422ef154819a.json"},{"id":96844810,"identity":"d82afb66-4212-453f-b8b8-845fcba59cd0","added_by":"auto","created_at":"2025-11-26 16:29:51","extension":"zip","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":97558274,"visible":true,"origin":"","legend":"","description":"","filename":"supplementaryimformation.zip","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/55e5ff1e11e9908f190f88b5.zip"},{"id":96918109,"identity":"1238a6cf-1532-4974-b208-5a24a22031ed","added_by":"auto","created_at":"2025-11-27 14:11:10","extension":"xml","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":159334,"visible":true,"origin":"","legend":"","description":"","filename":"2cf58a05b1854cee9de756746b46525f1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/5107cc5f53716b6328449d43.xml"},{"id":96844776,"identity":"f34351e6-bced-4792-bc01-dd361963ed87","added_by":"auto","created_at":"2025-11-26 16:29:49","extension":"tiff","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3146678,"visible":true,"origin":"","legend":"","description":"","filename":"figure1.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/f27ec3a112bcf2f9d2f8700c.tiff"},{"id":96919345,"identity":"c0e42ab4-c4c2-42df-8365-0d7cea6392a9","added_by":"auto","created_at":"2025-11-27 14:13:40","extension":"tiff","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3146678,"visible":true,"origin":"","legend":"","description":"","filename":"figure2.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/3bc2a4620b1af615d9bbf77a.tiff"},{"id":96844790,"identity":"2c0d7f1f-83c9-48de-9efd-e9a3c035b5f4","added_by":"auto","created_at":"2025-11-26 16:29:50","extension":"tiff","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3146678,"visible":true,"origin":"","legend":"","description":"","filename":"figure3.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/7a9af5e3bc5f10293a948b09.tiff"},{"id":96919374,"identity":"9f3b5cfd-f845-4ecc-a00b-d61f739331a1","added_by":"auto","created_at":"2025-11-27 14:13:45","extension":"tiff","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3146678,"visible":true,"origin":"","legend":"","description":"","filename":"figure4.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/33969e39459a804d05e80cf0.tiff"},{"id":96919004,"identity":"b631cf25-c6dd-4cd3-9e41-967f874585f6","added_by":"auto","created_at":"2025-11-27 14:12:58","extension":"tiff","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3146678,"visible":true,"origin":"","legend":"","description":"","filename":"figure5.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/823f1698a31138c6bf8fdc22.tiff"},{"id":96844783,"identity":"f24438e5-583f-4e7a-bba3-1cfd21828603","added_by":"auto","created_at":"2025-11-26 16:29:49","extension":"tiff","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3146678,"visible":true,"origin":"","legend":"","description":"","filename":"figure6.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/7ca272826d5e4828170a0b46.tiff"},{"id":96844792,"identity":"a3375513-228f-4166-9eb0-1625f482aa5b","added_by":"auto","created_at":"2025-11-26 16:29:50","extension":"tiff","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3146678,"visible":true,"origin":"","legend":"","description":"","filename":"figure7.tiff","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/add9091072bb6a9e95862d16.tiff"},{"id":96919161,"identity":"837427ff-0a98-4567-824e-27fc414ac336","added_by":"auto","created_at":"2025-11-27 14:13:15","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":32963,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefigure1.png","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/dac95e2048ac548ad1cf768e.png"},{"id":96844779,"identity":"73237a63-da79-4ac7-bd03-bca4010e45e0","added_by":"auto","created_at":"2025-11-26 16:29:49","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":56621,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefigure2.png","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/d4b5ba7724616483ef7a57a3.png"},{"id":96844772,"identity":"51d9f30e-59e7-494e-97ce-a8e8c20d04f6","added_by":"auto","created_at":"2025-11-26 16:29:49","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":68941,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefigure3.png","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/83df9da3cabbc95e81cda7d7.png"},{"id":96844788,"identity":"b91ccafb-b085-463c-8525-9d1a8fabd03f","added_by":"auto","created_at":"2025-11-26 16:29:50","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":44087,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefigure4.png","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/7dbee18567fba5bdf5529c52.png"},{"id":96918946,"identity":"24273dfb-f6a3-4a38-a8b3-0d9347bb9d51","added_by":"auto","created_at":"2025-11-27 14:12:53","extension":"png","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":37882,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefigure5.png","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/1bbb093b8738579a18d84120.png"},{"id":96918811,"identity":"b960e871-e49d-4633-aeea-f94369cc91b1","added_by":"auto","created_at":"2025-11-27 14:12:39","extension":"png","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":55671,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefigure6.png","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/1b0b72122ef4674e23019b01.png"},{"id":96844786,"identity":"21b817d1-0615-41f7-a234-e479b9979f83","added_by":"auto","created_at":"2025-11-26 16:29:49","extension":"png","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":51533,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefigure7.png","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/5a1837db5d3e1fc9a02f901b.png"},{"id":96844784,"identity":"03fadffd-a01e-403d-8b76-5fa7b6306d72","added_by":"auto","created_at":"2025-11-26 16:29:49","extension":"xml","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":160584,"visible":true,"origin":"","legend":"","description":"","filename":"2cf58a05b1854cee9de756746b46525f1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/26cedf82066b1df2f186bc21.xml"},{"id":96844787,"identity":"5629c6f5-3f47-4c8d-88a9-dc773e87a560","added_by":"auto","created_at":"2025-11-26 16:29:49","extension":"html","order_by":27,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":170108,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/7b939fe906e521b0c1be6524.html"},{"id":96918986,"identity":"9e39e9ee-63f5-4bcf-a431-ef5e0a34ccab","added_by":"auto","created_at":"2025-11-27 14:12:57","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":582074,"visible":true,"origin":"","legend":"\u003cp\u003eInclusion and exclusion flowchart. 443 eligible patients for genetic prognostic study, 327 eligible patients for Pathomics study.\u003c/p\u003e","description":"","filename":"figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/5e566c93c27b7ceead64800c.png"},{"id":96844763,"identity":"6a92ac3d-91fe-43d6-8c59-c7129955d53c","added_by":"auto","created_at":"2025-11-26 16:29:49","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1499459,"visible":true,"origin":"","legend":"\u003cp\u003eThe overall framework. First, Construction of Prognostic Risk Model for MRGs.Then, Developing a Mitochondrial Genotyping-Associated Pathomics Model. Finally, Biological Mechanisms of Mitochondrial Genotyping in Pathological Features.\u003c/p\u003e","description":"","filename":"figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/b33a75740af877f26e0b2fde.png"},{"id":96844761,"identity":"0713089d-e0b3-45f8-97d5-9d013d6ca446","added_by":"auto","created_at":"2025-11-26 16:29:49","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1791846,"visible":true,"origin":"","legend":"\u003cp\u003eNMF model establishment. A.NMF rank survey;B. Consensus matrix; C.NMF feature genes heat map.\u003c/p\u003e","description":"","filename":"figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/72959dfed1318b5966833509.png"},{"id":96918954,"identity":"f9fcd68f-6e7c-4e00-bfd9-09bf24f041a7","added_by":"auto","created_at":"2025-11-27 14:12:53","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":963712,"visible":true,"origin":"","legend":"\u003cp\u003ePrognostic Value of Mitochondrial Genotyping in LUAD. A.survival curve by risk group. B.Forest plot showing univariable and multivariable Cox proportional hazards regression analyses of clusters;C.Forest plot of subgroup analysis and interaction test of cluster\u003c/p\u003e","description":"","filename":"figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/5d5417ca3504837fc6c1dc30.png"},{"id":96918876,"identity":"55361a9a-fbd6-4b6c-b3e5-f506357b3114","added_by":"auto","created_at":"2025-11-27 14:12:47","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":521324,"visible":true,"origin":"","legend":"\u003cp\u003eConstruction and Evaluation of a Mitochondrial Genotyping-Integrated Pathomics Model. A.Feature filtering line chart; B.Feature importance derived from XG-Boost model ; C/D.Performance assessment of the Pathomics model based on XG-Boost model; E/F.Performance assessment of the Pathomics model based on XG-Boost model; I/J.Differential analysis of XG-Boost based Pathomics score between clusters.\u003c/p\u003e","description":"","filename":"figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/d3d4babc107756b81b75aaa7.png"},{"id":96918108,"identity":"adda4670-ad13-4e78-a14d-0d93efe20b30","added_by":"auto","created_at":"2025-11-27 14:11:10","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":884648,"visible":true,"origin":"","legend":"\u003cp\u003ePrognostic Significance of Pathomics Features in LUAD. A.KM curve for the PS subgroups;B.Forest plot showing univariable and multivariable Cox proportional hazards of PS subgroups;C.Forest plot of subgroup analysis and interaction test of PS subgroups.\u003c/p\u003e","description":"","filename":"figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/3423d69c0fb85b5693aa4d85.png"},{"id":96919684,"identity":"8f629bce-c44b-4996-9489-bd869c7314eb","added_by":"auto","created_at":"2025-11-27 14:14:19","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":1119087,"visible":true,"origin":"","legend":"\u003cp\u003eBiological Mechanisms of Mitochondrial Genotyping in Pathological Features.A.GSVA analysis with Hallmark/KEGG pathways gene sets in PS subgroups ;B.GSVA analysis with KEGG pathways gene sets in PS subgroups ;C.Differential immune-infiltrations between PS subgroups;D.Mutational landscape in high PS subgroups ;E .Mutational landscape in high PS subgroups. (Note: (1) The color scheme is based on the GSVA analysis results: Yellow indicates pathways significantly enriched in the high PS group; Red indicates pathways significantly enriched in the low PS group; Gray indicates pathways with no significant difference. (2) A/B pathway information are derived from the KEGG database. (3) D/E: The mutation annotation of Multi_Hit indicates that the gene has multiple mutations in the same sample.)\u003c/p\u003e","description":"","filename":"figure7.png","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/a295b125bd0c1445c2e596dd.png"},{"id":99317926,"identity":"8fcddb20-0b02-4bee-a8ff-503ec9db18c2","added_by":"auto","created_at":"2025-12-31 16:30:57","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":8553427,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/53997e78-698d-477d-b647-f865ebc13d86.pdf"},{"id":96844813,"identity":"01398abb-8941-4e26-bfe2-77a9f6522c5d","added_by":"auto","created_at":"2025-11-26 16:29:54","extension":"zip","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":97558274,"visible":true,"origin":"","legend":"","description":"","filename":"supplementaryimformation.zip","url":"https://assets-eu.researchsquare.com/files/rs-7252366/v1/66b60645c6fa3cc3540969d7.zip"}],"financialInterests":"No competing interests reported.","formattedTitle":"Application of Machine Learning Models Based on H\u0026E Staining for Mitochondrial-Related Genes Classification and Prognosis of Lung Adenocarcinoma","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eLung cancer, primarily known as primary bronchogenic carcinoma, originates in the bronchial or alveolar epithelium\u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e][\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]\u003c/sup\u003e. Among the various types of lung cancer, lung adenocarcinoma (LUAD) is categorized as a type of non-small cell lung cancer\u003csup\u003e[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/sup\u003e.The primary treatment approach for LUAD involves surgical resection; however, the prognosis for patients diagnosed at advanced stages remains unfavorable\u003csup\u003e[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/sup\u003e.For those with mid to late-stage lung cancer who are no longer candidates for surgery, treatment objectives shift towards prolonging life, improving quality of life, and striving for long-term survival despite the presence of tumors\u003csup\u003e[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]\u003c/sup\u003e .Traditional markers used for LUAD diagnosis include clinical pathological features, carcinoembryonic antigen, carbohydrate antigen 125, and CT pathological examination methods. Unfortunately, these markers are increasingly inadequate for the demands of precision medicine\u003csup\u003e[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]\u003c/sup\u003e, highlighting the urgent need to identify novel prognostic markers that can offer new basis for personalized precision therapy.\u003c/p\u003e\u003cp\u003eMitochondrial-related genes play a crucial role in essential metabolic processes such as oxidative phosphorylation, the tricarboxylic acid cycle, and fatty acid β-oxidation\u003csup\u003e[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e][\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e, thereby maintaining the cell's energy supply and balance. Previous studies have shown that alterations in mitochondrial are intricately linked to LUAD cell growth, evasion of apoptosis, and the development of drug resistance\u003csup\u003e[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e][\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/sup\u003e.Mitochondrial related genes like SLC25A4 and ALDOA exhibit high expression levels in LUAD tissues and correlate with poor patient prognosis\u003csup\u003e[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/sup\u003e.Subtype clustering based on the expression characteristics of mitochondrial-related genes has emerged as a promising approach to understanding the molecular heterogeneity of LUAD. Recent studies have demonstrated that mitochondrial-related genes can aid in the classification of LUAD, revealing notable differences in immune status, metabolic features, and survival rates among distinct subtypes\u003csup\u003e[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e][\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e][\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e.In a multi-omics integrated analysis, Zhang et al.\u003csup\u003e[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]\u003c/sup\u003e developed a Mitochondrial Pathway Signature (MitoPS) based on 149 mitochondria-related pathways. This signature stratified LUAD patients into distinct subtypes. The high-MitoPS subtype exhibited mitochondrial metabolic dysfunction and an immunosuppressive tumor microenvironment. This included reduced CD8⁺ T cell infiltration and an \"immune desert\" phenotype, along with upregulation of immune checkpoint molecules and poorer survival outcomes.By contrast, the low-MitoPS subtype was associated with an\u0026ldquo; immune-activated \u0026rdquo;phenotype and better treatment response.Furthermore, the core gene NDUFB10 was identified as a key immune regulator, whose expression level influences CD8⁺ T cell infiltration and immune checkpoint protein expression, thereby further highlighting the critical role of mitochondria in regulating the tumor immune microenvironment. \u003csup\u003e[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]\u003c/sup\u003e.Mennuni M et al. revealed that elevated mitochondrial DNA (mtDNA) levels fuel tumor proliferation\u0026mdash;a characteristic feature aligning with highly mitochondrial subtypes and tumor growth phenotypes \u003csup\u003e[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]\u003c/sup\u003e.Notably, this transcriptional signature extends its influence to the tumor immune microenvironment. Zhang Y et al. identified the mitochondrial-related circular RNA cEMSY as enhancing dendritic cell and CD8⁺ T cell infiltration, subsequently improving immunotherapy response in lung cancer \u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eIn conclusion, mitochondrial-related genes are not only crucial in the onset and progression of LUAD, but their expression patterns and subtype classifications may also serve as valuable foundations for diagnostic and therapeutic strategies. Although current research has advanced our understanding of mitochondrial typing in LUAD, no studies have precisely predicted mitochondrial genotyping based on pathomics. Furthermore, the prognostic value of such predictions in LUAD patients remains unreported.\u003c/p\u003e\u003cp\u003eMajor methods for detecting mitochondrial-related genes (e.g., qPCR, RNA-seq, mass spectrometry-based metabolomics, and mtDNA sequencing) are used to analyze genes expression, mutations, and metabolic alterations\u003csup\u003e[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e][\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]\u003c/sup\u003e. However, these methods have limitations such as strict sample quality demands, mtDNA degradation or contamination risks, high costs, and difficulties in data integration \u003csup\u003e[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]\u003c/sup\u003e. Deciphering the synergistic regulatory mechanisms between mitochondrial and nuclear genes presents significant challenges for biological interpretation. H\u0026amp;E-stained sections are core imaging data used for clinical diagnosis and form the basis of pathomics analysis\u003csup\u003e[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/sup\u003e. Pathomics applies artificial intelligence to convert pathological images into high-throughput quantitative features, such as texture, morphology, and biological characteristics\u003csup\u003e[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]\u003c/sup\u003e. This enables tumor diagnosis, molecular expression assessment, and prognostic evaluation\u003csup\u003e[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e][\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/sup\u003e.Nevertheless, its application in predicting mitochondrial subtypes remains exploratory. The potential of pathomics for accurate prediction of this subtyping lacks robust validation.The key bottleneck lies in biological interpretability: Integration of genomic, transcriptomic, and mutational data is essential to decode pathological features, making this cross-omics research critically valuable.\u003c/p\u003e\u003cp\u003eThis study proposes a novel approach to classify genes associated with mitochondrial through clustering analysis, aiming to investigate the prognostic significance of mitochondrial typing. Following this, we plan to develop a pathological model that predicts mitochondrial typing by utilizing paired H\u0026amp;E images alongside RNA transcriptome samples, and we will assess its prognostic value in relation to overall survival (OS). Ultimately, we will delve into the biological interpretability of the pathological model by integrating transcriptomic and mutation data.\u003c/p\u003e"},{"header":"2. Results","content":"\u003cdiv id=\"Sec3\"\u003e\n \u003ch2\u003e2.1 Construction of Prognostic Risk Model for Mitochondrial-Related Genes\u003c/h2\u003e\n \u003cdiv id=\"Sec4\"\u003e\n \u003ch2\u003e2.1.1 NMF typing\u003c/h2\u003e\n \u003cp\u003eIn the training set, the optimal number of cluster was determined as k\u0026thinsp;=\u0026thinsp;2 based on the steepest decline in the cophenetic correlation coefficient. Subsequently, samples were divided into two major molecular subtypes using the unsupervised clustering method of NMF and labeled Cluster 1 and Cluster 2. (Fig. 3)\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec5\"\u003e\n \u003ch2\u003e2.1.2 Prognostic Value of Mitochondrial Genotyping in LUAD\u003c/h2\u003e\n \u003cp\u003eA total of 443 patients with LUAD from the TCGA database were included in the survival analysis, with patients divided into the high-risk group (n\u0026thinsp;=\u0026thinsp;148) and the low-risk group (n\u0026thinsp;=\u0026thinsp;295). Patient clinical information is shown in Table 1.Gender, pathologic_stage, smoking_status, and radiotherapy were significantly different between the two groups (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05).\u003c/p\u003e\n \u003cp\u003eAccording to survival data from the Kaplan-Meier (KM) curve analysis, cluster 1 was defined as the high-risk group, and cluster 2 was defined as the low-risk group. The median overall survival for the high-risk group was 39.03 months, while that for the low-risk group was 51.03 months. The Log-rank test indicated that the difference in overall survival rates between the high and low-risk groups was statistically significant (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05)(Fig.\u0026nbsp;4A).\u003c/p\u003e\n \u003cdiv\u003e\n \u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 1\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003eThe clinical characteristics of cluster\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eVariables\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTotal (n\u0026thinsp;=\u0026thinsp;443)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eHigh-risk group (n\u0026thinsp;=\u0026thinsp;148)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eLow-risk group (n\u0026thinsp;=\u0026thinsp;295)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.346\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e~65\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e215 (49)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e77 (52)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e138 (47)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e66~\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e228 (51)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e71 (48)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e157 (53)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGender, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e244 (55)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e60 (41)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e184 (62)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e199 (45)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e88 (59)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e111 (38)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eResidual_tumor, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.905\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e297 (67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100 (68)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e197 (67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR1/R2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e16 (4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6 (4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10 (3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRX/Unknown\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e130 (29)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e42 (28)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e88 (30)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHistologic_type, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.465\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMixed Subtype\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e97 (22)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e32 (22)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e65 (22)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNOS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e270 (61)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95 (64)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e175 (59)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOthers\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e76 (17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e21 (14)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e55 (19)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePathologic_stage, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.018\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eI/II\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e353 (80)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e108 (73)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e245 (83)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIII/IV\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e90 (20)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40 (27)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e50 (17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTumor_location, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.935\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eL-Lower\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e69 (16)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e23 (16)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e46 (16)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eL-Upper\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e111 (25)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e39 (26)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e72 (24)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR-Lower\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e85 (19)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e25 (17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e60 (20)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR-Middle\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e20 (5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7 (5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e13 (4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR-Upper\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e158 (36)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e54 (36)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e104 (35)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSmoking_status, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCurrent\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e105 (24)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e50 (34)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e55 (19)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFormer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e271 (61)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e89 (60)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e182 (62)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNonsmoker\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e67 (15)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e9 (6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e58 (20)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRadiotherapy, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.006\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e395 (89)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e123 (83)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e272 (92)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYES\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e48 (11)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e25 (17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e23 (8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eChemotherapy, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.664\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e289 (65)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e94 (64)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e195 (66)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYES\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e154 (35)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e54 (36)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100 (34)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOS, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.034\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAlive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e289 (65)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e86 (58)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e203 (69)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDead\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e154 (35)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e62 (42)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e92 (31)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOS.time, Median (Q1,Q3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e21.9 (14.47, 35.87)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e20.72 (12.38, 35.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e22.3 (15.05, 37.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.192\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eThe univariate COX analysis showed that the high-risk group was significant associated with worse OS (HR\u0026thinsp;=\u0026thinsp;1.514, 95% CI: 1.097\u0026ndash;2.092, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.012). Additionally, Residual Tumor: R1/R2 vs R0 (HR\u0026thinsp;=\u0026thinsp;4.058, 95% CI: 2.260\u0026ndash;7.287, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), Histologic Type: Others vs. Mixed Subtype (HR\u0026thinsp;=\u0026thinsp;0.427, 95% CI: 0.224\u0026ndash;0.814, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.010), Pathologic Stage (HR\u0026thinsp;=\u0026thinsp;2.560, 95% CI: 1.822\u0026ndash;3.597, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and Radiotherapy (HR\u0026thinsp;=\u0026thinsp;1.67, 95% CI: 1.077\u0026ndash;2.589, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.022) were also significant associated with OS. In multivariate analysis, after adjustment, the high-risk group (HR\u0026thinsp;=\u0026thinsp;1.463, 95% CI: 1.036\u0026ndash;2.065, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.031) remained a significant risk factor for OS. Moreover, Residual Tumor: R1/R2 vs R0 (HR\u0026thinsp;=\u0026thinsp;4.343, 95% CI: 2.182\u0026ndash;8.643, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and Pathologic Stage (HR\u0026thinsp;=\u0026thinsp;2.472, 95% CI: 1.682\u0026ndash;3.632, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) were identified as independent risk factors for OS( Fig.\u0026nbsp;4B).\u003c/p\u003e\n \u003cp\u003eIn the subgroup analysis, interaction tests revealed no significant interactions between the risk group and various subgroups including age, sex, postoperative pathological stage of residual tumor, radiotherapy, chemotherapy, smoking status, residual tumor, pathological subtypes, and tumor location subgroups (all interaction \u003cem\u003ep-\u003c/em\u003evalues\u0026thinsp;\u0026gt;\u0026thinsp;0.05). This indicates that the impact of the high-risk group on overall survival did not significantly differ across these subgroups (Fig.\u0026nbsp;4C).\u003c/p\u003e\n \u003cp\u003eThese results indicate the potential clinical significance of mitochondrial subtyping in the prognosis of LUAD patients.\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec6\"\u003e\n \u003ch2\u003e2.2 Developing a Mitochondrial Genotyping-Associated Pathomics Model\u003c/h2\u003e\n \u003cdiv id=\"Sec7\"\u003e\n \u003ch2\u003e2.2.1 Construction and Evaluation of a Mitochondrial Genotyping-Integrated Pathomics Model\u003c/h2\u003e\n \u003cp\u003eA total of 327 samples with complete pathological images were divided into a training set (n\u0026thinsp;=\u0026thinsp;229) and a validation set (n\u0026thinsp;=\u0026thinsp;98). Patients\u0026apos; baseline clinicopathological characteristics of the patients are presented in Table\u0026nbsp;2. Feature selection was first performed using the mRMR method,which selected the top 20 most relevant features. Then, recursive feature elimination (RFE)was performed, as shown in the feature selection diagram, resulting in 8 selected features (Fig.\u0026nbsp;5A).The relative importance of these selected features in the XGBoost algorithm is demonstrated in Fig.\u0026nbsp;5B, with the three most significant features beening square GLDM Dependence Variance, square First Order Minimum, and wavelet HH GLCM MCC(Fig.\u0026nbsp;5B).The pathological model showed good predictive performance. The ROC curve indicated that the AUC was 0.836 in the training set and 0.745 in the validation set(Fig.\u0026nbsp;5C, Fig.\u0026nbsp;5D).The calibration curve and Hosmer-Lemeshow goodness-of-fit test showed good agreement between the model\u0026rsquo;s predicted probabilities and actual outcomes (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05).The decision curve analysis (DCA) results indicate the model\u0026rsquo;s clinical usefulness across a wide range of scenarios.In the training set, the threshold was 0.337 with an accuracy of 0.769, sensitivity of 0.844, specificity of 0.73, Brier score of 0.132, and 95% CI of 0.782\u0026ndash;0.891. In the validation set, accuracy was 0.704, sensitivity of 0.697, specificity 0.708, Brier score 0.119, and 95% CI of 0.638\u0026ndash;0.851.Details are shown in Fig.\u0026nbsp;5E-H. The inter-group difference analysis revealed significantly higher PS in the high-risk group than in the low-risk group for both the training set (Fig.\u0026nbsp;5I) and validation set (Fig.\u0026nbsp;5J) (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001).\u003c/p\u003e\n \u003cdiv\u003e\n \u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 2\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003eThe clinical characteristics of cluster in cases of TCGA\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eVariables\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTotal (n\u0026thinsp;=\u0026thinsp;327)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTrain (n\u0026thinsp;=\u0026thinsp;229)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eValidation (n\u0026thinsp;=\u0026thinsp;98)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRisk_group, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHigh\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e110 (33.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e77 (33.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e33 (33.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLow\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e217 (66.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e152 (66.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e65 (66.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.875\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e~65\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e164 (50.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e116 (50.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e48 (49)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e66~\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e163 (49.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e113 (49.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e50 (51)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGender, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.744\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e183 (56)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e130 (56.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e53 (54.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e144 (44)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e99 (43.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e45 (45.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eResidual_tumor, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.446\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e220 (67.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e150 (65.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e70 (71.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR1/R2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e13 (4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e11 (4.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2 (2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRX/Unknown\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e94 (28.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e68 (29.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e26 (26.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHistologic_type, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.196\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMixed Subtype\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e67 (20.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e46 (20.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e21 (21.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNOS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e204 (62.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e149 (65.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e55 (56.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOthers\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e56 (17.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e34 (14.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e22 (22.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePathologic_stage, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eI/II\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e265 (81)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e186 (81.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e79 (80.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIII/IV\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e62 (19)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e43 (18.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e19 (19.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTumor_location, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.694\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eL-Lower\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e56 (17.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e39 (17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e17 (17.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eL-Upper\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e76 (23.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e56 (24.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e20 (20.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR-Lower\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e63 (19.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e46 (20.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e17 (17.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR-Middle\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e14 (4.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8 (3.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6 (6.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR-Upper\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e118 (36.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e80 (34.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e38 (38.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSmoking_status, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.864\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCurrent\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e82 (25.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e58 (25.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e24 (24.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFormer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e204 (62.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e141 (61.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e63 (64.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNonsmoker\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e41 (12.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e30 (13.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e11 (11.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRadiotherapy, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.19\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e293 (89.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e209 (91.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e84 (85.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYES\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e34 (10.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e20 (8.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e14 (14.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eChemotherapy, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.824\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e219 (67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e152 (66.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e67 (68.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYES\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e108 (33)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e77 (33.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e31 (31.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOS, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.5\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAlive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e213 (65.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e146 (63.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e67 (68.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDead\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e114 (34.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e83 (36.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e31 (31.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOS.time, Median (Q1,Q3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e21.73 (14.48, 35.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e22.13 (14.93, 34.87)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e20.32 (13.65, 35.17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.293\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec8\"\u003e\n \u003ch2\u003e2.2.2 Prognostic Significance of Pathomics Features in LUAD\u003c/h2\u003e\n \u003cp\u003eUsing the R package \u0026quot;survminer,\u0026quot; the cutoff value of 0.3047 for the pathomics score (PS) predicted by the XG-Boost model was selected to divide 327 patients into a PS high-risk group (183) and a low-risk group (144).There was no significant difference in age distribution, residual tumor, pathologic stage, tumor location, smoking status, radiotherapy, and chemotherapy between the high and low groups (all interaction \u003cem\u003ep-values\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05). However, there were significant differences in gender and histologic type between the high and low PS groups (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05).Details are shown in Table\u0026nbsp;3.The median overall survival time for the high PS group was 40.967 months, while it was 59.933 months for the low PS group. High PS was significantly correlated with worse OS (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig.\u0026nbsp;6A). In univariate analysis, we found that the high PS group was a statistically significant risk factor for OS (HR\u0026thinsp;=\u0026thinsp;1.933, 95% CI: 1.298\u0026ndash;2.879, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.001).In multivariate analysis, after adjustment for confounders, the high PS group (HR\u0026thinsp;=\u0026thinsp;1.686, 95% CI: 1.118\u0026ndash;2.542, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.013) remained a significant risk factor for OS. Datails were shown in Fig.\u0026nbsp;6B.In the subgroup analysis, we found that high PS was a significant risk factor for OS in both the \u0026lt;\u0026thinsp;65 subgroup (HR\u0026thinsp;=\u0026thinsp;1.772, 95% CI: 1.011\u0026ndash;3.107, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.046) and the \u0026gt;\u0026thinsp;65 subgroup (HR\u0026thinsp;=\u0026thinsp;2.062, 95% CI: 1.169\u0026ndash;3.637, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.012).The interaction test indicated that there was no significant interaction between PS and different subgroups, including age, gender, residual tumor, pathological stage, radiotherapy, chemotherapy, smoking status, and tumor location (all \u003cem\u003ep\u003c/em\u003e for interaction\u0026thinsp;\u0026gt;\u0026thinsp;0.05). In other words, the effect of PS on OS was consistent across the age subgroups and other analyzed groups. This suggests that PS generated by the model serves as an independent prognostic factor for LUAD. Details are shown in Fig.\u0026nbsp;6C.\u003c/p\u003e\n \u003cdiv\u003e\n \u003ctable id=\"Tab3\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 3\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003eThe clinical characteristics of PS subgroups\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eVariables\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTotal (n\u0026thinsp;=\u0026thinsp;327)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePS Low (n\u0026thinsp;=\u0026thinsp;144)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePS High (n\u0026thinsp;=\u0026thinsp;183)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e~65\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e164 (50)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e73 (51)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e91 (50)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e66~\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e163 (50)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e71 (49)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e92 (50)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGender, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.046\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e183 (56)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e90 (62)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e93 (51)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e144 (44)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e54 (38)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e90 (49)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eResidual_tumor, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.271\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e220 (67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e97 (67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e123 (67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR1/R2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e13 (4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3 (2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10 (5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRX/Unknown\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e94 (29)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e44 (31)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e50 (27)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHistologic_type, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.026\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMixed Subtype\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e67 (20)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e22 (15)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e45 (25)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNOS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e204 (62)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e90 (62)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e114 (62)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOthers\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e56 (17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e32 (22)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e24 (13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePathologic_stage, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.099\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eI/II\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e265 (81)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e123 (85)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e142 (78)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIII/IV\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e62 (19)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e21 (15)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e41 (22)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTumor_location, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.938\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eL-Lower\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e56 (17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e24 (17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e32 (17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eL-Upper\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e76 (23)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e36 (25)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40 (22)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR-Lower\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e63 (19)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e28 (19)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e35 (19)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR-Middle\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e14 (4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5 (3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e9 (5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eR-Upper\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e118 (36)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e51 (35)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e67 (37)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSmoking_status, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.156\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCurrent\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e82 (25)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e31 (22)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e51 (28)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFormer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e204 (62)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e90 (62)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e114 (62)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNonsmoker\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e41 (13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e23 (16)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e18 (10)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRadiotherapy, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.205\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e293 (90)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e133 (92)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e160 (87)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYES\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e34 (10)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e11 (8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e23 (13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eChemotherapy, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.824\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e219 (67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95 (66)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e124 (68)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYES\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e108 (33)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e49 (34)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e59 (32)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\"\u003e\n \u003ch2\u003e2.3 Potential Biological Mechanisms of Mitochondrial Genotyping in Pathological Features\u003c/h2\u003e\n \u003cp\u003eWe analyzed GSVA differences between high and low PS groups in LUAD.The results showed that in the Hallmark genes set, the high PS group was significantly enriched in the G2M_CHECKPOINT, SPERMATOGENESIS, E2F-TARGETS, and other signaling pathways (Fig. 7A, Note:Pathway information are derived from the KEGG database\u003csup\u003e[30]\u003c/sup\u003e).The top 30 pathways visualized in KEGG indicated that in the KEGG genes set, the high PS group was significantly enriched in the LYSINE_DEGRADATION, GLYOXYLATE-AND- DICARBOXYLATE, BASE-EXCISION-REPAIR, and other signaling pathways( Fig.\u0026nbsp;7B, Note:Pathway information are derived from the KEGG database\u003csup\u003e[30]\u003c/sup\u003e ).We analyzed immune cell infiltration between high and low PS groups. Among 22 immune cell types, only resting dendritic cells, resting mast cells, and activated CD4 memory T cells showed significant differences between the groups (Fig. 7C).genes mutation profiling showed that missense mutations were the most common, followed by frame shift deletions and nonsense mutations.Among them, the high PS group had a higher overall tumor mutational burden (TMB), indicating that patients in this group had more tumor mutations. Mutation rates of TP53 and TTN genes in both high and low PS groups exceeded 40%.Mutation rates of TP53 and TTN were higher in the high PS group than in the low PS group.We speculated that the high MT was associated with a poor prognosis in LUAD, possibly due to common mutations in genes such as TP53 and TTN( Fig. 7D, Fig. 7E).\u003c/p\u003e\n\u003c/div\u003e"},{"header":"3. Discussion","content":"\u003cp\u003eWe developed a mitochondria subtyping-based pathomics model for predicting the prognosis of LUAD patients through PS. Additionally, we explored the model\u0026rsquo;s underlying biological interpretability.The main findings of this study were: (1) Mitochondrial genotyping based on NMF clustering demonstrated independent prognostic value in LUAD (HR\u0026thinsp;=\u0026thinsp;1.463, 95% CI: 1.036\u0026ndash;2.065, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.031); (2) The HE pathology-based pathomics model with 8 features achieved accurate prediction of mitochondrial subtypes with an AUC of 0.836 in the training set; (3) The biological interpretability analysis of the pathomics model showed that the PS, which reflects mitochondrial genotypes, was significantly associated with several factors. These included pathways such as the G2/M checkpoint, immune cell infiltration (including resting dendritic cells, resting mast cells, and CD4 memory T cells), genes mutations like TP53 and TTN.\u003c/p\u003e\n\u003cp\u003eMitochondria, as central organelles in cellular energy metabolism, exhibit functional abnormalities that are strongly correlated with cancer progression and poor prognosis across multiple malignancies.Extensive research on mitochondrial DNA, metabolic pathways, and genes expression in LUAD has revealed associations with patient survival rates and recurrence risk. For instance, a study by Hertweck KL et al.\u003csup\u003e[\u003cspan class=\"CitationRef\"\u003e31\u003c/span\u003e]\u003c/sup\u003e found that LUAD subtypes characterized by high mitochondrial-related genes expression exhibited significant enrichment in oxidative phosphorylation (OXPHOS), the tricarboxylic acid (TCA) cycle, and metabolic reprogramming pathways. Patients in this group experienced significantly poorer outcomes. Similarly, Han et al. \u003csup\u003e[\u003cspan class=\"CitationRef\"\u003e32\u003c/span\u003e]\u003c/sup\u003e, in a study published in Nature, showed that mitochondria-specific network architectures in non-small cell lung cancer (NSCLC) are linked to metabolic dependencies. Their findings suggest that structural metabolic reorganization of mitochondria is positively associated with tumor progression.Extending beyond LUAD, Bezwada et al.\u003csup\u003e[\u003cspan class=\"CitationRef\"\u003e33\u003c/span\u003e]\u003c/sup\u003e revealed in another Nature study that metastatic sites in clear cell renal cell carcinoma (ccRCC) exhibit distinct mitochondrial metabolic reprogramming. Notably, TCA cycle activity was more prominent in metastatic lesions than in primary tumors, suggesting a fundamental shift in metabolic strategy during metastasis.Together, these findings underscore the crucial role of mitochondria in cancer metabolism and support the exploration of mitochondria-based pathomic features for prognostic modeling in LUAD.\u003c/p\u003e\n\u003cp\u003eIn summary, existing research has established the prognostic foundation of mitochondria in LUAD and other cancers.This study further confirmed the independent predictive value of mitochondrial-related genes subtyping in LUAD(HR\u0026thinsp;=\u0026thinsp;1.463).This study further substantiates the independent predictive value of mitochondrial-related genes subtyping (HR\u0026thinsp;=\u0026thinsp;1.463) through NMF clustering. This underscores the potential of mitochondrial dynamics, such as metabolic or genomic alterations, as potential biomarkers to advance precision medicine, including early intervention for high-risk patients.Future studies should validate this subtyping approach in large prospective cohorts and integrate treatment response data to achieve clinical translation.\u003c/p\u003e\n\u003cp\u003eIn recent years, pathomics models based on HE-stained pathology images have emerged as vital tools for predicting molecular features of lung cancer, owing to their non-invasive nature, cost-effectiveness, and high-throughput advantages. Current research primarily focuses on driver genes and immune markers. For instance, Yuan L et al.\u003csup\u003e[\u003cspan class=\"CitationRef\"\u003e34\u003c/span\u003e]\u003c/sup\u003ereviewed 109 relevant publications published between 2018 and 2023. Their analysis revealed that integrating pathomics with multi-omics studies was crucial for in-depth assessment and characterization of the tumor microenvironment, enabling the identification of broader tumor feature spectra.This approach ultimately facilitates the development of multimodal fusion models to precisely evaluate personalized immunotherapy efficacy and prognosis in lung cancer patients.For example, MA L et al.\u003csup\u003e[\u003cspan class=\"CitationRef\"\u003e35\u003c/span\u003e]\u003c/sup\u003e employed multi-omics modeling to analyze progressive supranuclear palsy (PSP), identifying AKT1 driver mutations and impaired TP53 tumor suppressor pathways, thereby demonstrating a link between pathological morphology and genomic instability.Similarly, the pathomics model developed in this study to predict mitochondrial-related genes subtypes shows potential clinical significance in the prognostic assessment of LUAD.This study reveals significant biological associations of the PS.Specifically, in the LUAD model, PS reflects activation states of pathways such as the G2/M checkpoint. This finding is consistent with established research by Zhang L et al.\u003csup\u003e[36]\u003c/sup\u003e, which shows that mitochondrial-related genes promote tumor progression by activating this pathway.\u003c/p\u003e\n\u003cp\u003eRegarding immune cell infiltration, this study found that the high PS group had lower infiltration levels of resting dendritic cells(DCs), resting mast cells, and CD4⁺ memory T cells ,which was associated with poorer OS (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05).These findings align with the established understanding that immune cell infiltration shapes the tumor immune microenvironment (TME) and is closely associated with tumor progression, patient survival, and treatment response\u003csup\u003e[\u003cspan class=\"CitationRef\"\u003e11\u003c/span\u003e][37][38][39][40]\u003c/sup\u003e.Specifically ,resting mast cells have been shown to be strongly associated with better OS and disease-free survival (DFS),while activated mast cells correlate with adverse prognosis\u003csup\u003e[40][\u003cspan class=\"CitationRef\"\u003e41\u003c/span\u003e]\u003c/sup\u003e. This founding aligns with the observed enrichment of resting mast cells in the low-risk PS group. another study developed an immune prognostic signature based on ferroptosis-related genes. It demonstrated that resting dendritic cells (DCs) and resting mast cells are closely associated with better OS \u003csup\u003e[42]\u003c/sup\u003e, further supporting the link between PS and immune characteristics in LUAD.\u003c/p\u003e\n\u003cp\u003eFrom a mechanistic perspective, mitochondria, as the core of cellular energy metabolism, regulate immune activity within the TME, ether directly or indirectly through metabolic reprogramming when they become dysfunctional. Mitochondrial metabolic subtypes reshape the tumor immune microenvironment through several mechanisms. These include influencing antigen presentation, immune cell activation, and immune checkpoint expression. Together, these changes modulate patient prognosis and treatment response\u003csup\u003e[\u003cspan class=\"CitationRef\"\u003e11\u003c/span\u003e][37][38][39][40]\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eSpecific mechanisms include:\u003c/p\u003e\n\u003cp\u003e(1)Metabolic reprogramming promotes an immunosuppressive microenvironment \u003csup\u003e[\u003cspan class=\"CitationRef\"\u003e11\u003c/span\u003e][37]\u003c/sup\u003e. Enhanced mitochondrial metabolism and glycolysis suppress antigen-presenting cell function. Tumor cells consume glucose and amino acids. This impairs DC antigen presentation and inhibits CD8⁺ T lymphocytes. This leads to reduced immune cell infiltration.\u003c/p\u003e\n\u003cp\u003e(2)Dendritic cell dysfunction. Multi-omics analyses indicate that in high-risk subtypes, the expression of activation markers (e.g., CD80/CD86) in dendritic cells is reduced, while the secretion of immunosuppressive cytokines (e.g., IL-10, TGF-\u0026beta;) is increased\u003csup\u003e[37]\u003c/sup\u003e. These changes are closely associated with mitochondrial metabolic reprogramming. Moreover tumor-derived lactate and ketone bodies inhibit DC maturation through epigenetic mechanisms such as histone modification. Moreover, efferocytosis and the TGF-\u0026beta; and WNT/\u0026beta;-catenin signaling pathways have been reported to suppress DC maturation\u003csup\u003e[38][39]\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e(3)Immune checkpoint dysregulation and T cell exhaustion\u003csup\u003e[37][40]\u003c/sup\u003e. Abnormal mitochondrial metabolism can also lead to dysregulated expression of immune checkpoints such as PD-L1, CTLA-4, and LAG-3. Certain checkpoint molecules are upregulated, however, the overall signaling pathway suppression leads to T cell exhaustion. This contributes to an immune \u0026quot;desert\u0026quot; phenotype, which is associated with poor prognosis and resistance to immune checkpoint inhibitors.\u003c/p\u003e\n\u003cp\u003eFurthermore, our mutation analysis showed that patients in the high-PS group had significantly higher TMB and increased mutation frequencies of TP53 and TTN. These findings are visually summarized in the waterfall plot.TP53 is one of the most frequently mutated tumor suppressor genes in LUAD (mutation rate\u0026thinsp;~\u0026thinsp;40\u0026ndash;50%) and plays a critical role in DNA repair, apoptosis, and cell cycle regulation. Mutations in TP53 may enable cells to evade normal checkpoint controls, which promotes tumorigenesis and progression\u003csup\u003e[\u003cspan class=\"CitationRef\"\u003e43\u003c/span\u003e]\u003c/sup\u003e.TTN alterations are also common in LUAD and are linked to tumor heterogeneity and patient prognosis\u003csup\u003e[\u003cspan class=\"CitationRef\"\u003e44\u003c/span\u003e]\u003c/sup\u003e. Our findings align with existing research on TP53 and TTN, further validating mutation analysis based on PS stratification. Meanwhile, TP53 and TTN mutations have been considered potential biomarkers for predicting LUAD patients\u0026rsquo; response to immune checkpoint inhibitor (ICI) treatment\u003csup\u003e[40]\u003c/sup\u003e .\u003c/p\u003e\n\u003cp\u003eThis study has some limitations often found in similar bioinformatics mining and model construction studies. (1)Retrospective data and validation cohorts: The data used in this study were entirely from public databases such as TCGA and MitoCarta 3.0, which are retrospective. The research was based on a single TCGA cohort with a cross-sectional design, which to some extent limits the generalizability of the results and makes it difficult to assess dynamic changes related to longitudinal follow-up or treatment response. Nevertheless, we confirmed the reliability of mitochondrial subtyping and the pathomics model using multiple independent cohorts;however,their clinical translation still requires validation through large-scale, multi-center prospective studies. Therefore, future efforts should focus on improving the model\u0026rsquo;s reliability in longitudinal study designs and promoting its clinical application.\u003c/p\u003e\n\u003cp\u003eDepth of mechanistic exploration: The study explored the biological mechanisms of mitochondrial subtyping through pathway enrichment, immune infiltration, and genes mutation analysis. However, the specific molecular pathways through which mitochondrial-related genes influence immune infiltration, genes mutations, and other processes remain unclear and require further in-depth in vivo and in vitro experiments for clarification. Additionally, somatic mutation analysis was conducted in TCGA-LUAD samples. The samples were stratified by high and low pathomics score (PS) and visualized using the maftools package. However ,these results remain primarily descriptive and visual, lacking advanced statistical inference or interaction testing. More systematic analyses of statistical associations and interactions are needed to fully evaluate the relationship between PS and mutational profiles.\u003c/p\u003e\n\u003cp\u003e(3)Generalizability of the model: Although the model performed well in independent datasets, there is inherent selection bias in public database populations. Whether the predictive efficacy of the model remains consistent across different ethnic groups, clinical subtypes, or LUAD patient populations, receiving different treatment regimens requires further evaluation.\u003c/p\u003e\n\u003cp\u003eIn summary, this study demonstrates that pathomics score, derived from machine learning based image analysis ,can serve as an independent prognostic factor. It allows for low-cost and accurate assessments based on routine clinical H\u0026amp;E slides. This approach can help advance precision diagnosis and treatment for LUAD. The study also reveals an important association between mitochondrial-related genes subtyping and LUAD prognosis, providing a new perspective for personalized treatment.Future research should incorporate multi-center clinical data and experimental validation to strenghten generalizability. This approach will also enhance machanistic understanding and advance precision medicine in LUAD. These findings not only help optimize treatment plans but also provide important insights for early identification of high-risk individuals and improved clinical management.\u003c/p\u003e"},{"header":"4. Methods","content":"\u003cdiv id=\"Sec12\"\u003e\n \u003ch2\u003e4.1. Data Sources and Acquisition\u003c/h2\u003e\n \u003cp\u003eThis study used data from several public databases.Genomic and pathological data for LUAD were obtained from the TCGA database (https://portal.gdc.cancer.gov/), with data usage following DBPC policies. Mitochondrial-related genes were sourced from the MitoCarta 3.0 database, which includes 1,136 human mitochondrial-related genes. Among these, 213 genes have transcriptomic data in TCGA, presented as RNA-seq log2(FPKM\u0026thinsp;+\u0026thinsp;1) format.Clinical data related to LUAD were also collected. This included survival status, survival time, age, gender, tumor stage, pathological stage, history of radiotherapy, smoking history, pathological type, tumor location, and chemotherapy history.\u003c/p\u003e\n \u003cp\u003ePrimary, treatment-naive LUAD cases with RNA-seq data were included. Samples with missing follow-up data, survival time less than 30 days, or incomplete clinical data were excluded. This process left 443 cases for survival analysis.Additionally, among the 478 samples with available pathological images, those failing quality control were removed.After quality control, 327 cases with complete clinical data, RNA-seq, and qualified pathological images were selected to construct the pathology-based omics model.Specific inclusion and exclusion criteria are shown in Fig. 1, The overall framework was summarized in Fig. 2. Because the TCGA database is publicly accessible and all patient information is anonymized, this study did not require ethics approval.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\"\u003e\n \u003ch2\u003e4.2. Construction of Prognostic Risk Model for Mitochondrial-Related Genes\u003c/h2\u003e\n \u003cdiv id=\"Sec14\"\u003e\n \u003ch2\u003e4.2.1 NMF Clustering\u003c/h2\u003e\n \u003cp\u003eWe used the \u0026quot;NMF\u0026quot; package in R for typing analysis. NMF (non-negative matrix factorization) unsupervised clustering method was used to cluster mitochondrial-related genes, obtaining samples of different pathological types. During clustering, we applied the Brunet algorithm with the rank parameter set between 2 and 5.We selected the optimal rank value based on the point at which the cophenetic correlation coefficient showed the steepest decline.For mitochondrial typing, RNA-seq data in log2(FPKM\u0026thinsp;+\u0026thinsp;1) format were used.\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec15\"\u003e\n \u003ch2\u003e4.2.2 Prognostic Analysis\u003c/h2\u003e\n \u003cp\u003ePatients were divided into high-risk and low-risk groups based on model-predicted risk scores. The impact of MRGs on patient prognosis was systematically evaluated using survival analysis, Cox regression models, and subgroup analysis. The detailed methodologies are described as follows.(1) Survival analysis for the classification was conducted using the R package \u0026quot;survival,\u0026quot; with \u0026quot;survminer\u0026quot; applied to summarize and visualize the results. Kaplan-Meier survival curves were used to illustrate the survival probabilities of different risk groups over time, with the median survival time defined as the time when survival probability reaches 50%.Log-rank tests were used to assess differences in survival between groups.(2)Univariate-multivariate analysis was performed using the \u0026quot;survival\u0026quot; package in R, and tables and forest plots for univariate and multivariate analyses were created. The Cox proportional hazards model can study the relationship between one or more risk factors and the occurrence of survival outcomes. Univariate Cox regression was used to identify factors associated with OS, while multivariate Cox regression was applied to determine independent prognostic factors and examine the combined effects of multiple variables. When HR\u0026thinsp;\u0026gt;\u0026thinsp;1, the independent variable is considered a risk factor; when HR\u0026thinsp;\u0026lt;\u0026thinsp;1, the independent variable is considered a protective factor.(3)Subgroup Analysis and Interaction Test:The \u0026quot;cmprsk,\u0026quot; \u0026quot;survival,\u0026quot; and \u0026quot;forestplot\u0026quot; packages in R were used to conduct exploratory subgroup analyses to evaluate the impact of risk groups (high-risk vs low-risk) on patient prognosis across different covariate subgroups. Likelihood ratio tests were employed to assess the interaction between risk groups and other covariates.\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\"\u003e\n \u003ch2\u003e4.3. Pathological Methodology\u003c/h2\u003e\n \u003cdiv id=\"Sec17\"\u003e\n \u003ch2\u003e4.3.1 Image Acquisition, Segmentation, and Feature Extraction\u003c/h2\u003e\n \u003cp\u003eImage acquisition : Pathological images were downloaded from the TCGA (https://tcga-data.nci.nih.gov/tcga/) database. These images consist of formalin-fixed and paraffin-embedded pathological tissue slices in svs format, with a maximum magnification of 20\u0026times; or 40\u0026times; (H\u0026amp;E-stained histopathological images)\u003csup\u003e[20][21]\u003c/sup\u003e.\u003c/p\u003e\n \u003cp\u003eTrain/validation split : Patients were randomly allocated into a training set (n\u0026thinsp;=\u0026thinsp;229) and a validation set (n\u0026thinsp;=\u0026thinsp;98) in a 7:3 ratio \u003csup\u003e[22][23]\u003c/sup\u003e. A fixed random seed was used to ensure reproducibility.\u003c/p\u003e\n \u003cp\u003ePathological image processing and segmentation: The OTSU algorithm (https://opencv.org/) was used to segment the tissue area of the pathological slice. Also called the maximum inter-class variance method, this thresholding algorithm binarizes the image by separating the background from the tissue region of interest \u003csup\u003e[24]\u003c/sup\u003e.The 40\u0026times; images were segmented into multiple 1024\u0026times;1024-pixel sub-images. The 20\u0026times; images were segmented into multiple 512\u0026times;512-pixel sub-images and then upsampled to 1024\u0026times;1024 pixels.Pathologists reviewed the images and excluded those with poor quality (contamination, blurriness, or more than 50% blank areas) sub-images. Ten sub-images were randomly selected from each pathological image for subsequent analysis \u003csup\u003e[20][21]\u003c/sup\u003e.\u003c/p\u003e\n \u003cp\u003eFeature extraction: The PyRadiomics open-source package (https://pyradiomics.readthedocs.io/en/latest/) was used to standardize each sub-image and extract 93 original features, including first-order and second-order features. Additionally, higher-order features were extracted, such as wavelet (LL, LH, HL, HH), square, square root, logarithm, exponential, gradient, and LBP2D, resulting in a total of 1,023 features.After extracting features from the 10 sub-images of each patient\u0026apos;s pathological image, their average was calculated and used as the representative pathological feature for subsequent data analysis\u003csup\u003e[19][25][26]\u003c/sup\u003e.\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec18\"\u003e\n \u003ch2\u003e4.3.2 mRMR_RFE Feature Selection\u003c/h2\u003e\n \u003cp\u003eThe best feature subset was selected using the mRMR(maximum relevance minimum redundancy) algorithm and the RFE(recursive feature elimination) algorithm. The mRMR algorithm selects features by considering both the correlation between features and the target variable, as well as the correlation among features themselves. The method uses mutual information to measure relevance and redundancy. Relevance is calculated by the information gain of each feature with the category, while redundancy is measured by summing mutual information among features and dividing by the square of the feature subset size.\u003c/p\u003e\n \u003cp\u003eRFE feature selection ranks predictors before modeling and removes the least important features sequentially.The goal is to find a subset of predictors that can be used to generate an accurate model. The model is trained iteratively. After each training, n least important features are removed. The model is retrained with the remaining features, and their importance is reassessed. This process repeats until the optimal feature subset is obtained.The mRMR method selects the top 20 features, which are then refined using RFE feature selection.RFE with 10-fold cross-validation was used to further refine the 20 features. Based on the performance shown in Fig.\u0026nbsp;5A, the accuracy peaked at approximately 0.62 when the number of features was dropped to 8.Consequently, 8 features were retained for modeling \u003csup\u003e[27][28][29]\u003c/sup\u003e.\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec19\"\u003e\n \u003ch2\u003e4.3.3 XGBoost model establishment\u003c/h2\u003e\n \u003cp\u003eUsing the \u0026quot;caret\u0026quot; package in R, we built a model on the training set with features selected by mRMR_RFE. The model was developed using the XGBoost algorithm, which stands for Extreme Gradient Boosting.XGBoost is a boosting algorithm based on the gradient boosting framework that iteratively adds decision trees to reduce prediction errors. Compared to traditional gradient boosting machines (GBM), XGBoost offers significant improvements in computational efficiency, parallel processing, and model flexibility, resulting in more robust performance in classification and regression tasks.XGBoost improves performance through several key features: (1) introducing a regularization term to prevent overfitting; (2) using an approximate greedy algorithm during tree construction that evaluates gains of all features at each split and selects the one with the highest gain; (3) considering weighted quantiles of feature values when splitting nodes to enhance robustness; and (4) automatically handling missing values by assigning them to child nodes that increase gain after splitting.Using the XGBoost algorithm, we modeled the selected pathological features to predict risk group classification. Pathological model outputs included risk probabilities and pathomics scores (PS).\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec20\"\u003e\n \u003ch2\u003e4.3.4 XGBoost model evaluation\u003c/h2\u003e\n \u003cp\u003eThe R packages \u0026quot;pROC,\u0026quot; \u0026quot;measures,\u0026quot; \u0026quot;ResourceSelection,\u0026quot; \u0026quot;rms,\u0026quot; and \u0026quot;rmda\u0026quot; were used to evaluate the model\u0026apos;s performance. Evaluation metrics included accuracy (ACC), specificity (SPE), sensitivity (SEN), positive predictive value (PPV), and negative predictive value (NPV).The x-axis of the receiver operating characteristic curve (ROC) represents the false positive rate (1 - specificity), and the y-axis represents the true positive rate (sensitivity).A larger receiver operating characteristic(ROC)-area under the curve(AUC) means a greater area under the curve. When the curve bulges more towards the upper left corner, it indicates better model performance.Calibration curves were drawn and the Hosmer-Lemeshow goodness-of-fit test was conducted to evaluate the calibration of the pathological prediction model; the Brier score was used to quantify the overall performance of the pathological prediction model, with smaller values indicating better consistency in model predictions; decision curves (DCA) were plotted to show the clinical benefit of the pathological prediction model.\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec21\"\u003e\n \u003ch2\u003e4.4. Mechanism Analysis of Pathomics\u003c/h2\u003e\n \u003cdiv id=\"Sec22\"\u003e\n \u003ch2\u003e4.4.1 GSVA Enrichment Analysis between High and Low PS Groups\u003c/h2\u003e\n \u003cp\u003eThe expression matrix of 327 LUAD patients from TCGA was used to calculate pathway enrichment scores for KEGG and Hallmark genes sets in each sample. This calculation was performed using GSVA.The R package \u0026quot;limma\u0026quot; was used for differential analysis between high and low PS groups. A total of 184 KEGG pathways and 50 Hallmark genes sets were analyzed.\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec23\"\u003e\n \u003ch2\u003e4.4.2 Analysis of Differences in Immune Cell Abundance\u003c/h2\u003e\n \u003cp\u003eThe \u0026apos;CIBERSORT\u0026apos; package was used to analyze immune infiltration, and the Wilcoxon rank-sum test was applied to assess differences in immune cell infiltration between high and low PS groups.\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec24\"\u003e\n \u003ch2\u003e4.4.3 Genes Mutation Analysis between High and Low PS Groups\u003c/h2\u003e\n \u003cp\u003eThe relationship between the mutation profile of TCGA-LUAD patients and model-predicted PS groups was assessed using available somatic mutation data. Mutation data were downloaded from the TCGA data portal, with 322 samples overlapping with pathological data. Somatic variants were stored in mutation annotation format (MAF). The R package maftools was employed to analyze the mutation data and visualize the top 15 genes with the highest mutation frequencies.\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec25\"\u003e\n \u003ch2\u003e4.5. Statistical analysis\u003c/h2\u003e\n \u003cp\u003eData processing, statistical analyses, and visualization were performed using R software (v4.1.0). Categorical variables were compared using chi-square tests, while continuous variables were analyzed using Wilcoxon rank-sum tests or t-tests. For the pathological intersection sample set from TCGA, samples with complete pathological images, genes expression matrices, and clinical data were screened using the caret (v6.0-93) and survminer (v0.4.9) packages. Dataset partitioning was performed using the caret package with a fixed random seed (set.seed(123)); zero-variance features were removed, and samples were randomly split into training and validation sets at a 7:3 ratio. The 1,023 radiomic features extracted from the validation set using pyradiomics (v3.0.1) were standardized by Z-score normalization based on the (mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard) deviation of the training set.Inter-group differences in clinical characteristics were assessed using independent t-tests (continuous variables) or chi-square tests (categorical variables).Inter-group differences in XGBoost predictions were analyzed using the ggpubr package.Optimal cutoff values for PS were computed using survminer to create high/low binary groups. Statistical significance was defined as \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05, with notation as follows: ns (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026ge;\u0026thinsp;0.05), * (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05), ** (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.01), *** (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and **** (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.0001).\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eSijuan Xu :conceptuatization,experimental design(lead),writing-original draft(lead),and writing -review and deiting (equal).Jianzhong Lu:conceptualization (equal), data collection and analysis (lead), writing \u0026ndash; original draft (equal), and writing \u0026ndash; review and editing (equal).All authors read and approved the final manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eWe thank the authors of TCGA and MitoCarta 3.0 databas for making their data public for analysis.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe data used in this study were sourced from two publicly accessible databases: TCGA (https://portal.gdc.cancer.gov/) and MitoCarta 3.0 (https://www.broadinstitute.org/files/shared/metabolism/mitocarta/human.mitocarta3.0.html).The data used in this study came from the publicly accessible databases TCGA and MitoCarta 3.0.We confirm that other researchers can obtain the same datasets from these repositories for further analysis.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eSiegel, R. L., Miller, K. D., Fuchs, H. E., Jemal, A. \u0026amp; Cancer statistics CA: A Cancer Journal for Clinicians, 72(1), 7\u0026ndash;33. (2022). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3322/caac.21708\u003c/span\u003e\u003cspan address=\"10.3322/caac.21708\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e(2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLeiter, A., Veluswamy, R. R. \u0026amp; Wisnivesky, J. P. The global burden of lung cancer: current status and future trends. \u003cem\u003eNat. Reviews Clin. Oncol.\u003c/em\u003e, \u003cb\u003e20\u003c/b\u003e(9), 624\u0026ndash;639. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41571-023-00798-3(2023\u003c/span\u003e\u003cspan address=\"10.1038/s41571-023-00798-3(2023\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen, H. Correlation of LOXL2 expression in non-small cell lung cancer with immunotherapy. \u003cem\u003eInt. J. Clin. Exp. Pathol.\u003c/em\u003e \u003cb\u003e17\u003c/b\u003e (9), 268\u0026ndash;286. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.62347/zieg9007\u003c/span\u003e\u003cspan address=\"10.62347/zieg9007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhao, H., Li, W., Li, X., Ding, Z. \u0026amp; Zhao, S. Survival benefit of resection surgery for lung adenocarcinoma with bone metastases and a post-operative prognosis nomogram establishment and validation. \u003cem\u003eJ. Thorac. Disease\u003c/em\u003e, \u003cb\u003e14\u003c/b\u003e(12), 4877\u0026ndash;4893. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.21037/jtd-22-1514(2022\u003c/span\u003e\u003cspan address=\"10.21037/jtd-22-1514(2022\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang, Z. et al. Dynamics of Serum Tumor Markers Can Serve as a Prognostic Biomarker for Chinese Advanced Non-small Cell Lung Cancer Patients Treated With Immune Checkpoint Inhibitors. \u003cem\u003eFront. Immunol.\u003c/em\u003e \u003cb\u003e11\u003c/b\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fimmu.2020.01173\u003c/span\u003e\u003cspan address=\"10.3389/fimmu.2020.01173\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWu, P. et al. FOXK2 in skeletal muscle development: a new pathogenic gene for congenital myopathy with ptosis. \u003cem\u003eEMBO Mol. Med.\u003c/em\u003e, \u003cb\u003e17\u003c/b\u003e(7), 1599\u0026ndash;1630. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s44321-025-00247-x(2025\u003c/span\u003e\u003cspan address=\"10.1038/s44321-025-00247-x(2025\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBALLARD, J. W. O. \u0026amp; MELVIN, R. G. Linking the mitochondrial genotype to the organismal phenotype. \u003cem\u003eMol. Ecol.\u003c/em\u003e, \u003cb\u003e19\u003c/b\u003e(8), 1523\u0026ndash;1539. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/j.1365-294x.2010.04594.x(2010\u003c/span\u003e\u003cspan address=\"10.1111/j.1365-294x.2010.04594.x(2010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDong, L., Gopalan, V., Holland, O. \u0026amp; Neuzil, J. Mitocans Revisited: Mitochondrial Targeting as Efficient Anti-Cancer Therapy. \u003cem\u003eInt. J. Mol. Sci.\u003c/em\u003e \u003cb\u003e21\u003c/b\u003e (21), 7941. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/ijms21217941(\u003c/span\u003e\u003cspan address=\"10.3390/ijms21217941(\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJeon, J. H. et al. Migration and invasion of drug-resistant lung adenocarcinoma cells are dependent on mitochondrial activity. \u003cem\u003eExp. Mol. Med.\u003c/em\u003e \u003cb\u003e48\u003c/b\u003e (12), e277\u0026ndash;e277. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/emm.2016.129\u003c/span\u003e\u003cspan address=\"10.1038/emm.2016.129\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang, X. et al. Multi-omic and machine learning analysis of mitochondrial RNA modification genes in lung adenocarcinoma for prognostic and therapeutic implications. \u003cem\u003eTransl Oncol.\u003c/em\u003e \u003cb\u003e53\u003c/b\u003e, 102306. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://doi.org/10.1016/j.tranon.2025.102306\u003c/span\u003e\u003cspan address=\"10.1016/j.tranon.2025.102306\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2025).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang, P. et al. Mitochondrial Pathway Signature (MitoPS) predicts immunotherapy response and reveals NDUFB10 as a key immune regulator in lung adenocarcinoma. \u003cem\u003eJ. Immunother. Cancer\u003c/em\u003e. \u003cb\u003e13\u003c/b\u003e (7), e012069. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1136/jitc-2025-012069\u003c/span\u003e\u003cspan address=\"10.1136/jitc-2025-012069\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2025).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMennuni, M. et al. High mitochondrial DNA levels accelerate lung adenocarcinoma progression. \u003cem\u003eSci. Adv.\u003c/em\u003e \u003cb\u003e10\u003c/b\u003e (44). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1126/sciadv.adp3481(\u003c/span\u003e\u003cspan address=\"10.1126/sciadv.adp3481(\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang, Y. et al. The circRNA cEMSY Induces Immunogenic Cell Death and Boosts Immunotherapy Efficacy in Lung Adenocarcinoma. \u003cem\u003eCancer Res.\u003c/em\u003e \u003cb\u003e85\u003c/b\u003e (3), 497\u0026ndash;514. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1158/0008-5472.can-24-1484(\u003c/span\u003e\u003cspan address=\"10.1158/0008-5472.can-24-1484(\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWen, J. J., Cummins, C. B., Williams, T. P. \u0026amp; Radhakrishnan, R. S. The Genetic Evidence of Burn-Induced Cardiac Mitochondrial Metabolism Dysfunction. \u003cem\u003eBiomedicines\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e (12), 566. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/biomedicines8120566(\u003c/span\u003e\u003cspan address=\"10.3390/biomedicines8120566(\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJarrett, S. G., Rohrer, B., Perron, N. R., Beeson, C. \u0026amp; Boulton, M. E. Assessment of mitochondrial damage in retinal cells and tissues using quantitative polymerase chain reaction for mitochondrial DNA damage and extracellular flux assay for mitochondrial respiration activity. \u003cem\u003eMethods Mol. Biol.\u003c/em\u003e, \u003cb\u003e935\u003c/b\u003e:227\u0026ndash;243 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://doi.org/10.1007/978-1-62703-080-9_16(2013\u003c/span\u003e\u003cspan address=\"10.1007/978-1-62703-080-9_16(2013\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen, Y. et al. Multiphoton microscopy as a diagnostic imaging modality for pancreatic neoplasms without hematoxylin and eosin stains. \u003cem\u003eJ. Biomed. Opt.\u003c/em\u003e, \u003cb\u003e19\u003c/b\u003e(9):96008. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://doi.org/10.1117/1.JBO.19.9.096008(2014\u003c/span\u003e\u003cspan address=\"10.1117/1.JBO.19.9.096008(2014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChang, L. et al. Advancing precision medicine: the transformative role of artificial intelligence in immunogenomics, radiomics, and pathomics for biomarker discovery and immunotherapy optimization. \u003cem\u003eCancer Biology Med.\u003c/em\u003e, 1\u0026ndash;15 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.20892/j.issn.2095-3941.2024.0376(2025\u003c/span\u003e\u003cspan address=\"10.20892/j.issn.2095-3941.2024.0376(2025\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKunchur, N. N. \u0026amp; Mosta\u0026ccedil;o-Guidolin, L. B. Development of an image classification pipeline for atherosclerotic plaques assessment using supervised machine learning. \u003cem\u003eBMC Bioinform.\u003c/em\u003e, \u003cb\u003e23\u003c/b\u003e(1). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12859-022-05059-1(2022\u003c/span\u003e\u003cspan address=\"10.1186/s12859-022-05059-1(2022\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu, K. \u0026amp; Hu, J. Classification of acute myeloid leukemia M1 and M2 subtypes using machine learning. \u003cem\u003eComput. Biol. Med.\u003c/em\u003e, \u003cb\u003e147\u003c/b\u003e:105741. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://doi.org/10.1016/j.compbiomed.2022.105741(2022\u003c/span\u003e\u003cspan address=\"10.1016/j.compbiomed.2022.105741(2022\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen, L., Zeng, H., Zhang, M., Luo, Y. \u0026amp; Ma, X. Histopathological image and gene expression pattern analysis for predicting molecular features and prognosis of head and neck squamous cell carcinoma. \u003cem\u003eCancer Med.\u003c/em\u003e \u003cb\u003e10\u003c/b\u003e (13), 4615\u0026ndash;4628. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://doi.org/10.1002/cam4.3965\u003c/span\u003e\u003cspan address=\"10.1002/cam4.3965\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZeng, H., Chen, L., Zhang, M., Luo, Y. \u0026amp; Ma X.Integration of histopathological images and multi-dimensional omics analyses predicts molecular features and prognosis in high-grade serous ovarian cancer. \u003cem\u003eGynecol. Oncol.\u003c/em\u003e \u003cb\u003e163\u003c/b\u003e (1), 171\u0026ndash;180. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ygyno.2021.07.015\u003c/span\u003e\u003cspan address=\"10.1016/j.ygyno.2021.07.015\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eQu, W. F. et al. Development of a deep pathomics score for predicting hepatocellular carcinoma recurrence after liver transplantation. \u003cem\u003eHep. Intl.\u003c/em\u003e \u003cb\u003e17\u003c/b\u003e (4), 927\u0026ndash;941. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s12072-023-10511-2\u003c/span\u003e\u003cspan address=\"10.1007/s12072-023-10511-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCai, X. et al. Development of a Pathomics-Based Model for the Prediction of Malignant Transformation in Oral Leukoplakia. Laboratory investigation; a journal of technical methods and pathology, 103(8), 100173. https://doi.org/10.1016/j.labinv.2023.100173, LI FP,et al.MRI\u0026ndash;based radiomics nomogram for predicting temporal lobe injury after radiotherapy in nasopharyngeal carcinoma. European Radiology; 32:1106\u0026ndash;1114. https://doi.org/10.1007/s00330-021-08254-5 (2022) (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi, M. L. et al. A clinical-radiomics nomogram for the preoperative prediction of lymph node metastasis in colorectal cancer. \u003cem\u003eJ. Transl Med.\u003c/em\u003e \u003cb\u003e18\u003c/b\u003e, 46. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12967-020-02215-0\u003c/span\u003e\u003cspan address=\"10.1186/s12967-020-02215-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi, Y., Chen, C. M., Li, W. W., Shao, M. T. \u0026amp; Dong, Y., Zhang QC.Radiomic features based on pyradiomics predict CD276 expression associated with breast cancer prognosis.Heliyon 10, e37345. (2024). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.heliyon.2024.e37345\u003c/span\u003e\u003cspan address=\"10.1016/j.heliyon.2024.e37345\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang, X. et al. Weakly Supervised Deep Learning for Whole Slide Lung Cancer Image Analysis. \u003cem\u003eIEEE Trans. Cybernetics\u003c/em\u003e. \u003cb\u003e50\u003c/b\u003e (9), 3950\u0026ndash;3962. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/tcyb.2019.2935141\u003c/span\u003e\u003cspan address=\"10.1109/tcyb.2019.2935141\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSaednia, K. et al. Quantitative digital histopathology and machine learning to predict pathological complete response to chemotherapy in breast cancer patients using pre-treatment tumor biopsies. \u003cem\u003eSci. Rep.\u003c/em\u003e, \u003cb\u003e12\u003c/b\u003e(1). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-022-13917-4(2022\u003c/span\u003e\u003cspan address=\"10.1038/s41598-022-13917-4(2022\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNishio, M., Nishio, M., Jimbo, N. \u0026amp; Nakane, K. Homology-Based Image Processing for Automatic Classification of Histopathological Images of Lung Tissue. \u003cem\u003eCancers\u003c/em\u003e \u003cb\u003e13\u003c/b\u003e (6), 1192. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/cancers13061192\u003c/span\u003e\u003cspan address=\"10.3390/cancers13061192\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKanehisa, M., Furumichi, M., Sato, Y., Matsuura, Y. \u0026amp; Ishiguro-Watanabe M.KEGG: biological systems database as a model of the real world. \u003cem\u003eNucleic Acids Res. 53\u003c/em\u003e. D672\u0026ndash;D677. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://doi.org/10.1093/nar/gkae909\u003c/span\u003e\u003cspan address=\"10.1093/nar/gkae909\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2025).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHertweck, K. L. et al. Clinicopathological significance of unraveling mitochondrial pathway alterations in non-small‐cell lung cancer. \u003cem\u003eFASEB J.\u003c/em\u003e, \u003cb\u003e37\u003c/b\u003e(7). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1096/fj.202201724rr(2023\u003c/span\u003e\u003cspan address=\"10.1096/fj.202201724rr(2023\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHan, M. et al. Spatial mapping of mitochondrial networks and bioenergetics in lung cancer. \u003cem\u003eNature\u003c/em\u003e, \u003cb\u003e615\u003c/b\u003e(7953), 712\u0026ndash;719. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41586-023-05793-3(2023\u003c/span\u003e\u003cspan address=\"10.1038/s41586-023-05793-3(2023\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBezwada, D. et al. et,al. Mitochondrial complex I promotes kidney cancer metastasis. \u003cem\u003eNature\u003c/em\u003e, \u003cb\u003e633\u003c/b\u003e(8031), 923\u0026ndash;931. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://doi.org/10.1038/s41586-024-07812-3(2024\u003c/span\u003e\u003cspan address=\"10.1038/s41586-024-07812-3(2024\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYuan, L. et al. Unveiling the landscape of pathomics in personalized immunotherapy for lung cancer: a bibliometric analysis. \u003cem\u003eFront. Oncol.\u003c/em\u003e \u003cb\u003e14\u003c/b\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fonc.2024.1432212\u003c/span\u003e\u003cspan address=\"10.3389/fonc.2024.1432212\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMa, L. et al. An advanced molecular medicine case report of a rare human tumor using genomics, pathomics, and radiomics. \u003cem\u003eFront. Genet.\u003c/em\u003e 13. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fgene.2022.987175\u003c/span\u003e\u003cspan address=\"10.3389/fgene.2022.987175\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang, L. et al. Xanthatin Induces Cell Cycle Arrest at G2/M Checkpoint and Apoptosis via Disrupting NF-κB Pathway in A549 Non-Small-Cell Lung Cancer Cells. Molecules, 17(4), 3736\u0026ndash;3750. https://doi.org/10.3390/molecules17043736, Optimized Dynamic Network Biomarker Deciphers a High-Resolution Heterogeneity Within Thyroid Cancer Molecular Subtypes. Med Research, https://doi.org/10.1002/mdr2.70004(2025),,\u0026amp;Clinical prognostication and immunotherapy response prediction in esophageal squamous cell carcinoma using the DNA damage repair-associated signature. Environmental toxicology, 39(5), 2803\u0026ndash;2816. https://doi.org/10.1002/tox.24155(2024),, Integration of the bulk transcriptome and single-cell transcriptome reveals efferocytosis features in lung adenocarcinoma prognosis and immunotherapy by combining deep learning. Cancer cell international, 24(1), 388. https://doi.org/10.1186/s12935-024-03571-3(2024),,\u0026amp; Integrating multi-omics and machine learning survival frameworks to build a prognostic model based on immune function and cell death patterns in a lung adenocarcinoma cohort. Frontiers in immunology, 15, 1460547. https://doi.org/10.3389/fimmu.2024.1460547(2024) (2012).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang, C. et al. Patterns of immune infiltration in lung adenocarcinoma revealed a prognosis-associated microRNA-mast cells network. Human cell, 33(1), 205\u0026ndash;219. https://doi.org/10.1007/s13577-019-00300-1, Development and Validation of a Robust Ferroptosis-Related Prognostic Signature in Lung Adenocarcinoma. Frontiers in cell and developmental biology, 9, 616271. https://doi.org/10.3389/fcell.2021.616271 (2021) (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFrille, A. et al. TP53 co-mutations in advanced lung adenocarcinoma: comparative bioinformatic analyses suggest ambivalent character on overall survival alongside KRAS, STK11 and KEAP1 mutations. \u003cem\u003eFront. Oncol.\u003c/em\u003e \u003cb\u003e14\u003c/b\u003e, 1357583. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fonc.2024.1357583\u003c/span\u003e\u003cspan address=\"10.3389/fonc.2024.1357583\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu, Z. et al. Heterogeneous pattern of gene expression driven by TTN mutation is involved in the construction of a prognosis model of lung squamous cell carcinoma. \u003cem\u003eFront. Oncol.\u003c/em\u003e \u003cb\u003e13\u003c/b\u003e, 916568. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fonc.2023.916568\u003c/span\u003e\u003cspan address=\"10.3389/fonc.2023.916568\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Prognostic model, Lung adenocarcinoma, Mitochondrial-related genes, Pathomic, Machine learning algorithm","lastPublishedDoi":"10.21203/rs.3.rs-7252366/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7252366/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eLung adenocarcinoma (LUAD) is a leading cause of cancer-related mortality worldwide, necessitating the identification of reliable prognostic biomarkers. This study innovatively combines mitochondrial-related genes classification with pathological histology. We developed machine learning models to predict LUAD prognosis with improved accuracy.We utilized RNA sequencing data from 443 LUAD patients and paired H\u0026amp;E-stained pathological images from 327 patients, both sourced from The Cancer Genome Atlas (TCGA) database. Using a non-negative matrix factorization (NMF) clustering algorithm, we clustered 213 mitochondrial-related genes into high-risk (Cluster 1) and low-risk (Cluster 2) categories.Survival analysis confirmed that the high-risk group is an independent risk factor for overall survival (OS) (HR\u0026thinsp;=\u0026thinsp;1.463, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.031). In parallel, we constructed an eight-feature pathomics model using various machine learning techniques, achieving a strong predictive performance with an area under the curve (AUC) of 0.836. The pathomics score (PS) derived from this model was identified as an independent prognostic factor (HR\u0026thinsp;=\u0026thinsp;1.686, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.013). Moreover, functional enrichment analysis revealed that the high-PS group is associated with several critical pathways and alterations, including activation of the G2M checkpoint pathway, upregulated lysine degradation, reduced resting dendritic cell infiltration, TP53/TTN mutations, and increased tumor mutation burden.This study represents a novel integration of mitochondrial metabolic genes classification with pathological histology, demonstrating that pathomics features indicative of mitochondrial subtypes serve as potential prognostic markers for LUAD and might suggest promising avenues for personalized treatment strategies.\u003c/p\u003e","manuscriptTitle":"Application of Machine Learning Models Based on H\u0026amp;E Staining for Mitochondrial-Related Genes Classification and Prognosis of Lung Adenocarcinoma","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-26 16:29:44","doi":"10.21203/rs.3.rs-7252366/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"34551972-6e47-4644-b04c-cc8771b5f1f1","owner":[],"postedDate":"November 26th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":58308231,"name":"Health sciences/Biomarkers"},{"id":58308232,"name":"Biological sciences/Cancer"},{"id":58308233,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":58308234,"name":"Health sciences/Oncology"}],"tags":[],"updatedAt":"2025-12-30T06:09:42+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-26 16:29:44","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7252366","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7252366","identity":"rs-7252366","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}