Histone deacetylase 9 (HDAC9) exacerbates atherosclerosis through inflammatory pathways, yet its specific role in cholesterol metabolism remains to be fully elucidated. Here, we investigated whether HDAC9 promotes atherosclerosis by impairing hepatic cholesterol excretion via bile acid synthesis. Using ApoE-/- and ApoE-/-PSRC1-/- mice fed a high-fat diet (HFD), we observed that PSRC1 deficiency or HFD feeding up-regulated hepatic HDAC9 expression, concomitant with suppression of the rate-limiting bile acid enzyme CYP7A1. Pharmacological inhibition of HDAC9 (TMP195) or liver-targeted AAV8-shHdac9 knockdown restored CYP7A1 expression, reduced plasma LDL-C, and attenuated aortic plaque burden in HFD-fed ApoE-/- mice. In vitro assays in Hep1-6 and HepG2 hepatocytes confirmed that Hdac9 knockdown attenuated free fatty acid-induced cholesterol accumulation, while overexpression exacerbated it. Mechanistically, HDAC9 transcriptionally represses Cyp7a1, as the cholesterol-lowering effect of Hdac9 knockdown was abolished by concomitant Cyp7a1 silencing. Collectively, our findings indicate that HDAC9 promotes atherosclerosis by transcriptionally repressing Cyp7a1 and impairing bile acid synthesis, independent of its inflammatory roles. These results highlight hepatic HDAC9 as a promising therapeutic target for hypercholesterolemia and advocate for the development of liver-directed HDAC9 inhibitors.

e16052 Background: SYS6010 is a novel antibody-drug conjugate (ADC), consisting of an anti-epidermal growth factor receptor (EGFR) humanized IgG1 monoclonal antibody linked to topoisomerase I inhibitor (JS-1). Enlonstobart is a fully humanized anti-PD-1 IgG4 monoclonal antibody. In most recurrent or metastatic (R/M) esophageal cancer (EC) and selected advanced solid tumors, PD-1 based chemo-immunotherapy is standard treatment yet limited by resistance. Mechanistically, certain EGFR-targeted ADCs can complement PD-1 immunotherapy and hold promise for enhancing anti-tumor activity. This phase I/II study (ChiCTR2400089402) is to evaluate the safety, tolerability, pharmacokinetic profile and preliminary efficacy of SYS6010 combined with Enlonstobart ± chemotherapy in patients with EGFR and ALK wild type locally advanced or metastatic non-small cell lung cancer and other advanced solid tumors. Here, we present the preliminary safety and efficacy results of SYS6010 combined with Enlonstobart in patients with R/M EC from the phase II portion. Methods: In the phase II EC cohort, eligible patients were those aged 18 to 75 years with R/M EC who had not received prior systemic therapy, or who had received adjuvant/neoadjuvant therapy with disease progression occurring at least 6 months after completion of treatment. Patients received SYS6010 at 3.6 mg/kg and Enlonstobart at 240 mg every 2 weeks (Q2W). Primary endpoints were safety, tolerability and objective response rate (ORR) assessed by investigators. Results: As of January 05, 2026 (data cut-off), 32 patients were enrolled. The median age was 63 years (range, 36-72) and 81.3% of patients were male. The median treatment exposure was 19.64 weeks (IQR, 12.43-25.86). 96.9% of patients experienced at least one treatment-emergent adverse event (TEAE). The incidence of grade ≥3 TEAEs was 31.3%. The most common grade ≥3 TEAEs were neutropenia (12.5%), leukopenia (6.3%) and anemia (6.3%). One patient experienced death, which was assessed as unrelated to treatment. No patients discontinued treatment due to TEAEs. 29 patients were evaluable for efficacy, with confirmed ORR and disease control rate of 65.5% (95% CI, 45.7-82.1) and 93.1% (95% CI, 77.2-99.2), respectively. The 6-month progression-free survival (PFS) rate was 68.2% (95% CI, 40.6-85.0). The confirmed 3-month duration of response (DOR) rate was 91.7% (95% CI, 53.9-98.8). Median PFS, median DOR and overall survival were not yet mature. Conclusions: This study demonstrates that the combination therapy of SYS6010 and Enlonstobart has the potential to represent a novel and effective treatment option for patients with R/M EC. The promising clinical activity, together with a manageable safety profile, supports further clinical development of this regimen. Clinical trial information: ChiCTR2400089402.

The differential response of patients with unresectable colorectal cancer liver metastases to bevacizumab plus chemotherapy highlights the urgent need for high-specificity predictive biomarkers. Recent studies suggest that pretreatment contrast-enhanced abdominal CT can noninvasively predict subtypes of liver metastases resistant to bevacizumab-based combination therapy. However, it remains unclear whether integrating intelligent radiological imaging with computational pathology can further enhance predictive performance while simultaneously improving the model’s interpretability and elucidating the underlying mechanisms of bevacizumab resistance. Here, we developed a machine learning-based integrative model that combines abdominal CT radiomics with deep learning-based computational pathology to predict the efficacy of bevacizumab combination therapy. Using a multi-center cohort of 319 patients, we allocated patients to the following cohorts: training (n = 120), internal test (n = 50), external test (n = 47), transcriptomic validation (n = 14), and pathological validation (n = 88). The model demonstrated robust predictive performance across all cohorts. Radiomic analysis identified peritumoral features as the most significant contributors. Model interpretation using paired transcriptome data linked therapeutic resistance to inflammatory cancer-associated fibroblasts. Both deep learning-based histopathological analysis and single-cell RNA sequencing further revealed their pro-angiogenic activity. By integrating features derived from radiographic and histopathological data, we developed a dual-omics model to improve predictive performance and enable efficacy-based stratification. Collectively, our study established an integrative predictive model that combines imaging radiomics and histopathological profiling, thereby elucidating tumor microenvironment heterogeneity associated with therapeutic outcomes. This finding provides practical insights into developing non-invasive biomarkers to characterize the tumor’s drug-resistant microenvironment.

[This corrects the article DOI: 10.1016/j.isci.2026.114872.][This corrects the article DOI: 10.1016/j.isci.2024.111340.].

BACKGROUND:Pathological cardiac hypertrophy remains a major contributor to heart failure, with impaired glucose metabolism playing a central role. Although exercise is known to enhance myocardial glucose utilization, the long-term metabolic reprogramming effects of exercise and their role in preventing pathological hypertrophy are poorly understood. This study elucidates the mechanisms underlying the sustained metabolic memory induced by exercise-induced hypertrophic preconditioning (EHP) and its cardioprotective effects, with a focus on RNA methylation and arachidonic acid metabolism. METHODS:We used positron emission tomography/computed tomography to assess cardiac glucose uptake and bulk RNA sequencing to profile myocardial gene expression in sedentary and EHP mice. Genetic manipulation of Pdk4 (pyruvate dehydrogenase kinase 4) was achieved via adeno-associated virus-mediated overexpression and tamoxifen-inducible, cardiac-specific Pdk4 knockout. Pressure overload was induced by transverse aortic constriction in cardiac-specific Pdk4 knockout and control (MCM [Myh6-MerCreMer]) mice. Epigenetic regulation of Pdk4 by EHP was investigated using pyrosequencing, single-base elongation- and ligation-based quantitative polymerase chain reaction and dual-luciferase assays. Untargeted metabolomics and molecular docking, molecular dynamics simulation, and cellular thermal shift assay were performed on heart tissues and neonatal rat cardiomyocytes/fibroblasts to identify key metabolites and their mechanisms of action. RESULTS:EHP conferred sustained myocardial glucose preference even after regression of physiological hypertrophy, mediated through METTL3 (methyltransferase-like 3)-dependent m6A RNA methylation that suppressed Pdk4 expression. Pdk4 overexpression abolished EHP-mediated cardioprotection, whereas Pdk4 deletion enhanced cardiac function and attenuated fibrosis under pressure overload. Metabolomic profiling identified arachidonic acid-derived metabolites 5-KETE (5-oxo-6E,8Z,11Z,14Z-eicosatetraenoic acid), 12-keto-leukotriene B4, and 20-hydroxy-leukotriene B4 as novel inhibitors of hypertrophy and fibrosis. These metabolites attenuated cardiomyocyte hypertrophy and fibroblast transdifferentiation through inhibition of the ERK2 (extracellular signal-regulated kinase 2)/MAPK1 (mitogen-activated protein kinase 1) pathway. CONCLUSIONS:This study establishes a unified mechanism by which EHP induces metabolic memory through RNA methylation-dependent suppression of Pdk4, leading to altered arachidonic acid metabolism and the accumulation of protective lipid mediators. These findings highlight the therapeutic potential of targeting the PDK4-arachidonic acid metabolites axis to mitigate pathological cardiac remodeling.

Immunotherapy has significantly altered the treatment paradigm of non-small cell lung cancer (NSCLC), but not all patients experience durable benefits. Predictive biomarkers are needed to identify patients who may benefit from immunotherapy. We retrospectively collected tumor tissues from 65 patients with advanced NSCLC before treatment, and performed transcriptomic and genomic analysis. By performing single-sample gene set enrichment analysis, we constructed a predictor named IKCscore based on the tumor microenvironment characteristics. IKCscore is a robust biomarker predicting response to immunotherapy, and its predictive capacity was confirmed from public datasets across different cancer types (N = 892), including OAK, POPLAR, IMvigor210, GSE135222, GSE126044, and Kim cohorts. High IKCscore was characterized by inflammatory tumor microenvironment phenotype and higher T cell receptor diversity. The IKCscore exhibits promise as a bioindicator that can predict the efficacy of both immunotherapy and immunotherapy-based combination therapies, while providing guidance for personalized therapeutic strategies for advanced NSCLC patients.

Background: The switch to endothelial-to-mesenchymal transition (EndMT) in endothelial cells (ECs) induced by disturbed flow (d-flow) has been identified as the critical driver of the pathogenesis of inflammatory vascular disorders. We aimed to investigate the role of EndMT in abdominal aortic aneurysms (AAA) and the underlying mechanism. Methods: Immunoblotting, immunofluorescence and transmission electron microscope were used to assess d-flow-induced EndMT in human and mouse AAA models (Ang II/PPE). An Ibidi pump system was used to produce d-flow on human aortic endothelial cells (HAECs), and the expression of galectin-7 was enhanced and weakened using an adeno-associated virus. Furthermore, single-cell RNA sequencing was performed to explore the underlying mechanism of galectin-7-mediated EndMT. Results: EndMT induced by d-flow, which suppressed galectin-7 expression, was positively correlated with AAA. Enhanced galectin-7 expression inhibited d-flow-induced EndMT and AAA progression, whereas reduced galectin-7 expression resulted in the opposite effect. Mechanistically, we found a EndMT-related cluster in HAECs by single-cell RNA sequencing, and the SRGN gene in this cluster was considered the core gene. Galectin-7 bound competitively to the transcription factor CREB, resulting in the inhibition of SRGN transcription, which in turn prevented TGFβ/smad pathway activation, thereby restoring EndMT progression. Conclusions: EndMT transformation in ECs exposed to d-flow was the critical driver of AAA development. Furthermore, endothelium-enriched galectin-7 suppressed the EndMT process induced by d-flow and prevent AAA progression by transcriptionally inhibiting SRGN via competitive binding with CREB to restrict TGFβ/smad pathway.

166 Background: Maintenance therapy with bevacizumab (Bev) plus capecitabine (Cap) is widely recommended to unresectable mCRC patients (pts). Fruquintinib (Fru) is a highly selective TKI that inhibits vascular endothelial growth factor receptor (VEGFR)-1,2,3. This study is to compare the therapeutic potential of alternating treatment with fruquintinib and bevacizumab plus capecitabine as maintenance therapy for mCRC (NCT05659290). Methods: Eligible mCRC pts aged 18-75 years with stable disease or better after induction treatment with chemotherapy in combination with Bev, ECOG PS 0-2, adequate bone marrow, liver, and renal function were enrolled. Forty patients were included (20 in phase IIa, 40 in phase IIb). In phase IIa, pts were orally administered with Fru (5 mg, qd, d1-14, q3w) alternating with Bev (7.5 mg/kg, iv.gtt, d1, q3w) plus Cap (850 mg/m 2 , orally, twice daily, d1-14, q3w). In phase IIb, pts were randomly assigned (1:1) to either maintenance treatment with Fru alternating with Bev plus Cap or Bev plus Cap. The primary endpoint was progression-free survival (PFS). Secondary endpoints included objective response rate (ORR), disease control rate (DCR) and safety. Results: At cutoff date of Sep 5, 2024, In phase IIa, 20 pts (14 males and 6 females) were enrolled with the median age was 59.0 years (range 27-75), ECOG PS 1 (95.0%), liver metastasis (55.0%), left-sided colon and rectal primary (70.0%). 11 pts had received at least one tumor assessment. the DCR was 100.0% (11/11) and the mPFS was immature, but 4 pts showed the PFS of ≥ 8 months (8.3, 8.6, 9.2, 13.4 m, respectively). The most common treatment-emergent adverse events (TEAEs) were proteinuria (60.0%), hypoalbuminemia (40.0%), hypertension (35.0%); The most common grade ≥ 3 adverse events were hypertension (10.0%), proteinuria (5.0%), and platelet count decreased (5.0%). After fully considering about patient′s tolerance and safety, the phase IIb of Fru was adjusted from 5mg to 3mg, these results provided further evidence that 3mg can ensure the safety and tolerance. Conclusions: Fruquintinib alternating with bevacizumab plus capecitabine as maintenance therapy after first-line treatment in mCRC showed preliminary anti-tumor activity and manageable toxicity. The phase IIb is ongoing and warrants further exploration in mCRC. Clinical trial information: NCT05659290 .

BACKGROUND:Cardiomyocytes exhibit marked susceptibility to ferroptosis after myocardial infarction (MI), rendering ferroptosis inhibition a promising therapeutic strategy to mitigate ischemic myocardial injury. Although mitochondrial dysfunction is recognized as a core driver of ferroptosis, the potential role of mitochondrial DNA transcription in regulating cardiomyocyte ferroptosis remains unexplored. METHODS:To clarify the temporal role of the various modes of cell death in MI progression, we performed time-course echocardiography in MI models treated with various cell death inhibitors. To characterize the crucial process and molecular regulator in cardiomyocyte ferroptosis, we integrated RNA sequencing and single-nucleus RNA sequencing data from murine post-MI hearts and performed functional rescue experiments using mitochondrial protective agents. To determine the role of ABHD11 (αβ-hydrolase domain-containing protein 11) in cardiomyocyte ferroptosis and cardiac repair after MI, we used loss- and gain-of-function approaches. To elucidate the underlying mechanisms, we conducted transcriptomics, nontargeted lipidomics, site-specific mutagenesis, molecular docking, coimmunoprecipitation, native gel electrophoresis, proximity ligation assay, methylation-specific polymerase chain reaction, and chromatin immunoprecipitation assay. RESULTS:We found that cardiac ferroptosis peaked at day 7 after MI and was enriched in peri-infarct cardiomyocytes. Mitochondrial dysfunction was a key driver of cardiomyocyte ferroptosis after MI, and the lipid enzyme ABHD11 was identified as a potential regulator of both processes. ABHD11 expression was consistently reduced in mouse and human MI hearts, and its transcription was repressed by DNMT1 (DNA methyltransferase 1)-mediated promoter hypermethylation. Functionally, cardiac-specific overexpression of ABHD11 markedly alleviated cardiomyocyte ferroptosis and improved cardiac function after MI. Conversely, loss of ABHD11 in adult mice exacerbated pathological cardiac remodeling and heart failure. Mechanistically, independent of its canonical enzymatic activities, ABHD11 acted as a mitochondrial DNA transcription coactivator by enhancing the TEFM (mitochondrial transcription elongation factor)-POLRMT (mitochondrial RNA polymerase) interaction. This promoted mitochondrial DNA transcription, restored mitochondrial function, and reduced reactive oxygen species/PUFA-PLs (polyunsaturated fatty acid-containing glycerophospholipids)-driven lipid peroxidation and 4-hydroxynonenal generation. The reduction in 4-hydroxynonenal stabilized YY1 (Yin Yang 1), which subsequently regulated key ferroptosis-driving genes governing iron deposition, reactive oxygen species production, and polyunsaturated fatty acid lipids accumulation, further inhibiting lipid peroxidation and ferroptosis, and ultimately promoting cardiac recovery after MI. CONCLUSIONS:This study revealed that ABHD11-mediated mitochondrial DNA transcription attenuated cardiomyocyte ferroptosis after MI by orchestrating a mitochondrial-nuclear crosstalk, offering a novel therapeutic strategy for ischemic myocardial injury.

ABSTRACT The tumor microenvironment (TME) is a critical factor in antitumor immunity and treatment outcome in cancer therapy. We have developed an analysis tool called the immuno‐oncology biological research (IOBR) to investigate the TME and its role in antitumor immunity. Leveraging multi‐omics data, IOBR facilitates comprehensive analysis of TME characteristics, immune interactions, and their impact on immunotherapy outcomes. IOBR features six modules for TME analysis, including transcriptomic data preprocessing, TME profiling, TME pattern identification, ligand–receptor interaction analysis, genome‐TME interaction assessment, and visualization, along with modeling. Since its release, this tool has been widely applied in many studies. In the future, IOBR will gradually integrate TCR/BCR repertoire analysis, enhance genomic functionalities, and develop spatial transcriptomics modules, which will further our understanding of TME dynamics and tumor immunity.

Background Physical exercise is known to induce trained immunity and improve function of macrophages, and we recently reported that exercise-induced hypertrophic preconditioning (EHP) could protect pathological hypertrophic heart, but it remains unclear whether EHP can mitigate myocardial ischemic injury. Here we hypothesized that EHP would alleviate acute myocardial ischemic injury through trained immunity of macrophages. Methods Swimming or treadmill running training was used to induce EHP in C57 mice. After 1-week of exercise termination, mice were subjected to myocardial ischemia/reperfusion (IR). IR injury and inflammatory response and cell energy metabolism reprogramming of bone marrow-derived macrophages (BMDMs) were analysed. Moreover, the effects of exercise on myocardial ischemic injury or long-term survival in patients with acute myocardial infarction (AMI) or history of MI were retrospectively analysed. Results EHP mice had a significantly smaller infarct size and less apoptotic cardiomyocytes than IR mice without EHP. At 4 weeks after IR, EHP mice had better cardiac function and less myocardial fibrosis. We noted less infiltration and polarization of myocardial M1 macrophages, higher expression of M2 signature anti-inflammatory factors in BMDMs of EHP mice. EHP promoted tricarboxylic acid cycle, mitochondrial oxidative phosphorylation (OXPHOS) and adenosine 5’-monophosphate activated protein kinase (AMPK) phosphorylation, decreased glycolysis and phosphorylation of AKT and mTOR in BMDMs. Importantly, AMPK inhibitor treatment abrogated the cardioprotective effects of EHP, abolishing the improvement in infarct size and apoptosis and reversing the metabolic reprogramming of macrophages. In mice received treadmill running training for 6 weeks, no physiological hypertrophy occurred but myocardial infarct size after IR was significantly reduced. Similarly, in response to AMI, patients with exercise habit had higher left ventricular ejection fraction and lower plasma levels of cardiac troponin T than those sedentary AMI patients. In 2457 patients with self-reported history of MI, patients with regular exercise habit had significantly lower post-MI mortality. Conclusions EHP protects heart against myocardial ischemic injury through trained immunity of macrophage mediated by energy metabolism reprogramming.

Background Patients with microsatellite stable (MSS) colorectal cancer (CRC) often display resistance to immunotherapy. Epidermal growth factor receptor (EGFR)-targeted therapies have shown potential in enhancing immunotherapy, yet clinical benefits remain unfulfilled, which may relate to inadequate patient stratification.Methods Circulating tumor cells and tumor tissues were collected from multicenter cohorts of patients with CRC receiving cetuximab to analyze EGFR variant type III (EGFRvIII) expression and immune infiltration. Syngeneic mouse models of EGFRvIII CRC were used to investigate the combined efficacy of adenosine inhibition and antiprogrammed cell death protein 1 (anti-PD-1).Results EGFRvIII mutations are found in about 10% of MSS CRC and are associated with poor response to cetuximab therapy. EGFRvIII-mutated patients with CRC exhibit an adenosine-mediated immunosuppressive tumor microenvironment (TME) subtype. Combination therapy with adenosine inhibitors remodels the TME, reversing cetuximab resistance and enhancing anti-PD-1 efficacy in EGFRvIII CRC.Conclusions Our findings identified EGFRvIII-positive CRC as a distinct subtype characterized by adenosine-mediated immunosuppressive TME. Targeting adenosine significantly improved the efficacy of anti-PD-1 in MSS CRC.

The tumor microenvironment (TME) significantly influences cancer prognosis and therapeutic outcomes, yet its composition remains highly heterogeneous, and currently, no highly accessible, high-throughput method exists to define it. To address this complexity, the TMEclassifier, a machine-learning tool that classifies cancers into three distinct subtypes: immune Exclusive (IE), immune Suppressive (IS), and immune Activated (IA), is developed. Bulk RNA sequencing categorizes patient samples by TME subtype, and in vivo mouse model validates TME subtype differences and differential responses to immunotherapy. The IE subtype is marked by high stromal cell abundance, associated with aggressive cancer phenotypes. The IS subtype features myeloid-derived suppressor cell infiltration, intensifying immunosuppression. In contrast, the IA subtype, often linked to EBV/MSI, exhibits robust T-cell presence and improved immunotherapy response. Single-cell RNA sequencing is applied to explore TME cellular heterogeneity, and in vivo experiments demonstrate that targeting IL-1 counteracts immunosuppression of IS subtype and markedly improves its responsiveness to immunotherapy. TMEclassifier predictions are validated in this prospective gastric cancer cohort (TIMES-001) and other diverse cohorts. This classifier could effectively stratify patients, guiding personalized immunotherapeutic strategies to enhance precision and overcome resistance.

BACKGROUND:Amivantamab stands as the pioneering bispecific antibody that targets both EGFR and MET, utilized in the treatment of locally advanced or metastatic non-small cell lung cancer (NSCLC) harboring EGFR ex20ins mutations. Nevertheless, a thorough assessment of its safety characteristics in the real-world remains unknown. RESEARCH DESIGN AND METHODS:The adverse event (AE) reports were collected through a search of the FDA Adverse Event Reporting System (FAERS) database spanning from 2019 Q1 to 2024 Q1, and then disproportionality analysis was utilized. RESULTS:Totally, 9,252,269 AE reports were obtained from the FAERS database, with 893 reports of amivantamab classified as primary suspect AEs. Amivantamab-related AEs were distributed in 23 organ systems, and 87 significant preferred terms (PTs) met the reporting odds ratio criteria. Novel significant AEs were detected, and the median time to onset of amivantamab-associated AEs was 43 days. In subgroup analysis, a higher proportion of patients who were male, over 65 years, and with pneumonitis or pneumonia were reported in the death cases. We also found that AEs may vary between intravenous and subcutaneous administration. CONCLUSIONS:This investigation offered novel prospects for monitoring and addressing undesirable medication effects associated with amivantamab, which might improve the clinical medication safety.

Figure. S6 Matrix stiffness induces HSC activation.

Abstract Background The mechanisms underlying cardiac remodeling in aortic valvular (AoV) disease remain poorly understood, partially due to the insufficiency of appropriate preclinical animal models. Here, we present a novel murine model of aortic regurgitation (AR) generated by trans‐apical wire destruction of the AoV. Methods Directed by echocardiography, apical puncture of the left ventricle (LV) was performed in adult male C57BL/6 mice, and a metal guidewire was used to induce AoV destruction. Echocardiography, invasive LV hemodynamic and histological examination were conducted to assess the degree of AR, LV function and remodeling. Results AR mice exhibited rapid aortic regurgitation velocity (424 ± 15.22 mm/s) immediately following successful surgery. Four weeks post‐surgery, echocardiography revealed a 54.6% increase in LV diastolic diameter and a 55.1% decrease in LV ejection fraction in AR mice compared to sham mice. Pressure‐volume catheterization indicated that AR mice had significantly larger LV end‐diastolic volumes (66.2 ± 1.5 μL vs. 41.8 ± 3.4 μL), reduced LV contractility (lower dP/dtmax and Ees), and diminished LV compliance (smaller dP/dtmin and longer Tau) compared to sham mice. Histological examination demonstrated that AR mice had significantly larger cardiomyocyte area and more myocardial fibrosis in LV tissue, as well as a 107% and a 122% increase of heart weight/tibial length and lung weight/tibial length, respectively, relative to sham mice. Conclusions The trans‐apex wire‐induced destruction of the AoV establishes a novel and efficient murine model to develop AR, characterized by significant eccentric LV hypertrophy, heart failure, and pulmonary congestion.

Figure. S8 Matrix stiffness induced lipid crosstalk between HSCs and colon cancer cells promote tumor proliferation and angiogenesis.

Immune checkpoint inhibitors (ICIs) are widely used to treat advanced non-small cell lung cancer (NSCLC). However, it remains crucial to identify patients who are unlikely to benefit from immunotherapy and to explore potential combination treatment strategies. In this study, 1127 advanced NSCLC patients from multicenter randomized clinical trials (OAK, POPLAR, ORIENT-11) and an in-house cohort who received ICIs, ICIs combined with chemotherapy, or chemotherapy alone are analyzed. Using bulk RNA-seq transcriptomic data, an RNA-based model, named the Lung Cancer Immunotherapy Response Assessment (LIRA), is developed, utilizing interaction analysis and a random forest algorithm to predict immunotherapy outcomes. LIRA outperforms PD-L1 expression and tumor mutation burden in predicting responses, particularly in identifying early progression risk during ICI monotherapy (HR: 0.15, 95% CI: 0.11-0.20). Tumor profile analysis reveals that LRP8 and HDAC4 are associated with immunotherapy outcomes. Additionally, scRNA-seq analysis of NSCLC tumors indicates a higher prevalence of T cells and a reduced proportion of epithelial cells in samples with a high LIRA-score. The deep learning model pinpointed critical high-attention regions within whole-slide images that contributed decisively to the LIRA predictions. In summary, these results demonstrate that LIRA enables independent risk stratification of NSCLC patients and provides insights into potential resistance mechanisms.