Shamai and colleagues developed a multimodal deep-learning model that predicts Oncotype DX recurrence scores from routine H&E slides and clinicopathological variables in hormone receptor‑positive, HER2‑negative early breast cancer. Validated across the TAILORx trial and six external cohorts (over 5000 patients), the model achieved an AUC of 0.898 for identifying recurrence score ≥26 and recapitulated genomic assay patterns of chemotherapy benefit. Notably, 31% of clinically high-risk postmenopausal women were downgraded to low risk by AI, suggesting potential to reduce overtreatment. However, several limitations preclude immediate clinical substitution for genomic testing. First, intratumoural heterogeneity leads to discordant predictions with unclear management guidance. Second, the model's chemotherapy benefit estimates rely on TAILORx's age-based menopausal surrogates, which may not reflect real-world hormonal status or LHRH agonist use. Third, predictive value in node-positive disease remains untested in randomised datasets such as RxPONDER. Additionally, calibration uncertainty near risk thresholds and global scalability issues (including IHC requirements and digital pathology infrastructure) persist. While this represents a landmark step toward democratising precision oncology, the AI tool should currently serve as a complementary decision aid, with genomic testing remaining the gold standard for intermediate, borderline, or discordant cases.
BACKGROUND:Pancreatic ductal adenocarcinoma (PDAC) lacks biomarkers for accurate diagnosis and prognostic stratification. The standard, CA19-9, has suboptimal specificity in differentiating PDAC from mimics like chronic pancreatitis (CP). We investigated serum metabolomic signatures to address these challenges. METHODS:We conducted a prospective, multi-cohort (n = 221) untargeted metabolomics study, analyzing PDAC (n = 66), healthy control (n = 92), other cancer (n = 40), and CP (n = 23) groups. Machine learning and survival analysis were employed to build diagnostic and prognostic models from serum samples collected at diagnosis. RESULTS:A two-metabolite panel distinguished PDAC from other cancers (AUC=0.894), and a five-metabolite signature showed high discrimination between PDAC and CP (AUC=0.998; 95% CI, 0.993-1.000). A leakage-resistant nested cross-validation sensitivity analysis yielded a mean AUC of 0.955 (SD, 0.020). The exploratory 18-metabolite risk score separated high- and low-risk groups and remained associated with overall survival after adjustment for stage, age, sex, and CA19-9 (adjusted HR, 3.94; 95% CI, 2.05-7.58; P < 0.001), with a C-index of 0.601. CONCLUSIONS:Serum metabolomic profiles identified compact candidate panels that provided information complementary to CA19-9 for PDAC differential diagnosis, while the 18-metabolite risk score was associated with overall survival. These findings support targeted assay development and prospective multicenter evaluation of serum metabolomics for the clinical characterization of PDAC.
Patient-derived organoids (PDOs) have emerged as promising preclinical models for functional drug testing in ovarian cancer, with potential to support personalized treatment selection. However, their predictive value for clinical treatment response remains unclear. This systematic review evaluated the predictive accuracy of ovarian cancer PDOs for treatment efficacy.A systematic search of was conducted from inception to January 29, 2026. Eligible studies included patients with high-grade epithelial ovarian, fallopian tube, or primary peritoneal cancer from whom PDOs were generated for in vitro drug testing and directly correlated with clinical outcomes. The primary outcome was predictive accuracy, defined as concordance/ correlation between in vitro PDO drug response and in vivo patient outcomes. Because of substantial heterogeneity, a narrative synthesis was performed.Twelve studies published between 2019 and 2026 were included. Cohort sizes ranged from 6 to 61 patients and comprised two prospective validation studies and ten retrospective translational or feasibility studies. Clinical endpoints were heterogeneous and included radiologic, biochemical and histopathologic response, progression-free survival, and descriptive clinical course. Four studies reported statistically significant associations between PDO drug response and clinical outcome, six reported concordant findings without formal statistical testing, two showed limited or mixed concordance, and one relied on descriptive comparison only. The strongest evidence came from two prospective studies, which reported accuracies of 89% and 91.7%.Ovarian cancer PDOs show promise as functional predictors of treatment response, but current evidence is limited by small cohorts, methodological heterogeneity, and scarce prospective validation. Standardized prospective studies are needed to define their clinical applicability.
Accumulating evidence has established the sympathetic nervous system (SNS) as a critical regulator of tumor progression, making adrenergic signaling an attractive therapeutic target in oncology. Consequently, β-adrenergic blockade has been widely investigated as a strategy to inhibit cancer progression. However, recent advances in cancer neuroscience indicate that sympathetic regulation is considerably more complex than previously appreciated. Distinct adrenergic receptor subtypes, direct target cells, and tissue niches generate diverse biological outcomes, suggesting that indiscriminate inhibition of sympathetic signaling may not represent the optimal therapeutic approach. Here, we propose that the next challenge in cancer neuroscience is not whether sympathetic signaling should be targeted, but how it should be targeted. We introduce a context-dependent framework in which therapeutic strategies are guided by receptor subtype, direct target cell, and tissue-specific neural circuits rather than global sympathetic suppression. Finally, we discuss a future roadmap toward precision targeting and speculate that programmable neuromodulation may ultimately complement pharmacological intervention by enabling spatially and temporally selective control of tumor-associated neural circuits without influencing physiological functions.
INTRODUCTION:Prognosis of Relapsed/Refractory (R/R) neuroblastoma is still dismal. Chemoimmunotherapy has been reported to improve the response rate. PATIENTS AND METHODS:Dinutuximab beta added to chemotherapy was used in an off-label setting. Treatment was administered on a 21-day schedule: on day 1 Irinotecan was administered at the dose of 50 mg/m2/day for 5 days, together with Temozolomide, at the dose of 100 mg/m2/day for 5 days. Dinutuximab beta was administered from day 2 at the dose of 17,5 mg/m2/day for 4 consecutive days. RESULTS:Twenty-five children received chemo-immunotherapy. Hematological toxicity was the most common adverse event. The best overall response rate (ORR) was 52% (95% CI:31-72), with an ORR of 62% (95% CI: 38-82) at metastatic sites. Responses were higher at bone and bone marrow sites (ORR 67% and 92%, respectively). SIOPEN skeletal score declined during treatment, the best response being observed after the first 2/3 courses. Primary tumors and soft-tissue lesions were less responsive, though metabolic activity often decreased. Negative or low GD2 expression was observed in 10% and in 35% of tumor tissue, respectively at baseline and after treatment, with mean GD2-positive tumor cells decreasing from 88% to 62% in tumor tissue, mainly associated with differentiating histology. DISCUSSION:This real-world experience confirms that a short schedule is a feasible and effective option in R/R neuroblastoma. The dynamic and heterogeneous expression of GD2 deserves further evaluation, particularly in relation to selection of patients and subsequent GD2-directed treatment strategies.
Objectives: To develop and validate a multi-temporal magnetic resonance imaging (MRI)-based radiomics model for predicting one-year recurrence in hepatocellular carcinoma (HCC) after ablation. Methods: This retrospective study included 204 patients with HCC who underwent curative ablation between 2016 and 2021. Multi-sequence MRI was performed within one month before and three months after ablation. Radiomics features were extracted from arterial phase (T1A), portal venous phase (T1V), and T2-weighted images using pre-ablation, post-ablation, and delta features. Patients were randomly divided into training (n = 142) and validation (n = 62) cohorts. Models were evaluated using receiver operating characteristic analysis, and differences in AUC between models were assessed. Results: During a median follow-up of 750 days, 50 patients (24.5%) developed recurrence within one year. The integrated radiomics model (IRM) achieved the best discriminatory performance, with area under the curve values of 0.888 (95% CI: 0.801–0.976) in the training cohort and 0.858 (95% CI: 0.718–0.998) in the validation cohort. Kaplan-Meier analysis showed significant differences in recurrence-free survival between IRM-defined low- and high-risk groups in both cohorts (p < 0.001). Conclusion: The multi-temporal MRI radiomics model showed promising performance for early recurrence risk stratification after HCC ablation in the internal validation cohort, suggesting potential translational value for non-invasive recurrence-risk assessment. Further prospective external validation is warranted.
OBJECTIVE:Glioblastoma (GBM) is a highly aggressive brain tumor with a median survival of about 15 months and frequent recurrence. Liquid biopsy of cerebrospinal fluid (CSF) and plasma enables minimally invasive detection of tumor-associated molecular signals and may improve early diagnosis of recurrence. This study aims to elucidate the distinct molecular landscapes of cerebrospinal fluid and plasma in glioblastoma, and to identify recurrence-associated biomarkers through an integrated multi-omics framework. METHODS:Using integrated bioinformatic frameworks, gene set enrichment analysis, and receiver operating characteristic curves, we systematically mapped differential gene expression patterns and their associated regulatory mechanisms. RESULTS:Cerebrospinal fluid yielded substantially more differentially expressed genes than plasma (3,362 vs. 1,329), with upregulated genes predominating in both compartments. We identified four genes (THNSL2, SYT5, RUVBL1, and GSKIP) that showed significant expression differences between primary and recurrent tumors, with RUVBL1 and GSKIP demonstrating the strongest functional enrichment signals. These genes were preferentially linked to glioblastoma cell subpopulations in both fluid types. The plasma-based multigene prognostic model achieved area under the curve values of 1.0 for 1-, 2-, and 3-year survival prediction in the training cohort, which requires further validation in independent cohorts. CONCLUSION:Collectively, by using patients' CSF and plasma samples, this study identifies candidate molecular markers and regulatory networks associated with primary and recurrent GBM through integrated analysis of CSF and plasma samples, and highlights complementary molecular signatures between the two fluid compartments. Our findings provide a preliminary molecular basis and candidate targets for future research on precision diagnosis, prognostic stratification, and targeted therapy of GBM.
Seminoma, which is the most frequent testicular germ cell tumor, has stem cell-like features related to treatment resistance and recurrence, yet the molecular mechanisms that preserve its stemness are not well understood. This study aimed to discover the tumorigenesis and stemness regulation of seminoma. In vivo CRISPR/Cas9 knockout library screening was performed using Tcam-2 seminoma cells implanted into NSG mice to identify tumorigenesis drivers. Parallel FACS-based screening enriched CD117/CD133 double-negative cells to assess the regulators of seminoma stemness. Functional validation included clonogenicity assays, sphere formation tests, and tumor initiation experiments in vivo. RNA sequencing analyzed downstream pathways. In vivo and FACS-based CRISPR/Cas9 screening uncovers VENTX as a crucial element in maintaining seminoma stemness. Functional validation showed that VENTX knockout decreased clonogenicity, sphere formation, and CD117+/CD133+ populations in Tcam-2 cells, and also hindered tumor initiation in vivo. RNA sequencing identified a connection between VENTX and pluripotency pathways, demonstrating a downregulation of ID1/2/3 (essential differentiation inhibitors) and an imbalance in Wnt/β-catenin and TGF-β signaling. VENTX expression showed a significant association with DNA methylation-based stemness scores in testicular germ cell tumors. In conclusion, VENTX is a key regulator of seminoma stemness, presenting a promising target for therapy to address recurrence and resistance in seminoma.
Background High-grade serous ovarian carcinoma (HGSOC) features extensive intratumoral heterogeneity, frequent chemoresistance and poor prognosis. Tumor proliferation kinetics reflected by tumor doubling time (TDT) are linked to therapeutic response, yet molecular drivers of chemoresistance-associated proliferation remain incompletely defined. Methods We integrated scRNA-seq (GSE154600) with bulk data (TCGA-OV, GTEx). Differential analyses of refractory/resistant versus sensitive tumors and tumor versus normal tissues, intersected with TDT-associated genes, identified 105 chemoresistance-associated proliferation genes. Consensus clustering and immune analyses were performed. A 117-algorithm machine-learning framework constructed a prognostic signature trained on TCGA-OV and validated in GSE26193. Survival was assessed by Kaplan–Meier Plotter. Hub genes were validated by qPCR and Western blot in paired sensitive (A2780, TYKnu) and cisplatin-resistant (A2780-DDP, TYKnu-DDP) cell lines. Results The genes enriched in cell-cycle and p53 signaling. Consensus clustering defined C1 and C2 subtypes with distinct immune microenvironments; C2 showed upregulation of DNA replication and cell-cycle programs. A nine-gene signature (BIRC5, CENPH, CKAP2, PAK1IP1, PBK, SDF2L1, TEAD4, TPM3, UBE2T) was established. High CENPH, CKAP2, PBK, TEAD4, TPM3 and UBE2T associated with inferior overall survival, while SDF2L1 was protective. These genes were predominantly expressed in malignant epithelial cells, SPP1+/TREM2+ macrophages and CAFs. qPCR confirmed upregulation of six genes in TYKnu-DDP cells; Western blot validated elevated SURVIVIN (BIRC5), CENPH, CKAP2, PAK1IP1 and PBK in resistant lines. Conclusions This single-cell landscape of chemoresistance-associated proliferation genes delineates a nine-gene prognostic signature for HGSOC and confirms hub-gene overexpression in cisplatin-resistant cells, nominating therapeutic targets.
Background Cancer-associated fibroblasts (CAFs) are key components of the bladder cancer microenvironment, but how specific CAF subpopulations drive malignant progression remains unclear. This study aimed to identify CAF driver genes and elucidate the role of caveolin-1 (CAV1) in CAF-mediated bladder cancer progression. Methods Single-cell RNA sequencing and public bladder cancer datasets were integrated to identify fibroblast-associated candidate genes and evaluate the clinical significance of CAV1. Clinical samples were analyzed by immunofluorescence. The biological functions of CAV1 were further evaluated through CAF–bladder cancer cell co-culture systems, gain- and loss-of-function experiments, and comprehensive molecular analyses. The downstream mechanism was explored through LIF stimulation, STAT3 silencing, and validated in a subcutaneous xenograft model. Results Single-cell analysis identified CAV1 as a fibroblast-associated gene linked to stemness- and ferroptosis-related signatures. High CAV1 expression predicted poor prognosis and was independently associated with unfavorable survival. CAV1 expression positively correlated with fibroblast infiltration and was enriched in a subset of ACTA2-positive CAFs. CAFs markedly promoted multiple malignant behaviors of bladder cancer cells, whereas silencing CAV1 effectively counteracted these pro-tumorigenic effects. Mechanistically, CAV1 enhanced LIF expression in CAFs, thereby activating JAK2/STAT3 signaling in tumor cells. Pharmacological inhibition of STAT3 abolished the tumor-promoting effects of CAFs both in vitro and in vivo. Conclusions CAF-derived CAV1 promotes bladder cancer progression by activating the LIF-STAT3 signaling axis. Targeting stromal CAV1-mediated STAT3 signaling may represent a promising therapeutic strategy for bladder cancer.
Primary effusion lymphoma (PEL) and fluid overload–associated large B-cell lymphoma (FO-LBCL) are rare B-cell neoplasms presenting as serous effusions without solid masses, distinguished primarily by human herpesvirus-8 status. We analyzed three PEL and seven FO-LBCL cases using immunohistochemistry, Epstein–Barr virus–encoded RNA in situ hybridization, immunoglobulin heavy chain (IGH), T-cell receptor beta (TRB) and T-cell receptor gamma (TRG) gene rearrangement analyses, and targeted next-generation sequencing. The median age was 72 years for PEL and 83 years for FO-LBCL. All PEL cases lacked CD20, whereas FO-LBCL tumors retained pan–B-cell markers. Aberrant T-cell antigen expression was observed in two PEL cases, both showing clonal T-cell receptor rearrangements in addition to IGH clonality. FO-LBCL harbored recurrent alterations involving CD79B, PIM1, and MYD88, suggesting partial molecular overlap with the molecular features characteristic of the MCD subtype of diffuse large B-cell lymphoma (DLBCL). Clinically, all patients with PEL died of disease, whereas FO-LBCL appeared to have a more favorable clinical course, with four patients remaining alive at the last follow-up. Despite morphologic overlap, PEL and FO-LBCL showed distinct immunophenotypic, genetic, and clinical features. These preliminary findings support further investigation of B-cell receptor/NF-κB–related signaling pathways in FO-LBCL and may contribute to improved molecular classification and future therapeutic stratification.
Background Lung adenocarcinoma (LUAD) represents a significant global health challenge due to its elevated morbidity and mortality rates. Despite advancements in LUAD therapies, there remains an imperative need for novel therapeutic targets for patients who exhibit unresponsiveness or resistance to existing treatment modalities. Methods To investigate the potential role of SENP7 in LUAD progression, we analyzed SENP7 expression levels in LUAD and normal lung tissues using data from the TCGA, GTEx, and CPTAC databases, supplemented by immunohistochemistry (IHC) staining and western blotting assays. Additionally, the effects of SENP7 and SP1 on LUAD progression were assessed through CCK-8 assays, colony formation assays, EDU incorporation assays, and tumor xenograft experiments. To elucidate the role of the SENP7-SP1 pathway in LUAD progression, we conducted ubiquitination assays, co-immunoprecipitation (Co-IP), SUMOylation assays, and western blotting analyses. Results In this study, we demonstrated that the SENP7 protein is markedly expressed in LUAD, with a strong correlation to disease progression and adverse patient prognoses. Our findings reveal that the depletion of SENP7 substantially impedes LUAD progression in both in vitro and in vivo models. Importantly, we identified SP1 as a bona fide target of SENP7, with SENP7-mediated deSUMOylation, deubiquitination, and stabilization of SP1 being essential for LUAD development and chemoresistance. Conclusions These results underscore the pivotal role of the SENP7-SP1 axis in promoting chemoresistance and disease progression in LUAD, thereby providing a foundation for potential therapeutic strategies in the management of advanced cancer.
Pancreatic adenocarcinoma (PAAD) remains one of the most lethal cancers, largely due to the lack of reliable diagnostic biomarkers. Through integrative transcriptomic and proteomic analyses, we evaluated a four-gene signature comprising CLDN18, CEACAM5, FUT3, and FUT6, capable of distinguishing tumor from normal pancreatic tissue with high accuracy (AUC > 0.80). Among these genes, CLDN18 emerged as a structurally distinct and underappreciated biomarker, showing strong transcriptional association with CEACAM5 and the glycosyltransferases FUT3/FUT6, which are involved in the biosynthesis of CA19.9, the current clinical standard for PAAD diagnosis. Functional enrichment analyses indicated that CLDN18 is involved in pathways related to epithelial polarity, tight junction organization, and extracellular vesicle (EVs)-mediated transport. Machine learning models demonstrated robust classification performance, with CLDN18 consistently ranking among the top predictive features. In vitro validation confirmed the overexpression of CLDN18 in pancreatic cancer cells compared with non-tumoral pancreatic epithelial cells and its selective enrichment in tumor-derived EVs. Furthermore, by immunohistochemical evaluation of primary pancreatic adenocarcinomas, we demonstrated that CLDN18 was indeed overexpressed in tumor samples compared to corresponding normal tissues. Overall, these findings underscore the value of integrating junctional, glycosylation-related, and antigenic markers into multi-marker classifiers to improve diagnostic accuracy. Given its consistent overexpression, prognostic relevance, and compatibility with emerging immunotherapeutic strategies, CLDN18 represents a promising addition to current biomarker panels and a potential target for molecular imaging and drug development in PAAD.
Background: Resistance to androgen receptor signaling inhibitors (ARSIs) remains a major barrier of advanced prostate cancer (PCa) treatment. While RNA epitranscriptomic modifications are increasingly recognized as key regulators of tumor biology, the role of pseudouridine (Ψ) in therapeutic resistance is largely unexplored.Methods: A darolutamide-resistant PCa cell model was established and subjected to integrated multi-omics profiling using bulk RNA sequencing and photo-crosslinking-assisted Ψ sequencing (PA-Ψ-seq). Differential expression and pseudouridylation analyses were combined to identify Ψ-associated genes. Public datasets validated expression and prognosis. Functional assays including RNA knockdown, cell proliferation, colony formation, and xenograft models were conducted. Single-cell RNA sequencing investigated tumor microenvironment (TME) interactions.Results: We identified extensive transcriptomic and pseudouridylation alterations associated with ARSI resistance, with a significant positive correlation between Ψ modification and mRNA expression. Integrated analysis highlighted a subset of “hyper-up” genes enriched in resistance-related pathways. Thus, TIMM17A was identified as a novel candidate. TIMM17A expression was significantly elevated in PCa and correlated with disease progression and poor prognosis. Experimental validations demonstrated that TIMM17A promoted tumor growth and resistance, while its knockdown restored sensitivity to darolutamide both in vitro and in vivo. Mechanistically, TIMM17A expression may be regulated by PUS1‑mediated pseudouridylation. Single-cell analysis further revealed that TIMM17A is enriched in malignant epithelial cells and associated with enhanced cell–cell communication within the TME.Conclusions: This study delineates the pseudouridine epitranscriptomic landscape in advanced PCa and identifies TIMM17A as a key mediator of therapeutic resistance. Targeting the Ψ–TIMM17A axis may offer a novel strategy to overcome ARSI resistance.
Cancer remains one of the leading global health burdens, with increasing complexity in genomic, imaging, and clinical datasets presenting significant challenges for effective management. Artificial intelligence (AI) has emerged as a powerful tool to address these challenges by enabling pattern recognition, knowledge integration, and data-driven decision-making. This review highlights recent advances in the application of AI across cancer research, diagnosis, and therapy. In research, AI accelerates drug discovery and repurposing, enhances genomic data interpretation, and facilitates biomarker identification through multi-omics integration. In diagnosis, AI has demonstrated high technical performance in radiology for lesion detection and image segmentation, in pathology for tumour grading and molecular prediction, and in liquid biopsy for non-invasive biomarker analysis. In therapy, AI supports precision medicine by predicting treatment responses, monitoring disease progression, and optimizing clinical trial design. Despite these advances, barriers such as data heterogeneity, algorithmic bias, interpretability, and regulatory challenges remain. Future directions, including explainable AI, federated learning, multimodal modelling, and digital twins, hold promise for translating AI-driven innovations into routine oncology practice.Significance StatementThis review provides a timely synthesis of recent (2020–2025) advances in artificial intelligence across cancer research, diagnosis, and therapy, highlighting applications in drug discovery, genomics, multi-omics biomarker identification, and clinical decision-making. By integrating technological progress with translational and clinical relevance, this work serves as a valuable resource for bridging AI innovation with precision oncology practice. As a narrative review, the literature was identified through targeted PubMed, Scopus, and Google Scholar searches, combining terms for artificial intelligence, machine learning, and deep learning with cancer-related keywords, with priority given to peer-reviewed studies published between 2020 and 2025, seminal earlier works, and official regulatory or guideline documents. Within each domain, representative studies were selected to illustrate methodological diversity, clinical context, and current translational readiness rather than to provide exhaustive coverage of an extremely rapidly evolving field.
Introduction Immunotherapy recently approved with first-line (1L) chemotherapy for advanced biliary tract cancer (aBTC). Translational research is warranted to identify biomarkers for efficacy. Methods This prospective multicenter non-randomised phase-II study recruited 50 treatment-naïve proven aBTC patients for cisplatin and gemcitabine with pembrolizumab until disease progression or unacceptable toxicity. Primary endpoint was progression-free survival at 6 months (6mPFS, RECIST v.1.1). Secondary endpoints were overall survival (OS), radiological responses and safety. Exploratory translational research was to identify predictive biomarkers. Tumour tissue and blood were collected for MMR, PD-1/PD-L1 in T-cells, monocytes and myeloid derived suppressor cells (MDSCs). Results Fifty patients were 38% male, aged 63 (35–84) years. 6mPFS was 61.2% (80% CI 51.7–69.5) with mPFS 8.3 months (95% CI 5.6–10.6); ORR 40.8% (95% CI 27.0–55.8). Med. OS was 13.4 months (95% CI 8.3–20.5) with 24-months-OS 28.4% (95% CI 16.6–41.4). Higher PD-L1 expression correlated with improved PFS (HR=0.59; 95% CI: 0.26–1.37) and OS (HR=0.68; 95% CI: 0.28–1.64). Responders showed higher CD8/Treg ratio than non-responders (26.2% vs 21.2%) and lower CD4/PD1+ and CD8/PD1+ T-cells (33.5% vs 41.5% and 26.4% vs 44.7%) respectively. During treatment, profiling decreased in CD4/PD1+ and CD8/PD1+ Tcells (36.0% vs 22.2% and 41.6% vs 21.0%) respectively. MDSCs increased in non-responders. Responders, CPS-positive and low-stage patients showed normal Neutrophil-Lymphocyte-Ratios. Conclusion Pembrolizumab plus CisGem showed consistent efficacy to prior phase-III trials. Biomarkers suggest baseline immune landscape influences responses, T-cell exhaustion and immunosuppressive myeloid expansion. Immune phenotyping supports patient selection and further biomarker-driven trials in BTC.
Prostate cancer exhibits substantial microenvironmental heterogeneity, but endothelial plasticity and its clinical relevance remain insufficiently characterized. In this study, we integrated single-cell RNA sequencing, spatial transcriptomic data, and bulk transcriptomic cohorts to investigate endothelial heterogeneity in prostate cancer. Single-cell analysis identified a fibroblast-like endothelial (FLE) population characterized by the coexistence of endothelial and stromal transcriptional features. Endothelial subclustering further revealed a mesenchymal-like subset enriched in extracellular matrix remodeling, developmental, and lipid metabolic programs, supporting the presence of an EndMT-like endothelial state. Cell-cell communication analysis showed that FLE cells displayed strong incoming and outgoing signaling activity, with the MDK-NCL interaction emerging as a candidate ligand-receptor axis within the prostate cancer microenvironment. Based on an FLE-enriched co-expression module identified by hdWGCNA, we constructed a FLE-related prognostic model that consistently stratified patient outcomes across the TCGA-PRAD, GSE116918, and GSE46602 cohorts. High-risk patients exhibited poorer survival outcomes, while SHAP analysis showed cohort-dependent contributions of SPARC, IGFBP7, ESAM, CRIP2, and COL4A1 to risk prediction. Experimental validation demonstrated elevated NCL expression in prostate cancer tissues and cell lines. NCL knockdown suppressed cell viability, colony formation, and DNA synthesis in PC3 and DU145 cells. Collectively, these findings identify an FLE state associated with active intercellular communication and adverse clinical outcomes, highlighting endothelial plasticity and the MDK-NCL axis as potential components of prostate cancer microenvironment remodeling.
Lung adenocarcinoma (LUAD) is a highly heterogeneous malignancy with substantial differences in clinical outcomes and therapeutic responses. However, robust molecular signatures integrating prognostic stratification with tumor biological and immune characteristics remain limited. In this study, we established and validated a three-gene prognostic signature consisting of KRAS, KIAA1109, and BRCA1 through stepwise multivariate Cox regression analysis. The signature demonstrated robust prognostic performance across multiple independent cohorts. High-risk patients exhibited enrichment of cell cycle- and DNA repair-associated pathways, accompanied by distinct tumor microenvironment landscapes and altered immune infiltration patterns. Further analyses revealed that the high-risk group displayed lower TIDE scores, higher tumor mutation burden, and distinct immunotherapy-associated characteristics in the IMvigor210 cohort, suggesting potential differences in responsiveness to immune checkpoint blockade. Among the three signature genes, BRCA1 was selected for functional validation. Consistent with the bioinformatic analyses, BRCA1 was markedly upregulated in LUAD tissues and cultured LUAD cells, and its depletion suppressed malignant phenotypes in vitro. Collectively, our findings establish a clinically relevant prognostic signature integrating survival prediction, tumor microenvironment characteristics, and immune-related biological features in LUAD. This study provides a biologically interpretable framework for prognostic stratification and characterization of tumor heterogeneity in LUAD.
