Abstract Objective This review delineates the evolution of attention architectures in automated tumor segmentation across three pivotal paradigms: classic Pre‐Transformer attention, dominant Transformer‐based self‐attention, and emerging Mamba‐based state space models. Methods We synthesize their functional roles—from channel enhancement to long‐range dependency modelingand critically assess hybrid structures designed for multimodal clinical oncology scenarios. Findings/Results The review establishes attention mechanisms as a foundational pillar for the next generation of intelligent segmentation tools, highlighting their potential in feature enhancement and localization. Conclusion Interrogating current challenges and charting future trajectories, these mechanisms are poised to profoundly impact the landscape of precision medicine.

Abstract Background Hepcidin, a key regulator of iron homeostasis, is often overexpressed in non‐small cell lung cancer (NSCLC). While hepcidin renders cells susceptible to ferroptosis, which is linked to the activity of programmed cell death 1 (PD‐1)/programmed death ligand 1 (PD‐L1) inhibitors, its clinical impact on immunotherapy efficacy remains unclear. Methods We retrospectively analyzed patients with advanced NSCLC who received anti‐PD‐1/PD‐L1 monotherapy. Patients were stratified into two groups based on immunohistochemical analysis: high hepcidin expression (tumour proportion score [TPS] = 50%, n = 6) and none‐to‐moderate expression (TPS 0–49%, n = 39). Treatment duration, progression‐free survival (PFS), and overall survival (OS) were compared. In vitro mechanisms were also explored using hepcidin and interferon‐gamma (IFNg) co‐culture models. Results The high‐hepcidin group exhibited a longer median treatment duration (42 vs. 5 days; p = 0.051). This group also showed improved PFS (2222 vs. 126 days; hazard ratio [HR], 0.35 [95% CI, 0.16–0.79]; p = 0.0643) and OS (not reached vs. 459 days; HR, 0.46 [95% CI, 0.18–1.16]; p = 0.0994), although these differences did not reach statistical significance. In vitro experiments revealed that hepcidin enhanced IFNg‐induced reactive oxygen species production and cytotoxicity in A549 cells. Conclusion Advanced NSCLC patients with high hepcidin expression may derive greater benefit from anti‐PD‐1/PD‐L1 monotherapy. Our findings suggest that cellular sensitivity to IFNg varies by hepcidin expression levels.

Abstract Background Irritable bowel syndrome (IBS) is a prevalent condition in which disrupted communication between the gut and the brain plays a central role in disease manifestation. A dynamic, bidirectional regulatory network links the gut microbiota with the central nervous system (CNS), commonly referred to as the gut–brain axis, and this network exerts broad influence over both gastrointestinal physiology and psychological states. Aims This review specifically concentrates on direct microbiota–CNS communication pathways in the context of IBS. Materials and Methods This narrative review synthesizes existing experimental and clinical evidence related to gut microbiota–brain interactions, with a particular focus on neural and neuroendocrine signaling mechanisms. Results Gut microorganisms affect CNS function via microbial metabolites and neural signalling routes, whereas the CNS modulates gastrointestinal activity through integrated neural and endocrine pathways, involving the autonomic nervous system and the enteric nervous system. Given the complexity of these reciprocal interactions, this review specifically concentrates on direct microbiota–CNS communication pathways, with particular attention to neural circuits and the hypothalamic–pituitary–adrenal axis in the context of IBS. Conclusion Furthermore, we discuss how advances in precision stratification and multi‐omics approaches may facilitate the development of personalized therapeutic strategies, ultimately improving symptom control and restoring balance between gut microbial ecosystems and neural regulation.

Abstract Background The advent of cancer immunotherapy, particularly immune checkpoint inhibitors, has revolutionised the treatment landscape for many malignancies. However, a significant proportion of patients exhibit primary, adaptive, or acquired resistance to these therapies, limiting their broad clinical benefit. Tumour microenvironment (TME), a complex and dynamic ecosystem of cancer cells, immune cells, stromal cells and extracellular matrix, has emerged as a critical determinant of immunotherapy response and a major driver of resistance. Main body This review provides a comprehensive overview of the intricate mechanisms by which the TME fosters an immunosuppressive milieu, thereby impeding effective anti‐tumour immunity. We delineate the key cellular and molecular components of the TME that contribute to immunotherapy resistance, including tumour‐associated macrophages, myeloid‐derived suppressor cells, regulatory T cells, and cancer‐associated fibroblasts. We then discuss the diverse and innovative strategies being developed to reprogram the TME and overcome these resistance mechanisms. These emerging approaches include targeting immunosuppressive cell populations, modulating tumour metabolism, epigenetic reprogramming, normalising the tumour vasculature and employing engineered cell therapies such as chimeric antigen receptor T cells and oncolytic viruses. The potential of combination therapies to synergistically dismantle the immunosuppressive TME is also explored. Conclusion Reprogramming the TME represents a paradigm shift in cancer treatment, moving beyond direct tumour cell targeting to modulate the entire tumour ecosystem. While significant challenges remain, including TME heterogeneity and the need for predictive biomarkers, the strategies discussed herein hold immense promise for overcoming immunotherapy resistance and improving patient outcomes. A deeper understanding of the complex interplay between the tumour and its microenvironment will be crucial for the development of next‐generation, personalised immunotherapeutic strategies.

Abstract Background Patients receiving immunosuppressive regimens are at increased risk of life‐threatening infections from microorganisms. The comprehensive analysis of the types and composition of pathogenic microorganisms in clinical specimens from immunosuppressed patients is of great importance for prognosis and diagnosis. Methods Three hundred and sixty clinical samples from 199 patients were enrolled in this study. The results of metagenomics next‐generation sequencing (mNGS) detection were analysed, and diagnostic performance was compared with that of conventional methods (CMs). Differences in microbiological detection were analysed between immunosuppressant monotherapy and combination therapy. The differences in microflora between blood samples and non‐blood samples were also examined. Results The positive rate of the mNGS results was higher than that of the CMs results in all types of samples, and the number of pathogens detected by mNGS was also greater than that of CMs. The percentage of co‐infection detected by mNGS was 45.56% which was higher than that of CMs (6.67%, 24 out of 360). In the immunosuppressant single‐drug group, 58.1% of the samples were mixed infections, while in the two‐drug combination group, single infections accounted for 56% of the samples. Multiple pathogen infection (63.8%) was the most common type of infection in non‐blood samples, of which 28.4% were viral and bacterial co‐infections. A single pathogen accounted for 53.8% of the infections detected in blood samples, of which 35.9% were individual viruses. Non‐blood samples have a higher negative predictive value (88.9% vs. 77.5%) and accuracy (100% vs. 92%) than blood samples. Conclusions mNGS is superior to CMs for the diagnosis of infections in patients treated with immunosuppressants. The immunosuppressive medication regimen needs to be adjusted when the patients have serious infections. Blood samples could be used for rapid diagnosis of suspected viral infection, and samples from infectious sites have significant advantages in detecting bacterial and viral co‐infection.

Vascular endothelial cells (ECs) are critical guardians of vascular homeostasis, regulating angiogenesis, inflammation, and barrier integrity. However, their phenotypic and functional heterogeneity across vascular beds has posed challenges to traditional bulk analysis methods. Single-cell RNA sequencing (scRNA-seq) has emerged as a transformative tool, offering unparalleled resolution at the individual cell level. This technology has revolutionized our ability to dissect endothelial diversity and function and unveiled novel endothelial subtypes, unexpected signaling pathways, and dynamic responses to environmental stimuli (Figure 1). scRNA-seq now stands at the forefront of endothelial biology research, providing insights into both physiological and pathological processes. In vascular inflammation, scRNA-seq has transformed our ability to dissect endothelial plasticity and pathological transitions. For instance, McQueen et al. demonstrated how scRNA-seq in atherosclerotic lesions identified distinct endothelial subsets specializing in lipid handling, oxidative stress response, and leukocyte recruitment.1 Bondareva and Sheikh further highlighted that scRNA-seq platforms uncover vascular zonation patterns and region-specific endothelial responses to inflammation, thereby redefining our understanding of vascular homeostasis.2 In human heart failure, Rao et al. mapped fibrotic and non-fibrotic myocardial tissues, revealing that fibrotic-region ECs could upregulate adhesion molecules and foster leukocyte infiltration.3 Notably, tools for single-cell trajectory inference allowed dynamic modeling of endothelial activation, illustrating a continuum from quiescence to inflammation rather than discrete states. In oncology, scRNA-seq has improved the concept of tumour-associated ECs (TECs). Shiau et al. performed single-nucleus RNA-seq on pancreatic ductal adenocarcinoma samples and identified a “reactive EndMT” program enriched in hypoxia-driven signaling and mesenchymal traits, which correlates with poor outcomes.4 Yang et al. integrated scRNA-seq-derived TEC markers into a liver cancer prognostic model and then linked endothelial gene signatures to immune infiltration and therapy response.5 In gastric cancer, Chen et al. used longitudinal scRNA-seq to reveal endothelial expansion and pro-angiogenic activation following neoadjuvant chemotherapy,6 suggesting that endothelial remodeling is highly dynamic and therapy-responsive. Recent studies further highlight the critical role of vascular ECs in cancer progression. Using single-cell RNA sequencing, Zhang et al. characterized endothelial heterogeneity in colorectal cancer liver metastases and identified specific subpopulations that actively engage with immune cells.7 In gastric cancer, Yang et al. demonstrated that chemotherapy dynamically alters endothelial-immune crosstalk within the tumour microenvironment.8 In sum, these findings establish ECs as key regulators of both vascular structure and immune modulation in tumours. Beyond disease states, single-cell atlases have deepened our understanding of tissue-specific endothelial specialization. In human skin, He et al. identified six distinct blood vascular endothelial subtypes, each demonstrating specialized patterns of metabolism and immune modulation.9 Li et al. extended this outcome by showing that capillary ECs in the skin exhibited strong HLA-II expression profiles, suggesting active participation in immune surveillance.10 These dermal vascular maps reinforce the concept that endothelial identity is shaped by environmental cues and localized tissue demands. At the systems level, Augustin and Koh suggested rethinking the vascular endothelium as a widely distributed organ that actively shapes organ function, rather than merely acting as a passive barrier.11 This new perspective is increasingly supported by single-cell studies showing that ECs can secrete regulatory signals, adjust their metabolism, and remodel their local tissue environment in response to stress. Looking ahead, vascular research will increasingly depend on integrated single-cell multi-omics approaches—including transcriptomics, chromatin accessibility, proteomics, and spatial mapping—to fully capture the complexity of endothelial biology in both health and disease. As summarized by Shiau et al. and further expanded by Bondareva and Sheikh, ECs are inherently plastic, swiftly adapting to mechanical stress, metabolic changes, inflammatory cues, and therapeutic interventions.2, 4 The predictive power of scRNA-seq, especially when combined with trajectory analysis and intercellular communication analysis, firmly positions it as a cornerstone for precision vascular medicine. Beyond cataloging static cell types, single-cell technologies are evolving towards dynamic profiling of endothelial transitions over time. Longitudinal single-cell analyses have the potential to track how ECs shift during disease initiation, progression, and therapeutic response, moving our understanding from isolated snapshots to continuous biological narratives. Integrating complementary technologies, such as single-cell ATAC-seq to probe chromatin landscapes, spatial transcriptomics to preserve tissue architecture, and single-cell proteomics for functional validation, will further enhance mechanistic insights into endothelial regulation. To fully realize this potential, the development of comprehensive single-cell atlases across diverse populations, including pediatric, aging, and ethnically varied cohorts, will be critical for uncovering context-dependent vascular adaptations. Coupling large-scale single-cell datasets with machine learning may enable the prediction of endothelial states, disease trajectories, and therapeutic vulnerabilities. Ultimately, single-cell technologies not only promise to redefine vascular biology with unprecedented resolution but also to pave the way for precision interventions targeting endothelial dysfunction across a broad range of diseases. The authors declare no conflict of interest. Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.