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.
Purpose: Cancer progression is a non-linear, multiscale process driven by the interaction of molecular, cellular and tissue systems, which ultimately leads to tumour growth, invasion and metastasis. Understanding and predicting these dynamics is essential for improving diagnostics and personalising therapy. Mathematical and computational modelling has become central to this effort, enabling in silico simulations of progression and treatment response.
Methods: This survey outlines computational frameworks for modelling cancer across biological scales, incorporating mathematical formalisms and algorithmic paradigms. These frameworks include: ordinary and partial differential equation models of growth, angiogenesis and invasion; agent-based and hybrid multiscale approaches that capture heterogeneity and microenvironmental feedback; emerging digital twin platforms that integrate mechanistic and data-driven modelling for patient-specific predictions. Recent advances in artificial intelligence (AI), such as graph neural networks, transformer-based architectures and multimodal data fusion, enhance these frameworks by combining interpretability with predictive power through mechanistic learning pipelines.
Results: Based on these advances, the survey presents a conceptual roadmap for multiscale cancer modelling. This roadmap offers a structured workflow that begins with defining the biological question and modelling scale and continues through selecting appropriate paradigms, calibrating and validating models with experimental or clinical data and integrating multimodal information into individualised simulations. The long-term goal is to develop digital twins of cancer that can personalise and optimise therapy and ultimately guide clinical decisions. These models will capture processes ranging from single-cell dynamics and cancer progression to angiogenesis, tumour evolution and treatment response. To achieve this vision, we need advances in parameter calibration, multimodal dataset integration and cross-scale model validation.
Conclusion: Computational oncology bridges the gap between biological mechanisms and predictive modelling, offering a framework for reproducible, data-driven precision medicine. The convergence of mechanistic modelling, AI-assisted inference and digital twin technologies establishes a continuum for translating research into real-time, patient-specific decision support in oncology.
CD70, a tightly regulated immune co-stimulatory ligand under physiological conditions, becomes aberrantly and persistently overexpressed in a wide spectrum of malignancies, including renal cell carcinoma (RCC), glioblastoma and acute myeloid leukaemia. This tumour-restricted expression pattern, coupled with its potent immunomodulatory effects, positions CD70 as a uniquely actionable target in modern oncology. Unlike classical immune checkpoints, CD70 contributes to immune escape by chronically engaging CD27, leading to T cell exhaustion, regulatory T cell (Treg) expansion and myeloid-derived suppressor cell activation, collectively shaping a profoundly immunosuppressive tumour microenvironment. Mechanistically, CD70 expression is amplified by convergent oncogenic, inflammatory and hypoxic signals, most notably HIF-2α stabilization in VHL-deficient RCC, linking it to aggressive disease biology and poor prognosis. Notably, CD70 expression in tumours often coexists with dysregulation of receptor tyrosine kinase (RTK) pathways (e.g. MET, AXL) and immune checkpoint circuits (e.g. PD-1/PD-L1), supporting rationale combination therapies aimed at overcoming resistance and enhancing anti-tumour immunity. A diverse array of CD70-directed therapies, including antibody-drug conjugates, CAR-T and CAR-NK cells, monoclonal antibodies and radiotheranostic agents, are advancing through preclinical and early clinical pipelines. Beyond therapeutic applications, CD70 also holds promise as a biomarker for tumour detection, patient stratification and response monitoring. This mini-review synthesizes emerging insights into CD70-driven immune evasion, highlights evolving therapeutic strategies and discusses the integration of CD70 targeting into next-generation immunotherapy and precision oncology platforms. This review is novel in linking CD70-mediated immune escape to oncogenic RTK signalling and in foregrounding recent translational advances and rational combination strategies that have not been comprehensively covered in prior reviews.