Clear cell renal cell carcinoma (ccRCC), the most common subtype of RCC, requires accurate pathological grading for effective prognosis. However, current grading methods rely heavily on subjective pathologist assessment, leading to variability. While generative artificial intelligence (GenAI) has shown promise in medical imaging, its application in digital pathology remains underexplored. This study evaluates the performance of three multimodal GenAI models—GPT-4o, Claude-3.5-Sonnet, and Gemini-1.5-Pro—in ccRCC grading and prognosis prediction. A total of 499 ccRCC slides from The Cancer Genome Atlas and 349 external samples from two independent cohorts were analyzed. A standardized prompt repetition mechanism and variance-based stability validation method guided GenAI models in extracting 17 pathological features. Feature stability was assessed using intraclass correlation coefficient (ICC). These features, combined with 3 clinical variables, were used to build grading and prognostic models via logistic regression and 113 machine learning algorithms. Performance was benchmarked against CellProfiler, ResNet-50, DenseNet-121, attention-based multiple instance learning (MIL) and Pathology Language and Image Pre-training, using the concordance index (C-index) and area under the receiver operating characteristic curve (AUC). Claude-3.5-Sonnet outperformed the other two GenAI models (ICC = 0.76; micro-average AUC = 0.87), exceeding ResNet-50 (AUC = 0.78) and attention-based MIL (AUC = 0.70). Its top prognostic models achieved an average C-index of 0.739, effectively stratifying high- and low-risk patients. Key predictors included stage, calcification, sarcomatoid differentiation, and vascular networks. GenAI, particularly Claude-3.5-Sonnet, enhances accuracy and consistency in ccRCC pathology, showing strong potential for clinical use, especially in resource-limited settings.

Acute myeloid leukemia (AML) is a devastating hematological malignancy and one of the most prevalent forms of leukemia in adults. Despite recent advancements and approval of novel targeted therapies, drug resistance remains a formidable clinical challenge. In this study, we conducted an unbiased CRISPR-Cas9 knockout screen in AML cells to uncover novel mediators of resistance to the clinically approved FLT3 inhibitor gilteritinib. This screen identified chromodomain helicase DNA binding protein 7 (CHD7) as a new regulator of drug resistance. Strikingly, CHD7 loss not only conferred resistance to FLT3 inhibitors but also extended resistance to a broad range of therapeutics, including venetoclax and daunorubicin (DNR). Mechanistic investigations integrating transcriptomic and proteomic data revealed that CHD7 deletion upregulates angiopoietin-1 (ANGPT1), which drives resistance by activating the PI3K/AKT and MAPK/ERK signaling pathways. Significantly, genetic knockdown of ANGPT1 or pharmacological inhibition of its receptor TIE2 partially restored drug sensitivity in CHD7-deficient cells. Together, these findings identify the CHD7-ANGPT1 axis as a novel mechanism of multi-drug resistance in AML. Preclinical studies further suggest that combining targeted therapies with TIE2 inhibitors offers a promising strategy to overcome drug resistance in AML.

The management of osteoarthritis (OA) and osteochondral defects faces challenges due to heightened catabolic activity from pro-inflammatory mediators and a lack of reparative cells. Transforming growth factor β1 (TGFβ1) plays a crucial role in cartilage maintenance and cellular recruitment, making it a promising therapeutic target. However, the high cost and unpredictable effects of exogenous TGFβ1 limit its clinical application. Notably, TGFβ1 is primarily found in a latent form within joint tissues, especially during early injury stages. This study proposes that activating endogenous TGFβ1 may serve as an effective treatment strategy. We demonstrate that an ascorbic acid (AA) and ferric chloride (AA/Fe) Fenton reaction system can activate latent TGFβ1 in knee joint tissues. Treatment with AA/Fe-activated synovial fluid protects chondrocytes from interleukin1β-induced damage and enhances chemotactic responses in joint tissues. Intra-articular AA/Fe injections in rats significantly phosphorylated SMAD2/3 and reduced cartilage degradation. Additionally, we developed poly(lactic-co-glycolic acid) microspheres for sustained AA/Fe release within a thermosensitive or photocrosslinkable hydrogel, showing high biocompatibility. These formulations effectively prevented cartilage degeneration and promoted osteochondral repair. Our findings confirm that AA/Fe reliably activates endogenous TGFβ1, providing a novel cell-free and cost-efficient treatment approach for OA and osteochondral defects.

Microfluidics has rapidly advanced in medical laboratory science (MLS) applications due to its precise detection capabilities, allowing for analysis with minimal sample volume and fast response time. However, traditional data processing methods in MLS face considerable challenges when extracting meaningful insights from large, complex datasets. Artificial intelligence (AI), with its advanced algorithms in machine learning, image processing, and pattern recognition, provides robust analytical tools that enhance data extraction and interpretation, advancing the development of intelligent microfluidics applications in MLS. This review presents the first comprehensive summary of AI-driven microfluidics applications in MLS. Initially, the review introduces the basic concepts of AI and its advantages in data analysis. It then outlines the limitations of microfluidics, followed by a detailed discussion on the unique advantages of integrating AI with microfluidics. Next, the review presents various AI-driven applications in microfluidic systems for detecting cells, bacteria, nucleic acids, proteins, emphasizing innovative methodologies in these areas. Finally, the review discusses current challenges and explores potential solutions of AI-driven microfluidics for MLS.

Mitochondria are the foundation of cellular energy metabolism and are crucial for cell growth and development. Mitochondrial dysfunction can disrupt cellular energy metabolism and normal cellular functions, contributing to the onset of related diseases. The functionality of mitochondria is influenced by various associated proteins and molecules, including mitofusin 2, optic atrophy 1, dynamin related protein 1, translocase of the inner membrane 23, translocase of the outer membrane 40, PTEN-induced kinase 1, reactive oxygen species modulator 1, nicotinamide adenine dinucleotide dehydrogenase, mitochondrial voltage-dependent anion channel and mitochondrial DNA. We also discussed the role of mitochondrial targeting sequences in mitochondrial proteins. The abnormal expression of these proteins and molecules can impair mitochondrial network remodeling, which is essential for maintaining the quantity and quality of mitochondria and facilitating the exchange of substances between them. This review elucidates the relationship between mitochondrial network remodeling, dysfunction-induced diseases, and associated proteins and outlines current methods for detecting mitochondrial networks and functions, thereby providing strategies for the study of mitochondrial dysfunction -related diseases.

Extracellular vesicles (EVs) are lipid bilayer structures secreted by cells that act as intercellular messengers. Tissue-derived EVs (TEVs), harvested from the tissue interstitium, directly reflect the actual physiological or pathological state of the tissue microenvironment. However, the difficulty in tissue acquisition severely limits the development of TEVs. In contrast, organoids are 3D cell clusters cultured from stem cells, which have spatial structures and physiological functions that are highly similar to the source tissues. Although organoid-derived EVs (OEVs), isolated from culture supernatants, can reflect complex cellular interactions, they cannot directly reflect the state of the tissue microenvironment like TEVs. Building on the foundation of TEVs and OEVs, we introduce the innovative concept of organoid-tissue EVs (OTEVs), where residing in the organoid interstitium. Acting as a communication bridge between OEVs and TEVs, OTEVs can accurately represent the true microenvironment. They overcome the challenges associated with the limited availability of TEVs and the inability of OEVs to directly reflect the microenvironment. We believe that OTEVs will synergize with TEVs and OEVs to enhance the understanding of the pathogenesis of complex diseases, as well as to improve their diagnosis and treatment.

Patient-derived xenograft (PDX) models provide a robust preclinical platform that preserves the genetic and phenotypic heterogeneity of patient tumors while mirroring their tumor genetic characteristics, which retain malignant cells and the tumor pathological structure, making them valuable for studying tumor progression and developing anticancer therapies. This review outlines the establishment of PDX models and their applications in tumor research by comparing their attributes and limitations with other experimental models. It explores the use of PDX models in understanding tumor progression mechanisms, resistance mechanisms, and treatment strategies, including radiotherapy, chemotherapy, targeted therapy, immunotherapy, nanotherapy, cell therapy, antibody-drug conjugates, and combination therapy. Integration of PDX models with multi-omics technologies is also discussed, along with their application in co-clinical and clinical studies. Notably, this review covers approximately 30 cancer types and aims to guide future cancer research.

The integration of Surface-Enhanced Raman Scattering (SERS) with machine learning heralds a transformative era in cancer management, offering a non-invasive, expedited, and comprehensive approach for early diagnosis, targeted therapy, and continuous monitoring. As SERS penetrates the molecular intricacies of cancerous tissues, its conjunction with advanced machine learning algorithms enhances diagnostic accuracy, enabling the discernment of subtle biochemical cues critical for early-stage detection and precise therapeutic targeting, and holds promise for establishing a systematic platform for cancer from diagnosis to therapy. This review explores the synergistic potential of these technologies advocating for their expanded application across the diagnostic spectra and images to revolutionize the therapeutic landscape of cancer. By harnessing this integrated approach, we propose the development of an intelligent platform that promises to refine cancer management, thereby redefining oncological diagnostics and care.

Despite extensive research, currently, no biomarkers are available for clinical use in intracranial aneurysm (IA). However, recent advancements in high-throughput technologies have opened new possibilities for understanding IA's complex pathophysiology. These technologies generate extensive datasets across multiple molecular domains, including genomics, transcriptomics, proteomics, metabolomics, and microbiomics. By integrating multi-omics data, it is now possible to analyze thousands of genes, RNAs, proteins, metabolites, and microbiota simultaneously, revealing complex interaction networks across these molecular layers. This integrative approach offers significant insights into IA pathogenesis, helps identify potential therapeutic targets, and paves the way for discovering novel biomarkers. In this review, we provide an overview of the current understanding of IA pathology and highlight recent advances in biomarker research through multi-omics approaches. We also propose strategies for integrating multi-modal data to accelerate the development of molecular tools for the diagnosis, rupture prediction, prognosis, and personalized treatment of IA.

Isothermal amplification (IA) techniques have emerged as promising alternatives to polymerase chain reaction (PCR), enabling rapid and efficient nucleic acid amplification at constant temperatures. Meanwhile, clustered regularly interspaced short palindromic repeat (CRISPR)/Cas-based technology has revolutionized molecular diagnostics by harnessing collateral cleavage for programmable biomarker detection. Especially, the combination of IA techniques with CRISPR/Cas systems establishes a synergistic framework for next-generation diagnostic platforms, addressing the sensitivity-specificity trade-offs and challenges posed by resource-limited settings inherent in conventional methods. In this review, we trace the developmental milestones of IA techniques and elucidate the fundamental mechanisms of relevant IA techniques and CRISPR/Cas systems. Synergistic strategies for coupling IA techniques with CRISPR/Cas systems were summarized. Subsequently, personalized applications including point-of-care testing devices, droplet microfluidics platforms for rare biomarker detection, and IA/CRISPR/Cas-based nanosensors were introduced. Finally, several challenges and perspectives in the field are discussed.

Cancer immunotherapy is an innovative treatment approach that leverages the immune system to combat tumors, demonstrating significant therapeutic potential. In the past few years, single-cell RNA sequencing (scRNA-seq) has made significant progress in the field of cancer immunotherapy, enabling us to understand the complexity of anti-tumor immune processes with unprecedented depth and precision, thereby facilitating the design of more effective immunotherapy strategies. This review aims to summarize the recent applications of advanced scRNA-seq technologies in tumor immunology. First, we outline the most representative scRNA-seq technologies with different technical principles, with a particular focus on single-cell T cell receptor sequencing. Next, we describe how scRNA-seq technology is applied to identify the cellular composition and phenotypes within the tumor microenvironment for the construction of immune cell atlas, and uncover key cell types and molecular mechanisms underlying treatment responses for developing novel immunotherapies. Finally, we address the current challenges and future prospects of scRNA-seq technology in tumor immunology.