Abstract Effective screening for colorectal cancer (CRC) remains limited by the invasiveness of colonoscopy and the insufficient sensitivity of conventional liquid biopsy approaches. Here, we present Machine Learning-Assisted, Ago-Mediated Droplet Fluorescence Analysis with a Knowledge-Enabled Agent (MADAK-Agent), an integrated analytical platform that combines an Argonaute-mediated cascade amplification module with a monodisperse droplet-assisted, object-resolved post-amplification fluorescence measurement workflow and a knowledge-enabled interpretation interface. We developed a dual-channel cascade system coupling the exponential amplification reaction (EXPAR) with Clostridium perfringens Argonaute (CpAgo)-mediated cleavage, enabling the conversion of trace circulating miR-21 and miR-92a into distinct fluorescent signals. Following EXPAR–CpAgo cascade amplification, the reaction products were compartmentalized into monodisperse microdroplets using a coaxial flow-focusing microfluidic chip. Droplet-level fluorescence quantification yielded calculated limits of detection (LODs) of 45.6 aM for miR-21 and 59.2 aM for miR-92a. Compared with bulk fluorescence readout, the LODs were improved by 5.20-fold for miR-21 and 3.09-fold for miR-92a. Automated droplet segmentation and feature extraction were achieved using a Cellpose-based analysis module, reducing operator-dependent variability. Based on the extracted droplet features, a support vector machine (SVM) classifier achieved area under the receiver operating characteristic curve (AUC) values of 0.901 for miR-21 and 0.848 for miR-92a in a cohort of 60 clinical serum samples. To enable evidence-grounded interpretation, we further implemented a retrieval-augmented generation (RAG)-based question-answering (QA) agent with dual operating modes. Both the QA agent and the image analysis module are deployable as web-based and local tools, supporting flexible and privacy-preserving use. MADAK-Agent establishes a unified and deployable framework for sensitive droplet-based miRNA detection with integrated analysis, interpretation, and reporting.

Continuous in vivo monitoring of biomarkers remains challenging due to limited sensitivity, integration, and biocompatibility. Here, we report an integrated microneedle-based electrochemiluminescence device (MN-ECLD) for real-time detection of protein biomarkers in interstitial fluid. Leveraging hydrogen-bonded organic frameworks with ultrabright, biocompatible electrochemiluminescence, the emitters were incorporated into porous gold-coated microneedle arrays and regulated via interface-specific Y-shaped probes, enabling efficient coreactant-free signal generation. The device achieved ultrasensitive protein detection in vitro with a linear range of 100 fg/mL to 10 ng/mL, a detection limit of 21.3 fg/mL, and stability over 12 days, delivering an 87-fold sensitivity enhancement over conventional emitters. In vivo, MN-ECLD enabled real-time monitoring of cardiac biomarkers, achieving early warning of acute myocardial infarction in rats and pigs, with biomarker trends consistent with serum ELISA. This work establishes a versatile platform for continuous in vivo diagnostics of acute cardiovascular and metabolic disorders.

Aptamers are widely used in biosensing and targeted therapeutics, yet reported data remain fragmented across unstructured text, tables, and figures. Existing databases are limited in coverage and diversity, which constrains computational modeling. Here, we present AptaNexus (https://www.aptanexus.com/), a multitier aptamer database containing over 12,000 sequences targeting 1900 molecular entities, curated from literature published between 2005 and 2025. Its extraction pipeline, Dual-LLM Extraction with Reverse Grounding (Dual-LLMs+RG), achieves an F1 score of 0.970 at a fraction of the cost of the state-of-the-art model. Records are stratified into four quality tiers, supporting both experimental selection and machine learning applications. Beyond conventional keyword search, AptaNexus incorporates the Model Context Protocol (MCP) and an embedded conversational agent, Chat Aptamer, enabling natural-language queries that return structured, source-linked, and application-oriented recommendations. For example, users can request detection of a target in blood to receive a ranked list of aptamers with validated sensor platforms or query drug delivery targets to obtain functionally annotated candidates. By combining large-scale, evidence-grounded data with agent-accessible interfaces, AptaNexus makes aptamer information readily accessible across disciplines and, for the first time, provides native AI-agent interoperability.

Personalized health management aims to promote, maintain, and restore the health of individuals. Despite the ever-lasting research efforts involved in personalized healthcare bioelectronics, current healthcare platforms still face barriers such as costly facilities, specialized operations, and resource-limited applications. Therefore, personalized and user-friendly healthcare bioelectronics are urgently needed. Among emerging solutions, the integration of artificial intelligence (AI) and advanced bioelectronics is a pivotal approach that merges intelligent algorithms with multi-functional healthcare design. This review summarizes the latest advances in AI-assisted bioelectronics, aiming to provide a possible strategy for personalized healthcare applications. Initially, a brief survey is provided to discuss the material design, device fabrication, AI-hardware integration, and performance assessment of AI-assisted bioelectronics. The subsequent contents focus on the implementation of AI-assisted healthcare bioelectronics across health monitoring, early diagnosis, therapeutic treatment, and rehabilitation. Finally, we discuss the current challenges and prospective future developments in closed-loop healthcare bioelectronics, ultimately empowering individuals with control over their own health.

Recent progress in wearable electronics has pioneered a revolutionary technology for real-time monitoring of sweat biomarkers, particularly for chronic disease management. However, stable biosensing and accurate onbody monitoring remains challenging owing to sweat dynamics in real-world environment. Here we report a self-powered wearable biosensor for stable biosensing of sweat glucose, capable of autonomous sampling localized sweat. The wearable system integrates with an iontophoresis module, a capillary retention burst valvescontrolled microfluidic for sweat collection, MXene-functionalized stretchable biofuel cell-array patch, and reusable flexible printed circuit board. This wearable system allows for on-demand sweat collection up to a maximum volume of 41.4 & micro;L and monitoring glucose in the range of 1 & micro;M to 400 & micro;M, with the detection limit of 1 & micro;M. And the BFC array patch delivered an open circuit voltage of 0.412 V in the complicated environment. Its mechanism is comprehensively rationalized by theoretical and experimental results. This work illustrates the potential of autonomous wearable technology and self-powered flexible devices for complicated real-world settings.

Large language models (LLMs) hold significant promise in the field of medical diagnosis. There are still many challenges in the direct diagnosis of hepatocellular carcinoma (HCC). α-Fetoprotein (AFP) is a commonly used tumor marker for liver cancer. However, relying on AFP can result in missed diagnoses of HCC. We developed an artificial intelligence (AI) agent centered on LLMs, named ChatExosome, which created an interactive and convenient system for clinical spectroscopic analysis and diagnosis. ChatExosome consists of two main components: the first is the deep learning of the Raman fingerprinting of exosomes derived from HCC. Based on a patch-based 1D self-attention mechanism and downsampling, the feature fusion transformer (FFT) was designed to process the Raman spectra of exosomes. It achieved accuracies of 95.8% for cell-derived exosomes and 94.1% for 165 clinical samples, respectively. The second component is the interactive chat agent based on LLM. The retrieval-augmented generation (RAG) method was utilized to enhance the knowledge related to exosomes. Overall, LLM serves as the core of this interactive system, which is capable of identifying users' intentions and invoking the appropriate plugins to process the Raman data of exosomes. This is the first AI agent focusing on exosome spectroscopy and diagnosis, enhancing the interpretability of classification results, enabling physicians to leverage cutting-edge medical research and artificial intelligence techniques to optimize medical decision-making processes, and it shows great potential in intelligent diagnosis.

Extrachromosomal circular DNA (eccDNA) has emerged as a novel biomarker for cancer detection due to its tumor-specific amplification and stable structure in circulation. However, its clinical application is hindered by extremely low abundance in biofluids and the lack of robust detection techniques. To address this, we screened for tumor-associated eccDNA biomarkers and developed NPCC (Nested PCR-CRISPR/Cas12a), a novel method combining nested PCR for ultrasensitive amplification with CRISPR/Cas12a for sequence-specific detection. The assay employs two rounds of junction-specific PCR to enrich eccDNA, followed by CRISPR/Cas12a-mediated cleavage guided by target-specific crRNA. Validation using synthetic circular DNA standards demonstrated a limit of detection (LoD) of 10-6 fM, representing a >100-fold improvement over conventional PCR, with no cross-reactivity to linear or genomic DNA fragments. In plasma samples from 88 cancer patients, NPCC successfully detected multiple tumor-specific eccDNAs, including the hepatocellular carcinoma marker eccDNA-HCC-1 (AUC = 0.8977). NPCC overcomes key technical barriers in liquid biopsy, offering a cost-effective, highly sensitive, and specific platform for noninvasive cancer diagnostics.

Advancing clinical diagnostics requires platforms that combine catalytic efficiency, biocompatibility, and real-time, in vivo accessibility. Herein, this study reports a structurally integrated FePc-ZIF-8-MX nanozyme that combines the redox activity of FePc, the porous confinement of ZIF-8, and the electrical conductivity of MX. Synthesized via a low-energy, ambient-condition process, this hybrid enables efficient electron transfer, enhanced analyte enrichment, and sustained catalytic activity in physiological environments. To translate this functionality into a wearable diagnostic format, the hybrid is seamlessly incorporated into a microneedle array, offering minimally invasive access to interstitial fluid for continuous L-cysteine (L-Cys) monitoring. The resulting platform exhibits high selectivity and sensitivity across complex biological matrices, including serum, urine, cultured cells, and a murine model of myocardial infarction. This study presents a multifunctional electrochemical platform that enables on-body metabolite monitoring through a microneedle-integrated nanozyme interface. To the best of our knowledge, it constitutes the first realization of real-time, in vivo L-Cys sensing in this format, setting a new benchmark for precision biosensing in translational healthcare.

Wearable theranostics hold great promise in precision medicine and real-time monitoring of diseases, which are capable of performing both predictive analysis and therapeutics concurrently. This review, for the first time, provides a detailed description of wearable device applications in theranostic studies, including smart contact lenses, smart bandages, smart dressing, and wearable theranostic dental patch, as well as the practical use of wearable therapeutic devices in clinical settings. In addition, wearable technologies enhanced by machine learning techniques, which are capable of enhancing the precision of theranostics through autonomous learning in early stage diagnosis and treatment, image analysis, and predictive analytics, are analyzed. Third, the situation of commercialization of these bioelectronics is summarized. Finally, the existing challenges and future directions for translation and commercialization of the wearable theranostics are discussed in detail. AI-assisted wearable theranostic systems are transitioning from laboratory innovations to clinical applications, enabling intelligent and convenient clinical translation and deployment models, as a paradigm shift in modern personalized medicine.

Over the past decade, the number and diversity of identified protein post-translational modifications (PTMs) have grown significantly. However, most PTMs occur at relatively low abundance, making selective enrichment of modified peptides essential. To address this, we developed a thermodynamic model describing the free beads enrichment in suspension enrichment process and derived a theoretical relationship between material dosage and analyte recovery. The model predicts a non-linear trend, with enrichment efficiency increasing up to an optimal dosage and declining thereafter—a pattern confirmed by experimental data. We validated the model using centrifugation-based enrichment for glycosylated peptides and magnetic-based enrichment for phosphorylated peptides. In both cases, the results aligned with theoretical predictions. Additionally, the optimal dosage varied among peptides with the same modification type, highlighting the importance of tailoring enrichment strategies. This study provides a solid theoretical and experimental basis for optimizing PTMs enrichment and advancing more sensitive, accurate, and efficient mass spectrometry-based proteomic workflows.

In the original publication [...].

Background: Influenza viruses are major pathogens responsible for respiratory infections and pose significant risks to densely populated urban areas. RT-qPCR has made substantial contributions in controlling virus transmission during previous COVID-19 epidemics, but it faces challenges in terms of detection time for large sample sizes and susceptibility to nucleic acid contamination. Methods: Our study designed loop-mediated isothermal amplification primers for three common influenza viruses: A/H3N2, A/H1N1, and B/Victoria, and utilized a 4-channel microfluidic chip to achieve simultaneous detection. The chip initiates amplification by centrifugation and allows testing of up to eight samples at a time. Results: By creating a closed amplification system in the microfluidic chip, aerosol-induced nucleic acid contamination can be prevented through physically isolating the reaction from the operating environment. The chip can specifically detect A/H1N1, A/H3N2, and B/Victoria and has no signal for other common respiratory viruses. The testing process can be completed within 1 h and can be sensitive to viral RNA at concentrations as low as 10−3 ng/μL for A/H1N1 and A/H3N2 and 10−1 ng/μL for B/Victori. A total of 296 virus swab samples were further analyzed using the microfluidic chip method and compared with the classical qPCR method, which resulted in high consistency. Conclusions: Our chip enables faster detection of influenza virus and avoids nucleic acid contamination, which is beneficial for POCT establishment and has lower requirements for the operating environment.

Loop-mediated isothermal amplification (LAMP) holds great promise for rapid nucleic acid detection. However, its application is hindered by the notable occurrence of false positives. In this work, gold nanorod (AuNR)-mediated hot start effect, which adsorbs primers and controls the release of single-stranded DNA (ssDNA) primers at about 50 °C, was demonstrated. The AuNR-mediated LAMP reaction for detecting African horse sickness virus (AHSV) showed high selectivity against other DNA fragments, even at high magnesium concentrations. Furthermore, the assay achieves excellent sensitivity with a detection limit of 100 copies/μL. Additionally, the method displays good repeatability from 50 to 65 °C (optimal), and the coefficient of variation of Tt at the same concentration of nucleic acids is less than 5%. Our study optimizes the LAMP reaction using AuNRs, thereby reducing false positives and offering great potential for clinical diagnosis.

Rapid and sensitive detection of Epstein-Barr virus cell-free DNA (EBV cfDNA) is crucial for early diagnosis and monitoring of nasopharyngeal carcinoma (NPC), but accessibility to screening is limited by complicated and costly conventional DNA isolation and purification approaches. Here, a fully integrated ion concentration polarization (ICP)-enriched and nanozyme-catalyzed lateral flow assay (ICP-cLFA) is developed, enabling total analysis of EBV cfDNA in whole blood samples, with DNA isolation, pre-concentration, and amplification performed on a microfluidic chip, consequently providing the signal readout within 75 min. Specifically, ICP preconcentration and amplification steps, together with target recognition catalyzed by a platinum-decorated mesoporous gold nanosphere (MGNS@Pt) nanozyme, result in an ultralow detection limit of 4 aM in standard cfDNA samples and 100 aM in whole blood from NPC-bearing rats. The high sensitivity and specificity of the ICP-cLFA suggest strong potential for cfDNA screening in resource-limited clinical and field applications.

While massive studies are focused on platinum (Pt)-based nanozyme for antitumor therapies, their therapeutic efficiency is deficient due to the weak catalytic activity in the highly complex tumor microenvironment. Herein, mesoporous gold nanospheres confined platinum nanoclusters (MGNSs@Pt) as robust hydroxyl radical and oxygen nanogenerators are achieved for multimodal therapies. Benefiting from the confinement effect of the mesopores in the MGNSs, the Pt nanoclusters (Pt NCs) demonstrate enhanced stability and catalytic activity, with a catalytic constant (Kcat) of 1.42 × 106 s-1, which is 2 and 5 orders magnitude higher than Kcat values of Pt-decorated non-porous gold nanoparticles and pure Pt NCs respectively. Density functional theory (DFT) calculations reveal the proper interaction of intermediates contributes to the ultra-high catalytic activity of MGNSs@Pt. Meanwhile, owing to the local surface plasmon resonance (LSPR) effect in the second near-infrared (NIR-II) bio-window of MGNSs, the nanozymes exhibited high photothermal conversion efficiency up to 43.4%, which enhanced the nanocatalytic damage on cancer cells. This process can induce robust oxidative stress and oxygenation within the tumor, thereby activating the apoptosis pathway for tumor eradication by mitochondrial dysfunction, cell membrane disruption, HIF-1α downregulation as well as caspase 3 activation, which pave the way for multimodal and effective cancer treatment.

Continuous measurement of macromolecular biomarkers in vivo could enable diverse implantable applications in personalized medicine. However, technical obstacles remain: current technologies are limited to only one-way tracking of increases or decreases in macromolecule levels and lack real-time feedback on disease progression. Here, we propose an integrated diagnosis-therapy sensing system for dynamic tracking of cell-free DNA and drug delivery, based on a semi-implantable indwelling needle modified with clustered regularly interspaced short palindromic repeats (CRISPR)-dCas9. The specific binding-dissociation sensing mechanism of CRISPR-dCas9 with target DNA on the surface under fluctuating blood flow is discussed in detail, with various reaction equilibrium constants. Owing to the matched mechanical properties and geometrical structure of the semi-implantable device, it shows the ability to withstand interference of 60 % fetal bovine serum, sensitivity of 300 fM, 3-day stability, and real-time feedback on target DNA level in animal models. For sepsis patients bearing Staphylococcus aureus, the biosensor exhibited clinical sensitivity and specificity of 92.3 % and 100 % respectively. It could enable dynamic monitoring of cell-free DNA in vivo and timely intervention for patients with acute syndromes such as sepsis in the intensive care unit (ICU).

The early diagnosis of lung cancer is crucial for improving patient prognosis, and liquid biopsy plays an important role in early lung cancer screening. In the field of liquid biopsy, label-free Surface-Enhanced Raman Scattering (SERS) possesses its unique advantages as it can provide comprehensive information about the insights into the chemical makeup of serum. However, the nonuniform SERS substrates poses challenges for reliable clinical diagnosis. We design a Surface Controlled SERS with Improved Ensemble Learning (SCSIEL) for lung cancer screening and diagnosis. This finely controlled SERS substrate ensures the uniformity of the surface and the high reproducibility of SERS spectra. The improved ensemble learning used in SCSIEL consists of a multi- layer structure which is inspired from the residual connection of deep learning networks. Although the framework is lightweight and integrates only a few simple base models, it achieves impressive results under the carefully constructed network. This SCSIEL system is also validated by direct analysis of clinical serum samples from 168 lung cancer patients and 100 healthy controls and the excellent performance is obtained with an area under the curve (AUC) of 97.0 % and accuracy of 93.4 %, which outperforms that of the clinical biomarkers for lung cancer. This SCSIEL is also explainable, which indicates the enhanced protein degradation in lung cancer. The SCSIEL method is a reliable and cost-effective method in the screen of lung cancer, shows great promise in clinical implementation.

Rapid and accurate diagnostic methods are crucial for managing viral gastroenteritis in children, a leading cause of global childhood morbidity and mortality. This study introduces a novel microfluidic-Flap endonuclease 1 (FEN1)-assisted isothermal amplification (MFIA) method for simultaneously detecting major viral pathogens associated with childhood diarrhea—rotavirus, norovirus, and adenovirus. Leveraging the specificity-enhancing properties of FEN1 with a universal dspacer-modified flap probe and the adaptability of microfluidic technology, MFIA demonstrated an exceptional detection limit (5 copies/μL) and specificity in the simultaneous detection of common diarrhea pathogens in clinical samples. Our approach addresses the limitations of current diagnostic techniques by offering a rapid (turn around time <1 h), cost-effective, easy design steps (universal flap design), and excellent detection performance method suitable for multiple applications. The validation of MFIA against the gold-standard PCR method using 150 actual clinical samples showed no statistical difference in the detection performance of the two methods, positioning it as a potential detection tool in pediatric diagnostic virology and public health surveillance. In conclusion, the MFIA method promises to transform pediatric infectious disease diagnostics and contribute significantly to global health efforts combating viral gastroenteritis.

Argonaute proteins (Agos) have emerged as key tools in molecular diagnosis because of their ability to precisely target specific bases and perform multi-target cleavage. Previous reviews on Agos mainly covered their roles in nucleic acid detection and biological functions. However, recent advances in molecular diagnostics have led to the development of new technologies, methods, platforms, and targets using Agos. This review provides a succinct overview of Agos, including their classification, structural domains, and functional mechanisms, with a focus on recent advancements in Ago-based biosensors for molecular diagnosis from 2019 to 2024. The signal transduction strategies used in Ago-based biosensors and novel technological applications are reviewed and discussed, exploring methods to achieve high sensitivity and the multiplex detection of nucleic acids and non-nucleic acid biomarkers (such as proteins and small molecules). We also present our perspectives on the future development and challenges in creating next-generation molecular diagnostic technologies based on Agos.