Organ- and organoid-on-a-chip technologies provide critical human-relevant models for drug screening. Despite their promise, translating their complex biological outputs into reproducible, quantifiable, and pharmacologically interpretable readouts remains a significant challenge. To address this, artificial intelligence (AI) is increasingly employed to process the high-content imaging, sensor, and molecular data derived from these platforms. Crucially, this integration elevates AI from a conventional post-experimental analytical tool into a comprehensive framework that actively drives quality control, response quantification, model integration, and critical screening decisions. This review examines AI-augmented microphysiological systems across the entire drug screening pipeline by connecting biological readouts with specific computational strategies and pharmacological endpoints. We evaluate representative platforms, analytical methodologies, and specific applications where computational frameworks enable model standardization, robust phenotype interpretation, mechanism-informed evaluation, and compound prioritization. Furthermore, we outline primary barriers to clinical translation, including inherent biological and engineering variability, model generalizability, the need for external validation, and clinical dose relevance. Ultimately, these insights establish a comprehensive framework for evaluating the reproducibility, pharmacological applicability, interpretability, and translational potential of AI-driven microphysiological screening.

Peptide therapeutics are increasingly attractive due to their high potency and specificity, yet their clinical application is often limited by short systemic half-life, enzymatic instability, and rapid clearance. Here, we report a minimally invasive injectable in situ forming gel (ISG) composed of phospholipids dissolved in ethyl oleate (EO), replacing conventional hydrophilic organic solvents, which rapidly forms a depot upon exposure to aqueous environments. By adjusting the phospholipid-to-EO ratio, we systematically characterized the ISG's microstructural evolution, rheological behavior, water absorption, and drug release profile, revealing a watertriggered phase transition mechanism toward a lamellar gel structure. Compared with conventional hydrophilic organic solvent-based formulations, the EO-based ISG exhibited superior biocompatibility and enhanced protection of encapsulated peptides. Sustained in vivo release of peptides with distinct physicochemical properties was maintained for over one month, and the liraglutide-loaded ISG demonstrated prolonged glucosecontrol efficacy in a diabetic mouse model. Collectively, this study establishes a simple, scalable ISG platform for long-term peptide delivery, offering a promising strategy to enhance peptide stability and therapeutic performance.

Elucidating the mechanisms of organotropic metastasis requires the in situ profiling of exosomes within the tumor microenvironment. However, conventional assays are normally based on the isolation of exosomes from biological fluids, which disrupts their native spatial context. Moreover, these assays frequently exhibit limited analytical sensitivity, and the types of exosome phenotypes detected simultaneously are inherently restricted by fluorescent spectral overlap. To address the issue, we present an integrated microfluidic platform coupling a biomimetic tumor-stroma coculture system with a nanoplasmonic antibody barcode array, enabling the in situ profiling of exosomes within the cellular microenvironment on-chip. By utilizing gold nanoislands (GNIs) as substrates, the device leverages localized surface plasmon resonance (LSPR) to achieve fluorescence enhancement and improved sensitivity. By employing spatial barcoding, this system overcomes the constraints imposed by spectral overlap and circumvents the limitations associated with the number of conventional multicolor fluorescence detection channels. This platform possesses the capability of conducting multiplexed qualitative analyses of exosome biomarkers. Downstream machine learning decoded these high-dimensional profiles, resolving distinct exosomal subpopulations on the basis of their unique phenotypic profiles. Overall, this platform establishes a robust, universal strategy for investigating exosome-mediated intercellular communication and metastatic niche formation.

Cell-cell interactions are crucial for understanding various physiological and pathological processes, yet conventional population-level methods fail to disclose the heterogeneity at a single-cell resolution. Single-cell coculture systems that isolate and cultivate single-cell pairs can help reveal heterogeneous interactions between different types of individual cells. However, precise and high-throughput pairing of individual cells for long-term coculture remains challenging. Meanwhile, tools for analyzing single-cell data sets have lagged due to the increased data throughput. Herein, we report a deep learning-assisted high-throughput addressable single-cell coculture system (DL-HASCCS), enabling fast pairing of individual heterogeneous cells and quantitative analysis of single-cell interactions in a high-throughput manner by integrating high-throughput single-cell cocultivation and automated data processing. By analyzing the interaction between single breast cancer cells and single endothelial cells under normal and chemotherapy conditions, the effect of cell-cell interactions on cell proliferation and migration is revealed at the single-cell level, providing valuable insights into cellular heterogeneity.

Sweat, a biofluid rich in various biomarkers, offers significant potential for non-invasive health monitoring and disease screening. Colorimetric detection is well-suited for multi-analyte quantification and point-of-care testing in sweat analysis, while conventional platforms often suffer from detection inaccuracies due to subjective interpretation and environmental interference. Furthermore, many existing systems rely on complex fabrication or computationally demanding artificial intelligence models, limiting their scalability and practical use in resource-limited settings. To address these challenges, we developed a YOLOv5-aided paper-based microfluidic intelligent sensing platform that integrates an easily fabricated paper-based microfluidic chip, smartphone imaging, and a deep learning framework which attains a mean average precision of 99.5%. This platform provides a cost-effective, portable, and reproducible approach for the detection of multiplex biomarkers in sweat, and its functionality has been successfully validated through the colorimetric analysis of iron ions, chloride ions, and glucose in sweat.

The efficacy of CDK4/6 inhibitors as anti-tumor agents, especially in breast cancer, has been constrained by direct treatment-associated toxicities and the development of drug resistance. To overcome these limitations, we developed a novel lipid nanoparticle (LNP) platform utilizing microfluidic technology, a pioneering approach for the co-loading of small interfering RNA (siRNA) and hydroxychloroquine (HCQ). This innovative strategy leverages the synergistic effects of siRNA-mediated CDK4/6 silencing, which induces cell cycle inhibition, and HCQ-facilitated suppression of autophagy, enhancing anti-tumor therapy. Additionally, HCQ plays a pivotal role in improving the delivery efficiency of nucleic acid drugs by facilitating endosomal escape. In vitro studies demonstrated that co-delivery of siCDK4/6 and HCQ effectively blocked autophagy, arrested the cell cycle, induced cellular senescence, and significantly reduced tumor cell proliferation. Subsequent in vivo experiments confirmed the superior anti-tumor efficacy of this co-administration strategy compared to single-agent treatments, without observable adverse effects. This microfluidics-based LNP platform offers an engineerable strategy for the simultaneous delivery of small molecule drugs and nucleic acids, thereby providing immense potential for broader applications in cancer therapy.

Injectable hydrogels have wide applications in clinical practice. However, the development of tough and bioadhesive ones based on biopolymers, along with biofriendly and robust crosslinking strategies, still represents a great challenge. Herein, we report an injectable hydrogel composed of maleimidyl alginate and pristine gelatin, for which the precursor solutions could self-crosslink via mild Michael-type addition without any catalyst or external energy upon mixing. This hydrogel is tough and bioadhesive, which can maintain intactness as well as adherence to the defect of porcine skin under fierce bending and twisting, warm water bath, and boiling water shower. Besides, it is biocompatible, bioactive and biodegradable, which could support the growth and remodeling of cells by affording an extracellular matrix-like environment. As a proof of application, we demonstrate that this hydrogel could significantly accelerate diabetic skin wound healing, thereby holding great potential in healthcare.

This review outlines the current advances of high-throughput microfluidic systems accelerated by AI. Furthermore, the challenges and opportunities in this field are critically discussed as well.

Targeted gels have been attractive and regarded as a kind of promising adsorptive media for bioanalysis due to their advantages of high specific surface area, enough stability, and adjustable porous structure. Recently, targeted gel media have been applied for separation and enrichment of various biomolecules from different biological samples. Moreover, targeted gel media have been introduced into surface-enhanced Raman scattering (SERS) technology to eliminate matrix effect for rapid and accurate analysis of biological samples. In this article, we introduced the preparation methods of targeted gel media including in-situ polymerization method, self-assembly method and one-pot method. Then, the functions of targeted gel media in bioanalysis including separation-enrichment media, SERS substrates, and target-responsive units were also reviewed in detail. Additionally, the applications of targeted gel media for biomarker analysis, bioimaging analysis and cell engineering were further discussed. Finally, we tried to elucidate the trend and perspective into the future landscape of targeted gel media for bioanalysis.

Monitoring changes in the expression of marker proteins in biological fluids is essential for biomarker-based disease diagnosis. Epithelial cell adhesion molecule (EpCAM) has been identified as a broad-spectrum biomarker for various chronic diseases and as a therapeutic target. However, the development of simple and reliable methods for quantifying EpCAM changes in biological fluids faces challenges due to the variability of its expression across different diseases, the presence of soluble forms, and matrix effects. In this paper, a surface-enhanced Raman scattering (SERS)-fluorescence (FL) dual-mode sensing method was established for quantification of trace EpCAM in biological fluids based on bimetallic Au@Ag nanoparticles and nitrogen-doped quantum dots encapsulated DNA hydrogel hybrid with graphene oxide (Au@Ag-NQDs/GO). The DNA hydrogel was constructed based on three-dimensional (3D) structure DNA-mediated strategy using an aptamer DNA (AptDNA) linker. The interaction of the AptDNA with EpCAM triggered the disassembly of the DNA hydrogel. Consequently, the release of Au@Ag nanoparticles induced an "on-off" switch in the SERS signal while the weakened FL quenching effect in Au@Ag-NQDs/GO system achieved "off-on" switch of FL signal, enabling the simultaneous SERS-FL quantification of EpCAM. The established dual-mode method exhibited outstanding sensitivity and stability in quantifying EpCAM in the range of 0.5-60.0 pg/mL, with the limits of detection (LODs) of SERS and FL as 0.17 and 0.35 pg/mL, respectively. When applied for real sample analysis, the method showed satisfactory specificity and recoveries in cancer cells lysate, serum, and urine samples with RSDs of 2.8-6.3%, 4.0-6.3%, and 2.8-5.7%, respectively. The developed SERS-FL sensing method offered a sensitive, reliable, and practical quantification strategy for trace EpCAM in diverse biological fluid samples, which would benefit the early diagnosis of disease and further health management.

Organs-on-chips (OoCs) are in vitro models that simulate the anatomical and physiological structure of human organs at the micro-scale level. With the advantages of miniaturization, low reagent consumption, and precise controllability, they exhibit great potential in the studies of human disease, development of new drugs, and personalized medicine. To achieve accurate control and record in OoCs, it is required to integrate biosensors for precisely monitoring cells (i.e., cellular metabolism, functions, and responses to stimuli) and cellular microenvironment. This review outlines the current advances of biosensors-integrated OoC platforms for various biomedical applications, including monitoring cells, monitoring cellular microenvironment and simultaneous monitoring cells and the conditions of their microenvironment. Furthermore, different types of biosensors integrated into the OoCs are compared thoroughly by discussing their working principles, integration techniques, key findings, advantages and disadvantages. In the end, the challenges and future perspectives in this field are critically discussed.

DNA-mediated self-assembly technology with good sensitivity and affinity ability has been rapidly developed in the field of probe sensing. The efficient and accurate quantification of lactoferrin (Lac) and iron ions (Fe3+) in human serum and milk samples by the probe sensing method can provide useful clues for human health and early diagnosis of anemia. In this paper, contractile hairpin DNA-mediated dual-mode probes of Fe3O4/Ag-ZIF8/graphitic quantum dot (Fe3O4/Ag-ZIF8/GQD) NPs were prepared to realize the simultaneous quantification of Lac by surface-enhanced Raman scattering (SERS) and Fe3+ by fluorescence (FL). In the presence of targets, these dual-mode probes would be triggered by the recognition of aptamer and release GQDs to produce FL response. Meanwhile, the complementary DNA began to shrink and form a new hairpin structure on the surface of Fe3O4/Ag, which produced hot spots and generated a good SERS response. Thus, the proposed dual-mode analytical strategy possessed excellent selectivity, sensitivity, and accuracy due to the dual-mode switchable signals from "off" to "on" in SERS mode and from "on" to "off" in FL mode. Under the optimized conditions, a good linear range was obtained in the range of 0.5-100.0 μg/L for Lac and 0.01-5.0 μmol/L for Fe3+ and with detection limits of 0.14 μg/L and 3.8 nmol/L, respectively. Finally, the contractile hairpin DNA-mediated SERS-FL dual-mode probes were successfully applied in the simultaneous quantification of iron ion and Lac in human serum and milk samples.

The assessment of cell migration and proliferation is essential in the field of oncology. It has been widely performed for cancer prognosis and the prediction of treatment outcome. The microdevice-based methods have enabled the assessment of these two processes at single-cell resolution, which could acquire unique information on cell heterogeneity and subtypes. However, most of the current platforms show limited throughput due to design defects or lack of modules for high-speed data analysis, which greatly hampers the extraction of statistically unbiased biological data. To address this challenge, we propose a high-throughput system consisting of an addressable dual-nested microwell array chip (DNMA chip) and an artificial intelligence (AI)-based image analysis algorithm. Our DNMA chip allows single-cell trapping, label-free encoding, and long-term incubation. Combined with AI-aided data processing, the migration and proliferation of single tumor cells under normal culture or chemotherapy are quantitatively analyzed in a high-throughput and non-destructive manner.

The development of an in vitro colorectal cancer (CRC) model that reconstructs the physiopathological microenvironment of human colorectal cancer shows great potential in accelerating the study of CRC mechanisms and the development of anti-CRC drugs. To this end, a pathomimetic colorectal tumor-on-a-chip (CRT-chip) that can effectively recapitulate the in vivo dynamic physiology and pathology of primary CRC is proposed. It not only allows controllable seeding and cultivation of CRC cells to emulate the site-specific occurrence of primary CRC, but also provides continuous low-speed flows and peristalsis-like deformation on colorectum epithelium to simulate the fluid shearing and peristalsis in human colorectum. Moreover, a channel that transports nutrients to CRC cells is designed specifically to mimic the function of tumor vessels. Based on the pathomimetic CRT-chip, the therapeutic effect of a photothermal anticancer drug was tested and quantified, exhibiting its great potentials in in vitro evaluation of anti-CRC drugs.

DNA-mediated nanotechnology has become a research hot spot in recent decades and is widely used in the field of biosensing analysis due to its distinctive properties of precise programmability, easy synthesis and high stability. Multi-mode analytical methods can provide sensitive, accurate and complementary analytical information by merging two or more detection techniques with higher analytical throughput and efficiency. Currently, the development of DNA-mediated multi-mode analytical methods by integrating DNA-mediated nanotechnology with multi-mode analytical methods has been proved to be an effective assay for greatly enhancing the selectivity, sensitivity and accuracy, as well as detection throughput, for complex biological analysis. In this paper, the recent progress in the preparation of typical DNA-mediated multi-mode probes is reviewed from the aspect of deoxyribozyme, aptamer, templated-DNA and G-quadruplex-mediated strategies. Then, the advances in DNA-mediated multi-mode analytical methods for biological samples are summarized in detail. Moreover, the corresponding current applications for biomarker analysis, bioimaging analysis and biological monitoring are introduced. Finally, a proper summary is given and future prospective trends are discussed, hopefully providing useful information to the readers in this research field.

Cell-laden hydrogel microstructures have been used in broad applications in tissue engineering, translational medicine, and cell-based assays for pharmaceutical research. However, the construction of cell-laden hydrogel microstructures in vitro remains challenging. The technologies permitting generation of multicellular structures with different cellular compositions and spatial distributions are needed. Herein, we propose a laser-guided programmable hydrogel-microstructures-construction platform, allowing controllable and heterogeneous assembly of multiple cellular spheroids into spatially organized multicellular structures with good bioactivity. And the cell-laden hydrogel microstructures could be further leveraged for in vitro drug evaluation. We demonstrate that cells within hydrogels exhibit significantly higher half-maximal inhibitory concentration values against doxorubicin compared with traditional 2D plate culture. Moreover, we reveal the differences in drug responses between heterogeneous and homogeneous cell-laden hydrogel microstructures, providing valuable insight into in vitro drug evaluation.

Circular RNA (circRNA) is a novel class of non-coding RNA that is a promising biomarker for the diagnosis and prognosis of diseases, such as cancer. Sensitive and quantitative detection of circRNA is of particular significance. However, conventional qRT-PCR is not appropriate for accurate quantification due to the strand-substitution potential during the reverse transcription, which has severely hindered the studies of circRNA. Here, we present a sensitive assay for the detection of circRNA using reverse transcription-hyperbranched rolling circle amplification (RT-HRCA). We further developed droplet digital RT-HRCA (ddRT-HRCA) for circRNA quantification using the microfluidic device. We demonstrate the ddRT-HRCA, modeled on circHIPK3, can achieve a limit of detection (LOD) as low as 6.6 aM with excellent selectivity. As far as we know, this is the first digital isothermal platform for quantitative analysis of low-abundance circRNA, which could be applied in liquid biopsy for early diagnosis and prognosis. This novel strategy paves new avenues for the development of the digital platform for accurate quantification of circRNA, to further promote this cutting-edge research area.

Microfluidic platforms have been employed as an effective tool for drug screening and exhibit the advantages of lower reagent consumption, higher throughput and a higher degree of automation. Despite the great advancement, it remains challenging to screen complex antibiotic combinations in a simple, high-throughput and systematic manner. Meanwhile, the large amounts of datasets generated during the screening process generally outpace the abilities of the conventional manual or semi-automatic data analysis. To address these issues, we propose an artificial intelligence-accelerated high-throughput combinatorial drug evaluation system (AI-HTCDES), which not only allows high-throughput production of antibiotic combinations with varying concentrations, but can also automatically analyze the dynamic growth of bacteria under the action of different antibiotic combinations. Based on this system, several antibiotic combinations displaying an additive effect are discovered, and the dosage regimens of each component in the combinations are determined. This strategy not only provides useful guidance in the clinical use of antibiotic combination therapy and personalized medicine, but also offers a promising tool for the combinatorial screenings of other medicines.

Using low Cd accumulation cultivars and managing field water regimes are effective measures to mitigate Cd accumulations in rice grains. However, the effect of the cultivar-water condition interaction (CWI) on grain Cd accumulations has largely been ignored. To solve this problem, pot and hydroponic experiments were conducted using 14 rice cultivars and two contrasting water conditions. The results showed that CWI significantly affected Cd concentrations in rice grains and roots, explaining 8.8% and 22.8% of the total variance, respectively. These CWI effects were derived from cultivar-dependent variations in rhizosphere soil properties [Eh, pH and available Cd associated with root radial oxygen loss (ROL)] and root Cd uptake. In this context, cultivar HH61 exhibited low, stable Cd accumulations, owing to its stably lower translocation rate, root Cd uptake ability and available Cd in its rhizosphere than the other cultivars, which was induced by its lower ROL. Root-to-grain Cd translocation rates were vital in determining Cd accumulations in grain of different cultivars but were independent from CWI. These results indicated that CWI could play an important role in Cd accumulation in rice while stable low-Cd cultivar should possess low ROL under flooding and low root-to-grain Cd translocation rate. The results will provide novel theoretical basis for cultivar selection and hence benefit the extensive use of low-accumulation cultivars and public health.