Background:Lung adenocarcinoma (LUAD) is shaped by the tumor microenvironment, yet endothelial cell (EC) regulatory programs and their biological roles remain insufficiently defined. Methods:We analyzed scRNA-seq data to map EC-associated programs and applied hdWGCNA to identify EC modules and communication patterns. Network pharmacology integrated EC-module genes with LUAD-related targets to prioritize MYLK. MYLK expression and function were evaluated by RT-qPCR, immunohistochemistry, and gain-/loss-of-function assays in endothelial and LUAD cell models. We then performed network-based in silico knockout in LUAD tumors (GSE164789) and exploratory immune-cell eQTL analysis. Results:EC modules were enriched for junction organization, angiogenesis, and immune-related pathways, with extensive epithelial-stromal-endothelial interactions. Network pharmacology nominated MYLK as an EC-linked LUAD candidate. MYLK expression was reduced in LUAD and associated with unfavorable clinical outcomes. In endothelial cells, MYLK perturbation altered junction integrity and trans-endothelial tumor cell migration; in LUAD cells, MYLK gain/loss affected migration, invasion, and proliferation. In silico knockout of MYLK produced regulatory shifts enriched for tight junction organization, endothelial apoptosis, angiogenesis, vascular permeability, and vascular/cancer-related pathways. Immune-cell eQTL analysis identified an association between increased MYLK expression in dendritic cells and elevated LUAD risk. Conclusions:These findings define an endothelial-centered regulatory framework in LUAD and highlight the context-dependent, cell-type-specific roles of MYLK at the tumor-endothelial-immune interface.

Purpose:To investigate the prognostic value of cell cyclin A2 (CCNA2) in lung adenocarcinoma (LUAD) and to explore its mechanisms in promoting cancer progression. Patients and Methods:In this study, we employed an integrated strategy combining bioinformatics, clinical analysis and molecular biology to elucidate the role of CCNA2 in LUAD. First, comprehensive bioinformatics analyses were performed using public datasets. This included detecting the differential expression of CCNA2 in LUAD versus normal tissues, analyzing its correlation with patient survival and clinical characteristics, and employing Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes pathway (KEGG) analysis to predict the functions of CCNA2-associated genes. The relationship between CCNA2 expression and immune infiltration was further examined via the tumor immune estimation resource (TIMER) platform. The expression level of CCNA2 was also confirmed through reverse transcription-quantitative PCR and Western blotting. Additionally, the biological function of CCNA2 was evaluated by constructing an in vitro transfection model. Results:The results of the present study indicated that CCNA2 was upregulated in LUAD tissues. Cox regression analysis revealed that CCNA2 upregulation is a independent prognostic biomarker for LUAD. Additionally, CCNA2 was markedly associated with immune cell infiltration and immune checkpoint molecules. The results of in vitro experiments confirmed that knockdown of CCNA2 significantly inhibited the proliferation, invasion and migration of H1975 and H1299 cells. Furthermore, CCNA2 was found to promote the invasion and migration of lung cancer cells through the PI3K/AKT signaling pathway. Conclusion:The present research identified the prognostic signature and biological function of CCNA2 in LUAD, which suggested that CCNA2 may be a potential prognostic biomarker and a pivotal oncogenic driver for this disease.

Ammonium is a central component of the marine nitrogen cycle and a critical indicator of water quality. However, its susceptibility to environmental changes during sampling and analysis poses significant challenges for accurate measurement in marine waters. Here, we present iSEA-in situ, a submersible wet-chemistry sensor readily adapted from our previously developed programmable syringe pump-based flow analyzer. The sensor features a newly designed optical detection module, filtration module, and auto-calibration program tailored for in situ deployment. This sensor design offers a practical complement to microfluidics-and electrochemistry-based in situ sensors, circumventing the technical barrier of sophisticated microfabrication and sample matrix interference. Using a modified indophenol blue spectrophotometric method, the iSEA-in situ sensor demonstrates salinity-independent performance with high measurement frequency (6 h-1), precision (relative standard deviation of 2.7% for underwater analysis of 5 mu M standard solutions over 14 days), sensitivity (detection limit of 0.12 mu M), wide dynamic range (up to 20 mu M), and low reagent consumption (a total of 140 mu L per run). Field deployment in a fishing tank and coral tank in a nearshore marine station yielded a total of 795 measurements with only 4.2% outliers (>= 3 sigma residuals). Regular auto-calibration via measuring quality control samples confirmed sustained accuracy of the sensor (relative standard deviation <= 3.8%) over a week-long deployment. The iSEA-in situ sensor is capable of providing robust, high-temporal-resolution data to capture the dynamic interplay between marine biological activity and environmental fluctuations.

Accurate in-situ volume measurement of small (1 mm–10 cm) drifting underwater particles is critical for marine ecology and pollutant monitoring, yet it demands snapshot 3D imaging to avoid motion artifacts. Existing imaging techniques—including digital holography and conventional light field imaging—face a fundamental limitation in recovering the complete surface geometry of opaque and semi-transparent particles due to optical occlusion and limited perspective sampling. We overcome this challenge with a face-to-face dual light field camera (F2F-DLFC) system, which simultaneously captures both sides of a target under incoherent dark-field illumination. This dual-side snapshot strategy enables full 3D reconstruction of opaque particles, with experimental results showing volume errors below 6% for targets such as live fish and irregular pellets. While semi-transparent objects still present reconstruction challenges, this work establishes a foundational methodology for in-situ volumetric instrument development, providing a viable approach for accurate volumetry of a wide range of underwater particles.

Monitoring of bioluminescent algae blooms, referred to in China as “Blue Tears”, is essential for managing ecological risks, ensuring public safety, and accommodating the growing tourist attention they attract in coastal regions. However, this task remains challenging due to the nocturnal nature of these events, limited visibility, the lack of real-time monitoring infrastructure, and the absence of publicly available training datasets. To address this, we present BT3.8k, a novel dataset comprising 3827 images extracted from 241 user-uploaded videos shared on social media platforms. These samples cover diverse lighting, sea states, and viewpoints. Annotations were generated using Roboflow and SAM2, and subsequently refined using a YOLOv11m-seg model. We applied augmentations including Gaussian blur, salt-and-pepper noise, flipping, rotation, hue, brightness, and contrast to enhance generalizability. Using this dataset, we trained BT-YOLO, a YOLOv11 framework enhanced with the Convolutional Block Attention Module (CBAM), designed for coastal Blue Tear monitoring. Among multiple models evaluated, BT-YOLO achieved the highest performance with a Dice score of 85%, IoU of 77%, AP@50 of 80%, and 90 FPS on GPU. This approach offers a scalable, real-time solution for BT monitoring, with potential impact across marine ecology, eco-tourism management, and environmental governance.

Significance As a foundational component of marine ecosystems, plankton is an essential driver of biogeochemical cycles and global fisheries productivity. Consequently, obtaining accurate, high-resolution data on their abundance, distribution, and traits is a fundamental objective in oceanography. For decades, research relied on the manual collection and microscopic analysis of discrete water samples-a process that was inherently low-resolution, disruptive to delicate organisms, and incapable of capturing rapid dynamics. The development of in situ optical imaging enabled a decisive shift from disruptive "physical sampling" to continuous, non-invasive "optical sampling", vastly expanding the observational scale. However, this advance created a new bottleneck: the manual analysis of millions of resulting images became impractical. The integration of deep learning is now propelling a second transformative phase. By automating the interpretation of image data, artificial intelligence is converting imaging systems from passive data loggers into active perceptual platforms. This evolution from simply imaging plankton to programmatically understanding their ecology is critical for advancing both fundamental marine science and operational monitoring capabilities. Progress The field has progressed through an integrated technological chain encompassing imaging hardware, data foundations, intelligent algorithms, and ecological quantification. A diverse suite of in situ imaging instruments has been developed, each representing a fundamental compromise among core parameters such as resolution, field of view, and sampling volume, which define their observational capacity and limitations. The data they generate are primarily organized into full-frame images, which retain spatial context, and cropped regions-of-interest for individual analysis. A major bottleneck for subsequent analysis is the high cost and expertise required for manual annotation, which shapes the available public datasets and motivates the use of AI-based techniques to enhance image quality degraded by the underwater environment. Deep learning now drives automated information extraction, with research advancing along two main paths: the classification of individual targets, which is progressing from closed-set to more ecologically realistic open-set recognition, and object detection directly in full-frame images, though the latter is constrained by scarce pixel-level annotations, spurring interest in synthetic data generation. The ultimate goal is to invert these image features into ecological parameters. This involves estimating individual size and biovolume, which is subject to significant uncertainty due to the inherent challenge of converting 2D projections to 3D volume, as well as deriving biomass via allometric models and extracting functional traits. Furthermore, automated recognition feeds into the calculation of biodiversity indices, although managing the propagation of algorithmic errors, such as species misidentification, into these higher-order ecological metrics remains a key challenge. Conclusions and Prospects In situ optical imaging, powered by AI, is advancing plankton observation from data sampling to intelligent perception. Future progress hinges on integrated advances in three key areas. The first is perceptual intelligence, achieved through novel multi-modal imaging sensors. The second is algorithmic intelligence, moving towards adaptive and efficient models. The third is system intelligence, realized via edge-cloud networked observational systems. This technological convergence will both deepen fundamental ecological research and enable critical operational applications, ranging from ecological forecasting to environmental monitoring. Ultimately, it will cement the role of this approach as a cornerstone of next-generation marine observing systems.

In situ pH sensing is crucial for the real-time monitoring of ocean acidification and investigations into the marine carbon cycle. Although ion sensitive field-effect transistor (ISFET) has been proven suitable for marine pH monitoring, its supply and implementation remain challenging. An underwater pH sensor for environmental analysis (uSEA-pH) based on ISFET was developed herein, incorporating a modified commercial laboratory pH probe through engineering design. Laboratory characterization demonstrated that uSEA-pH exhibited a Nernstian response (slope -57.60 ± 1.05 mV/pH, R2 > 0.999), rapid response time (∼7 s), and low measurement uncertainty (<0.01 pH). The sensor supports a sampling frequency of 1 Hz with an average power consumption of only 0.72 W. Its compact design (self-contained with battery: Φ15 × 45 cm; miniaturized version: Φ6.4 × 21 cm) facilitates deployment on various observational platforms. During high-frequency underway monitoring in the Pearl River Estuary and Dongshan Bay, uSEA-pH successfully detected subtle pH variations (<0.05 pH). In extended in situ deployments, buoy-mounted uSEA-pH reliably recorded tidal-driven pH fluctuations in Dapeng Bay (27 days) and Xiamen Bay (7 days), generating over 2.3 million field measurements. This study presents a viable, robust, and high-resolution approach for continuous pH monitoring in estuarine and coastal areas.

We propose what we believe to be a novel LED-based light-sheet microscopy-in-flow system that integrates scattering and chlorophyll a fluorescence imaging for phytoplankton analysis. Compared to conventional laser light sources, LED light sources exhibit significantly lower coherence, effectively reducing speckle noise. This enhancement improves the quality of scattering imaging and helps preserve the fine structural details of phytoplankton cells. The fluorescence images facilitate phytoplankton detection and reveal the intracellular distribution of chlorophyll/chloroplasts, while the scattering images provide complementary morphological information, particularly for non-fluorescent cell structures. This dual-modality imaging approach enables a more comprehensive characterization of phytoplankton and other particles in natural water samples. Experimental results demonstrate that the system can simultaneously capture both chlorophyll a fluorescence and scattering images of flowing phytoplankton cells. With its ability to provide enhanced imaging quality and structural insights, this LED-based light-sheet microscopy-in-flow system has strong potential for automated and accurate phytoplankton water sample analysis, contributing to advancements in aquatic environment monitoring and oceanographic research.

Automated plankton recognition models face significant challenges during real-world deployment due to distribution shifts (Out-of-Distribution, OoD) between training and test data. This stems from plankton's complex morphologies, vast species diversity, and the continuous discovery of novel species, which leads to unpredictable errors during inference. Despite rapid advancements in OoD detection methods in recent years, the field of plankton recognition still lacks a systematic integration of the latest computer vision developments and a unified benchmark for large-scale evaluation. To address this, this paper meticulously designed a series of OoD benchmarks simulating various distribution shift scenarios based on the DYB-PlanktonNet dataset [27], and systematically evaluated twenty-two OoD detection methods. Extensive experimental results demonstrate that the ViM [57] method significantly outperforms other approaches in our constructed benchmarks, particularly excelling in Far-OoD scenarios with substantial improvements in key metrics. This comprehensive evaluation not only provides a reliable reference for algorithm selection in automated plankton recognition but also lays a solid foundation for future research in plankton OoD detection. To our knowledge, this study marks the first large-scale, systematic evaluation and analysis of Out-of-Distribution data detection methods in plankton recognition. Code is available at https://github.com/BlackJack0083/PlanktonOoD.

Estuaries are critical for land-ocean carbon exchange, but coupling mechanisms between air-sea CO2 fluxes (FCO2) and phytoplankton gross primary productivity (GPP) remain poorly understood. This study used high-frequency underway monitoring in the Lingdingyang Estuary to resolve spatiotemporal interplays between FCO2 and GPP. Annual mean FCO2 was 20.29 +/- 23.34 mol C m(-2) yr(-1), with flooding season (82.97 +/- 80.49 mmol C m(-2) d(-1)) an order of magnitude higher than dry season. Gross primary productivity averaged 2.23 +/- 2.07 mol C m(-2) yr(-1), increasing significantly during flooding. The results revealed a distinct "source-to-sink" FCO2 gradient, with a 116% reduction over similar to 40 km, primarily driven by phytoplankton activity. Biological processes explained 30-50% of FCO2 variability. While net autotrophy in the mid-estuary reduced FCO2 by 48.6 mmol C m(-2) d(-1) during flooding, heterotrophic activity downstream offset 40-60% of GPP-driven uptakes. This study quantifies how urban estuary oscillate between carbon source and sink states, providing parameters for blue carbon frameworks and demonstrating that eutrophication-driven loads reduce overall carbon sequestration efficiency through enhanced heterotrophic activities.

Underwater microscopes have been extensively developed to acquire in situ images for marine plankton observation. However, existing software relies on traditional thresholding methods to isolate single targets, followed by further processing of extracted regions of interest (ROI) for species classification and enumeration. This two-stage approach lacks end-to-end efficiency and struggles in real oceanic environments. We propose a novel pipeline that synthesizes large-scale, controllable, and diverse darkfield images with realistic content mimicking raw images captured from in situ settings, facilitating the training of plankton detection algorithms without the need of human annotation. A YOLO-v9 model trained exclusively on this synthetic dataset demonstrates excellent localization performance and strong generalizability to real images. This method well utilizes established ROI datasets and significantly reduces the reliance on cumbersome manual annotation, paving the way for advancements in deep object detection and segmentation networks, ultimately enhancing the automation and performance of plankton imaging systems.

Sepsis is a common systemic disease characterized by various physiological and pathological disorders. It can result from infection by various pathogens, such as bacteria, viruses, and fungi. The rate of culture-negative sepsis is almost 42%, indicating that most patients may have nonbacterial infections. With the outbreak of coronavirus disease 2019, viral sepsis has attracted growing attention because many critically ill patients develop sepsis. Viral sepsis can be caused by viral infections and combined with, or secondary to, bacterial infections. Understanding the common types of viral sepsis and the main characteristics of its pathogenesis will be helpful for effective diagnosis and treatment, thereby reducing mortality. Early identification of the causative agent of viral sepsis can help reduce the overuse of broad-spectrum antibiotics. In this article, we reviewed the common viruses of sepsis, their potential pathophysiology, targets of diagnosis, and remedies for viral sepsis.

The red Noctiluca scintillans (RNS) blooms often break out near Pingtan Island, in the northern Taiwan Strait from April to June. It is essential to gain insights into their formation mechanism to predict and provide early warnings for these blooms. Previous studies and observations showed that RNS blooms are the most likely to occur when winds are weak and shifting in direction. To explore this phenomenon further, we employed a high-resolution coastal model to investigate the hydrodynamics influencing RNS blooms around Pingtan Island from April to June 2022. The model results revealed that seawater exhibited weak circulation but strong stratification during RNS blooms. Residence time were examined through numerical experiments by releasing passive neutrally buoyant particles in three bays of Pingtan Island. The results showed a significantly longer residence time during RNS blooms, indicating reduced flushing capabilities within the bays, which could give RNS a stable environment to multiply and aggregate. This hydrodynamic condition provided a favorable basis for RNS blooms breakout near Pingtan Island. The shifts and weakening of the prevailing northeast wind contributed substantially to weakening the flow field around Pingtan Island and played a crucial role in creating the hydrodynamics conducive to RNS blooms. Our study offers fresh insights into the mechanisms underpinning RNS blooms formation near Pingtan Island, providing a vital framework for forecasting RNS blooms in this region.

Darkfield imaging can achieve in situ observation of marine plankton with unique advantages of high-resolution, high-contrast and colorful imaging for plankton species identification, size measurement and abundance estimation. However, existing underwater darkfield imagers have very shallow depth-of-field, leading to inefficient seawater sampling for plankton observation. We develop a data-driven method that can algorithmically refocus planktonic objects in their defocused darkfield images, equivalently achieving focus-extension for their acquisition imagers. We devise a set of dual-channel imaging apparatus to quickly capture paired images of live plankton with different defocus degrees in seawater samples, simulating the settings as in in situ darkfield plankton imaging. Through a series of registration and preprocessing operations on the raw image pairs, a dataset consisting of 55 000 pairs of defocused-focused plankter images have been constructed with an accurate defocus distance label for each defocused image. We use the dataset to train an end-to-end deep convolution neural network named IsPlanktonFE, and testify its focus-extension performance through extensive experiments. The experimental results show that IsPlanktonFE has extended the depth-of-field of a 0.5× darkfield imaging system to ~7 times of its original value. Moreover, the model has exhibited good content and instrument generalizability, and considerable accuracy improvement for a pre-trained ResNet-18 network to classify defocused plankton images. This focus-extension technology is expected to greatly enhance the sampling throughput and efficiency for the future in situ marine plankton observation systems, and promote the wide applications of darkfield plankton imaging instruments in marine ecology research and aquatic environment monitoring programs.

The development of remaining oil plays an important role in increasing late production, and low-grade faults seriously impact residual oil exploration and development. Low-grade faults have a small fault displacement, brief extension, and strong concealment, which hinders their prediction using the typical semantic segmentation network. To intelligently identify low-grade faults, we designed a codec target edge detection technique. For the network to fully learn the low-grade faults information, we constructed the encoder using dilated convolution. Next, we introduced an attention mechanism to the decoder to improve the capture of location information from a shallow network and semantic information from a deep network. Finally, the multiscale fusion decoder outputs fault information of different scales, which further improves the identification accuracy of low-grade faults. The training model is applied to simulated data and actual seismic data through ablation experiments. The results show that this method can effectively identify low-grade faults and overcomes problems such as blurred fault cross-location, thicker edge contour lines, lower detection accuracy, and less training data. Compared with conventional holistically nested edge detection (HED) and semantic segmentation (UNet), fault misidentification is reduced, fault continuity is increased, and fault accuracy is improved, providing technical support for the exploration and development of remaining oil and increasing the recovery rate of old oil fields.

Biofouling is a common challenge for underwater sensors, especially for long-term in situ monitoring in marine environments. In this study, we assessed the antifouling efficacy of a paint containing a natural product camptothecin (CPT) on six materials (316 L stainless steel, TC4 titanium alloy, 7075 aluminum alloy, polyoxymethylene, polyvinyl chloride, and Teflon), which are frequently used in the construction of underwater sensor housings. Additionally, a buoy-based sea-trial was performed to test the antifouling performance of the CPT-based paint on housings of three in situ sensors used for practical seawater monitoring applications, namely a spectrophotometer for chemical oxygen demand (COD) measurements and two fluorimeters for biochemical oxygen demand (BOD) and chlorophyll a (Chl a) concentration measurements. The results showed significantly lower macrofouling coverage on the areas painted with the CPT-based paint compared to the unpainted areas for each tested material over 9 months of seawater immersion. The CPT-based paint exhibited different antifouling performance for the different materials; in particular, it exhibited better antifouling performance on the plastic materials compared to the metal materials. Furthermore, when applied on submersible sensor housings in the sea-trial test, the CPT-based paint kept the housings of the COD sensor and the Chl a sensor clean for over 4 months. In addition, the paint prevented fouling of the BOD sensor housing even after 6 months of seawater immersion. Thus, our results suggest that the CPT-based paint could be used as a potential solution to control the biofouling of sensor housings for long-term in situ applications in marine environments.

Sorafenib (SOR) is an effective chemotherapy drug for hepatocellular carcinoma, renal cell carcinoma, and differentiated thyroid carcinoma. However, a long-standing clinical issue associated with SOR use is an increased risk of cardiotoxicity, but the underlying mechanisms remain obscure. Here we report that ferroptosis of cardiomyocytes is responsible for SOR-induced cardiotoxicity. The specific ferroptosis inhibitor ferrostatin-1 and deferoxamine mesylate, an iron chelator, significantly alleviate SOR-induced cardiac damage. RNA-sequencing revealed that endoplasmic reticulum (ER) stress and the unfolded protein response were predominately activated, which might be attributed to the lipid reactive oxygen species-mediated perturbation of the ER. Activating transcription factor 4 (ATF4) is one of the most significantly up-regulated genes, knockdown of ATF4 exacerbates cardiomyocyte ferroptosis induced by SOR, while overexpression of ATF4 promotes cell survival. Mice with AAV-mediated ATF4 knockdown exhibit lipid peroxidation and more severe cardiomyopathy. Further experiments demonstrated that ATF4 exerts its protective role by elevating SLC7A11 expression, a transport subunit of system Xc-, which promotes cystine uptake and glutathione biosynthesis. The cardioprotective effect of ATF4 was diminished by SLC7A11 knockdown in cardiomyocytes subjected to SOR treatment. Taken together, these findings show that ferroptosis of cardiomyocytes is an important cause of SOR-related cardiotoxicity. ATF4 acts as a key regulator to promote cardiomyocytes survival by up-regulation of SLC7A11 and suppression of ferroptosis.

In the dataset for IsPlanktonCLR model training, the images in the Training Set and Testing Set 1 are color ROIs. We use their L channel as the input for training network and ab channel as the ground truth to validate the network, respectively. The Testing Set 2 contains 60 grayscale-color image pairs. The grayscale images are used as network input to generate colorized images, and their color counterparts are used as ground truth to validate the colorization performance. We train the IsPlanktonCLR network for 600 epochs with batch size of 16 on an NVIDIA Tesla A100 GPU. We use Adam optimizer with an initial learning rate 0.001, which decays every 400 epochs. The input image size is standardized to 224× 224 pixels as done in [8].

Underwater imaging with red-NIR light illumination can avoid phototropic aggregation-induced observational deviation of marine plankton abundance under white light illumination, but this will lead to the loss of critical color information in the collected grayscale images, which is non-preferable to subsequent human and machine recognition. We present a novel deep networks-based vision system IsPlanktonCLR for automatic colorization of in situ marine plankton images. IsPlanktonCLR uses a reference module to generate self-guidance from a customized palette, which is obtained by clustering in situ plankton image colors. With this self-guidance, a parallel colorization module restores input grayscale images into their true color counterparts. Additionally, a new metric for image colorization evaluation is proposed, which can objectively reflect the color dissimilarity between comparative images. Experiments and comparisons with state-of-the-art approaches are presented to show that our method achieves a substantial improvement over previous methods on color restoration of scientific plankton image data.

Rapid and quantitative analysis of phytoplankton cells in natural seawater is of great need for marine ecological science research and harmful algae bloom monitoring applications. In this paper, we propose a YOLOX network-based object detection algorithm exclusively for high-throughput real-time analysis of phytoplankton fluorescence images collected by the FluoSieve ® imaging flow cytometer. Based on an active learning strategy, we first annotate and construct a FluoPhyto dataset of red tide phytoplankton species fluorescence images commonly found in the South and East China Sea, which contains a total of 30,339 images in 32 different categories. Using the dataset, we train the Faster-RCNN, SSD, YOLOv3 and YOLOX networks, and the comparison result shows that the performance of YOLOX network outperforms the other networks, which can reach a mean average precision (mAP) of 90.9%. The trained YOLOX model is then deployed on an embedded GPU module and the inference speed is tested to reach 20 fps with the help of TensorRT optimization, which can hopefully meet the real-time detection requirements of the instrument. In addition, the algorithm is run on the embedded platform for detection of images collected in a red tide event that happened near the Pearl River Estuary, and the key parameters such as abundance and size spectrum of the dominant species, Cochlodinium geminatum, are obtained, which confirms the feasibility and superior performance of the detection algorithm.