Knowledge graph construction for heart failure using large language models with prompt engineering.

Introduction: Constructing an accurate and comprehensive knowledge graph of specific diseases is critical for practical clinical disease diagnosis and treatment, reasoning and decision support, rehabilitation, and health management. For knowledge graph construction tasks (such as named entity recognition, relation extraction), classical BERT-based methods require a large amount of training data to ensure model performance. However, real-world medical annotation data, especially disease-specific annotation samples, are very limited. In addition, existing models do not perform well in recognizing out-of-distribution entities and relations that are not seen in the training phase. Method: In this study, we present a novel and practical pipeline for constructing a heart failure knowledge graph using large language models and medical expert refinement. We apply prompt engineering to the three phases of schema design: schema design, information extraction, and knowledge completion. The best performance is achieved by designing task-specific prompt templates combined with the TwoStepChat approach. Results: Experiments on two datasets show that the TwoStepChat method outperforms the Vanillia prompt and outperforms the fine-tuned BERT-based baselines. Moreover, our method saves 65% of the time compared to manual annotation and is better suited to extract the out-of-distribution information in the real world.

Flexible Phenanthracene Nanotubes for Explosive Detection.

Phenanthracene nanotubes with arylene-ethynylene-butadiynylene rims and phenanthracene walls are synthesized in a modular bottom-up approach. One of the rims carries hexadecyloxy side chains, mediating the affinity to highly oriented pyrolytic graphite. Molecular dynamics simulations show that the nanotubes are much more flexible than their structural formulas suggest: In 12, the phenanthracene units act as hinges that flip the two macrocycles relative to each other to one of two possible sites, as quantum mechanical models suggest and scanning tunneling microscopy investigations prove. Unexpectedly, both theory and experiment show for 13 that the three phenanthracene hinges are deflected from the upright position, accompanied by a deformation of both macrocycles from their idealized sturdy macroporous geometry. This flexibility together with their affinity to carbon-rich substrates allows for an efficient host-guest chemistry at the solid/gas interface opening the potential for applications in single-walled carbon nanotube-based sensing, and the applicability to build new sensors for the detection of 2,4,6-trinitrotoluene via nitroaromatic markers is shown.

Scalable biomimetic sensing system with membrane receptor dual-monolayer probe and graphene transistor arrays.

Affinity-based biosensing can enable point-of-care diagnostics and continuous health monitoring, which commonly follows bottom-up approaches and is inherently constrained by bioprobes' intrinsic properties, batch-to-batch consistency, and stability in biofluids. We present a biomimetic top-down platform to circumvent such difficulties by combining a "dual-monolayer" biorecognition construct with graphene-based field-effect-transistor arrays. The construct adopts redesigned water-soluble membrane receptors as specific sensing units, positioned by two-dimensional crystalline S-layer proteins as dense antifouling linkers guiding their orientations. Hundreds of transistors provide statistical significance from transduced signals. System feasibility was demonstrated with rSbpA-ZZ/CXCR4QTY-Fc combination. Nature-like specific interactions were achieved toward CXCL12 ligand and HIV coat glycoprotein in physiologically relevant concentrations, without notable sensitivity loss in 100% human serum. The construct is regeneratable by acidic buffer, allowing device reuse and functional tuning. The modular and generalizable architecture behaves similarly to natural systems but gives electrical outputs, which enables fabrication of multiplex sensors with tailored receptor panels for designated diagnostic purposes.

Low-thermal-budget synthesis of monolayer molybdenum disulfide for silicon back-end-of-line integration on a 200 mm platform.

Two-dimensional (2D) materials are promising candidates for future electronics due to their excellent electrical and photonic properties. Although promising results on the wafer-scale synthesis (≤150 mm diameter) of monolayer molybdenum disulfide (MoS2) have already been reported, the high-quality synthesis of 2D materials on wafers of 200 mm or larger, which are typically used in commercial silicon foundries, remains difficult. The back-end-of-line (BEOL) integration of directly grown 2D materials on silicon complementary metal-oxide-semiconductor (CMOS) circuits is also unavailable due to the high thermal budget required, which far exceeds the limits of silicon BEOL integration (<400 °C). This high temperature forces the use of challenging transfer processes, which tend to introduce defects and contamination to both the 2D materials and the BEOL circuits. Here we report a low-thermal-budget synthesis method (growth temperature < 300 °C, growth time ≤ 60 min) for monolayer MoS2 films, which enables the 2D material to be synthesized at a temperature below the precursor decomposition temperature and grown directly on silicon CMOS circuits without requiring any transfer process. We designed a metal-organic chemical vapour deposition reactor to separate the low-temperature growth region from the high-temperature chalcogenide-precursor-decomposition region. We obtain monolayer MoS2 with electrical uniformity on 200 mm wafers, as well as a high material quality with an electron mobility of ~35.9 cm2 V-1 s-1. Finally, we demonstrate a silicon-CMOS-compatible BEOL fabrication process flow for MoS2 transistors; the performance of these silicon devices shows negligible degradation (current variation < 0.5%, threshold voltage shift < 20 mV). We believe that this is an important step towards monolithic 3D integration for future electronics.

Chemiresistive Hydrogen Sensing with Size-Limited Palladium Nanoparticles in Iptycene-Containing Poly(arylene ether)s.

Metal nanoparticles have been widely employed in chemical sensing due to their high reactivity toward various gases. The size of the metal nanoparticles often dictates their reactivity and hence their performance as chemiresistive sensors. Herein, we report that iptycene-containing poly(arylene ether)s (PAEs) have been shown to limit the growth of palladium nanoparticles (Pd NPs) and stabilize the Pd NPs dispersion. These porous PAEs also facilitate the efficient transport of analytes. Single-walled carbon nanotube (SWCNT)-based chemiresistors and graphene field-effect transistors (GFETs) using these PAE-supported small Pd NPs are sensitive, selective, and robust sensory materials for hydrogen gas under ambient conditions. Generalizable strategies including presorting SWCNTs with pentiptycene-containing poly(p-phenylene ethynylene)s (PPEs) and thermal annealing demonstrated significant improvements in the chemiresistive performance. The polymer:NP colloids produced in this study are readily synthesized and solution processable, and these methods are of general utility.

Integrated biosensor platform based on graphene transistor arrays for real-time high-accuracy ion sensing.

Two-dimensional materials such as graphene have shown great promise as biosensors, but suffer from large device-to-device variation due to non-uniform material synthesis and device fabrication technologies. Here, we develop a robust bioelectronic sensing platform  composed of  more than 200 integrated sensing units, custom-built high-speed readout electronics, and machine learning inference that overcomes these challenges to achieve rapid, portable, and reliable measurements. The platform demonstrates reconfigurable multi-ion electrolyte sensing capability and provides highly sensitive, reversible, and real-time response for potassium, sodium, and calcium ions in complex solutions despite variations in device performance. A calibration method leveraging the sensor redundancy and device-to-device variation is also proposed, while a machine learning model trained with multi-dimensional information collected through the multiplexed sensor array is used to enhance the sensing system's functionality and accuracy in ion classification.

Unraveling the Correlation between Raman and Photoluminescence in Monolayer MoS2 through Machine-Learning Models.

2D transition metal dichalcogenides (TMDCs) with intense and tunable photoluminescence (PL) have opened up new opportunities for optoelectronic and photonic applications such as light-emitting diodes, photodetectors, and single-photon emitters. Among the standard characterization tools for 2D materials, Raman spectroscopy stands out as a fast and non-destructive technique capable of probing material's crystallinity and perturbations such as doping and strain. However, a comprehensive understanding of the correlation between photoluminescence and Raman spectra in monolayer MoS2 remains elusive due to its highly nonlinear nature. Here, the connections between PL signatures and Raman modes are systematically explored, providing comprehensive insights into the physical mechanisms correlating PL and Raman features. This study's analysis further disentangles the strain and doping contributions from the Raman spectra through machine-learning models. First, a dense convolutional network (DenseNet) to predict PL maps by spatial Raman maps is deployed. Moreover, a gradient boosted trees model (XGBoost) with Shapley additive explanation (SHAP) to bridge the impact of individual Raman features in PL features is applied. Last, a support vector machine (SVM) to project PL features on Raman frequencies is adopted. This work may serve as a methodology for applying machine learning to characterizations of 2D materials.

SynCells: A 60 × 60 µm2 Electronic Platform with Remote Actuation for Sensing Applications in Constrained Environments.

Autonomous electronic microsystems smaller than the diameter of a human hair (<100 µm) are promising for sensing in confined spaces such as microfluidic channels or the human body. However, they are difficult to implement due to fabrication challenges and limited power budget. Here we present a 60 × 60 µm electronic microsystem platform, or SynCell, that overcomes these issues by leveraging the integration capabilities of two-dimensional material circuits and the low power consumption of passive germanium timers, memory-like chemical sensors, and magnetic pads. In a proof-of-concept experiment, we magnetically positioned SynCells in a microfluidic channel to detect putrescine. After we extracted them from the channel, we successfully read out the timer and sensor signal, the latter of which can be amplified by an onboard transistor circuit. The concepts developed here will be applicable to microsystems targeting a variety of applications from microfluidic sensing to biomedical research.

Wireless optoelectronic photometers for monitoring neuronal dynamics in the deep brain.

Capabilities for recording neural activity in behaving mammals have greatly expanded our understanding of brain function. Some of the most sophisticated approaches use light delivered by an implanted fiber-optic cable to optically excite genetically encoded calcium indicators and to record the resulting changes in fluorescence. Physical constraints induced by the cables and the bulk, size, and weight of the associated fixtures complicate studies on natural behaviors, including social interactions and movements in environments that include obstacles, housings, and other complex features. Here, we introduce a wireless, injectable fluorescence photometer that integrates a miniaturized light source and a photodetector on a flexible, needle-shaped polymer support, suitable for injection into the deep brain at sites of interest. The ultrathin geometry and compliant mechanics of these probes allow minimally invasive implantation and stable chronic operation. In vivo studies in freely moving animals demonstrate that this technology allows high-fidelity recording of calcium fluorescence in the deep brain, with measurement characteristics that match or exceed those associated with fiber photometry systems. The resulting capabilities in optical recordings of neuronal dynamics in untethered, freely moving animals have potential for widespread applications in neuroscience research.