Search

Abstract Targeted (nano-)drug delivery is essential for treating respiratory diseases, which are often confined to distinct lung regions. However, spatio-temporal profiling of drugs or nanoparticles (NPs) and their interactions with lung macrophages remains unresolved. Here, we present LungVis 1.0, an AI-powered imaging ecosystem that integrates light sheet fluorescence microscopy with deep learning-based image analysis pipelines to map NP deposition and dosage holistically and quantitatively across bronchial and alveolar (acinar) regions in murine lungs for widely-used bulk-liquid and aerosol-based delivery methods. We demonstrate that bulk-liquid delivery results in patchy NP distribution with elevated bronchial doses, whereas aerosols achieve uniform deposition reaching distal alveoli. Furthermore, we reveal that lung tissue-resident macrophages (TRMs) are dynamic, actively patrolling and redistributing NPs within alveoli, contesting the conventional paradigm of TRMs as static entities. LungVis 1.0 provides an advanced framework for exploring pulmonary delivery dynamics and deepening insights into TRM-mediated lung immunity.

Abstract The rapid emergence of large language model (LLM) technology presents promising opportunities to facilitate the development of synthetic reactions. In this work, we leveraged the power of GPT-4 to build an LLM-based reaction development framework (LLM-RDF) to handle fundamental tasks involved throughout the chemical synthesis development. LLM-RDF comprises six specialized LLM-based agents, including Literature Scouter, Experiment Designer, Hardware Executor, Spectrum Analyzer, Separation Instructor, and Result Interpreter, which are pre-prompted to accomplish the designated tasks. A web application with LLM-RDF as the backend was built to allow chemist users to interact with automated experimental platforms and analyze results via natural language, thus, eliminating the need for coding skills and ensuring accessibility for all chemists. We demonstrated the capabilities of LLM-RDF in guiding the end-to-end synthesis development process for the copper/TEMPO catalyzed aerobic alcohol oxidation to aldehyde reaction, including literature search and information extraction, substrate scope and condition screening, reaction kinetics study, reaction condition optimization, reaction scale-up and product purification. Furthermore, LLM-RDF’s broader applicability and versability was validated on various synthesis tasks of three distinct reactions (SNAr reaction, photoredox C-C cross-coupling reaction, and heterogeneous photoelectrochemical reaction).

Abstract Assessing fertilized human embryos is crucial for in vitro fertilization, a task being revolutionized by artificial intelligence. Existing models used for embryo quality assessment and ploidy detection could be significantly improved by effectively utilizing time-lapse imaging to identify critical developmental time points for maximizing prediction accuracy. Addressing this, we develop and compare various embryo ploidy status prediction models across distinct embryo development stages. We present BELA, a state-of-the-art ploidy prediction model that surpasses previous image- and video-based models without necessitating input from embryologists. BELA uses multitask learning to predict quality scores that are thereafter used to predict ploidy status. By achieving an area under the receiver operating characteristic curve of 0.76 for discriminating between euploidy and aneuploidy embryos on the Weill Cornell dataset, BELA matches the performance of models trained on embryologists’ manual scores. While not a replacement for preimplantation genetic testing for aneuploidy, BELA exemplifies how such models can streamline the embryo evaluation process.

Abstract Concentrating solar power plants are a clean energy source capable of competitive electricity generation even during night time, as well as the production of carbon-neutral fuels, offering a complementary role alongside photovoltaic plants. In these power plants, thousands of mirrors (heliostats) redirect sunlight onto a receiver, potentially generating temperatures exceeding 1000°C. Practically, such efficient temperatures are never attained. Several unknown, yet operationally crucial parameters, e.g., misalignment in sun-tracking and surface deformations can cause dangerous temperature spikes, necessitating high safety margins. For competitive levelized cost of energy and large-scale deployment, in-situ error measurements are an essential, yet unattained factor. To tackle this, we introduce a differentiable ray tracing machine learning approach that can derive the irradiance distribution of heliostats in a data-driven manner from a small number of calibration images already collected in most solar towers. By applying gradient-based optimization and a learning non-uniform rational B-spline heliostat model, our approach is able to determine sub-millimeter imperfections in a real-world setting and predict heliostat-specific irradiance profiles, exceeding the precision of the state-of-the-art and establishing full automatization. The new optimization pipeline enables concurrent training of physical and data-driven models, representing a pioneering effort in unifying both paradigms for concentrating solar power plants and can be a blueprint for other domains.

Abstract Curbing methane emissions is among the most effective actions that can be taken to slow down global warming. However, monitoring emissions remains challenging, as detection methods have a limited quantification completeness due to trade-offs that have to be made between coverage, resolution, and detection accuracy. Here we show that deep learning can overcome the trade-off in terms of spectral resolution that comes with multi-spectral satellite data, resulting in a methane detection tool with global coverage and high temporal and spatial resolution. We compare our detections with airborne methane measurement campaigns, which suggests that our method can detect methane point sources in Sentinel-2 data down to plumes of 0.01 km2, corresponding to 200 to 300 kg CH4 h−1 sources. Our model shows an order of magnitude improvement over the state-of-the-art, providing a significant step towards the automated, high resolution detection of methane emissions at a global scale, every few days.

Abstract Autonomous reaction network exploration algorithms offer a systematic approach to explore mechanisms of complex chemical processes. However, the resulting reaction networks are so vast that an exploration of all potentially accessible intermediates is computationally too demanding. This renders brute-force explorations unfeasible, while explorations with completely pre-defined intermediates or hard-wired chemical constraints, such as element-specific coordination numbers, are not flexible enough for complex chemical systems. Here, we introduce a STEERING WHEEL to guide an otherwise unbiased automated exploration. The STEERING WHEEL algorithm is intuitive, generally applicable, and enables one to focus on specific regions of an emerging network. It also allows for guiding automated data generation in the context of mechanism exploration, catalyst design, and other chemical optimization challenges. The algorithm is demonstrated for reaction mechanism elucidation of transition metal catalysts. We highlight how to explore catalytic cycles in a systematic and reproducible way. The exploration objectives are fully adjustable, allowing one to harness the STEERING WHEEL for both structure-specific (accurate) calculations as well as for broad high-throughput screening of possible reaction intermediates.

Abstract Millions of people require custom ocular prostheses due to eye loss or congenital defects. The current fully manual manufacturing processes used by highly skilled ocularists are time-consuming with varying quality. Additive manufacturing technology has the potential to simplify the manufacture of ocular prosthetics, but existing approaches just replace to various degrees craftsmanship by manual digital design and still require substantial expertise and time. Here we present an automatic digital end-to-end process for producing custom ocular prostheses that uses image data from an anterior segment optical coherence tomography device and considers both shape and appearance. Our approach uses a statistical shape model to predict, based on incomplete surface information of the eye socket, a best fitting prosthesis shape. We use a colour characterized image of the healthy fellow eye to determine and procedurally generate the prosthesis’s appearance that matches the fellow eye. The prosthesis is manufactured using a multi-material full-colour 3D printer and postprocessed to satisfy regulatory compliance. We demonstrate the effectiveness of our approach by presenting results for 10 clinic patients who received a 3D printed prosthesis. Compared to a current manual process, our approach requires five times less labour of the ocularist and produces reproducible output.

Targeted (nano-)drug delivery is essential for treating respiratory diseases, which are often confined to distinct lung regions. However, spatio-temporal profiling of drugs or nanoparticles (NPs) and their interactions with lung macrophages remains unresolved. Here, we present LungVis 1.0, an AI-powered imaging ecosystem that integrates light sheet fluorescence microscopy with deep learning-based image analysis pipelines to map NP deposition and dosage holistically and quantitatively across bronchial and alveolar (acinar) regions in murine lungs for widely-used bulk-liquid and aerosol-based delivery methods. We demonstrate that bulk-liquid delivery results in patchy NP distribution with elevated bronchial doses, whereas aerosols achieve uniform deposition reaching distal alveoli. Furthermore, we reveal that lung tissue-resident macrophages (TRMs) are dynamic, actively patrolling and redistributing NPs within alveoli, contesting the conventional paradigm of TRMs as static entities. LungVis 1.0 provides an advanced framework for exploring pulmonary delivery dynamics and deepening insights into TRM-mediated lung immunity., Competing Interests: Competing interests: The authors declare no competing interests., (© 2024. The Author(s).)

Abstract Image-based prediction models for disease detection are sensitive to changes in data acquisition such as the replacement of scanner hardware or updates to the image processing software. The resulting differences in image characteristics may lead to drifts in clinically relevant performance metrics which could cause harm in clinical decision making, even for models that generalise in terms of area under the receiver-operating characteristic curve. We propose Unsupervised Prediction Alignment, a generic automatic recalibration method that requires no ground truth annotations and only limited amounts of unlabelled example images from the shifted data distribution. We illustrate the effectiveness of the proposed method to detect and correct performance drift in mammography-based breast cancer screening and on publicly available histopathology data. We show that the proposed method can preserve the expected performance in terms of sensitivity/specificity under various realistic scenarios of image acquisition shift, thus offering an important safeguard for clinical deployment.

Abstract Structural superlubricity, a state of nearly zero friction and no wear between two contact surfaces under relative sliding, holds immense potential for research and application prospects in micro-electro-mechanical systems devices, mechanical engineering, and energy resources. A critical step towards the practical application of structural superlubricity is the mass transfer and high throughput performance evaluation. Limited by the yield rate of material preparation, existing automated systems, such as roll printing or massive stamping, are inadequate for this task. In this paper, a machine learning-assisted system is proposed to realize fully automated selective transfer and tribological performance measurement for structural superlubricity materials. Specifically, the system has a judgment accuracy of over 98% for the selection of micro-scale graphite flakes with structural superlubricity properties and complete the 100 graphite flakes assembly array to form various pre-designed patterns within 100 mins, which is 15 times faster than manual operation. Besides, the system is capable of automatically measuring the tribological performance of over 100 selected flakes on Si3N4, delivering statistical results for new interface which is beyond the reach of traditional methods. With its high accuracy, efficiency, and robustness, this machine learning-assisted system promotes the fundamental research and practical application of structural superlubricity.

Abstract Registration of data to a common frame of reference is an essential step in the analysis and integration of diverse neuroscientific data. To this end, volumetric brain atlases enable histological datasets to be spatially registered and analyzed, yet accurate registration remains expertise-dependent and slow. In order to address this limitation, we have trained a neural network, DeepSlice, to register mouse brain histological images to the Allen Brain Common Coordinate Framework, retaining registration accuracy while improving speed by >1000 fold.

As cryo-EM becomes commonplace in drug discovery, tools for automating small molecule structure determination are needed. Here, authors show a map-guided ligand modeling approach to building ligand structures at resolutions common in cryo-EM.

The authors present an intelligent metasurface system that uses a target detection algorithm combined with a depth camera, to automatically detect the position of moving targets and achieve real-time wireless communications. The system can operate for multiple targets in limited ambient light, outdoor and other realistic environments.

Time-dependent calculations are widely employed in simulating spectra. Here, the authors present a basis set truncation scheme to analyse and accelerate TDDFT calculations with negligible change in the resulting electronic spectra.

One challenge in cryo-EM is to build atomic models into intermediate resolution maps. Here, the authors present a deep learning-guided iterative assembling method by integrating AlphaFold, FFTbased fitting, and domain-based refinement.

Accurate delineation of individual teeth and alveolar bones from dental cone-beam CT images is an essential step in digital dentistry for precision dental healthcare. Here, the authors present a deep learning system for efficient, precise, and fully automatic segmentation of real-patient CBCT images presenting highly variable appearances.

High resolution descriptions of social interactions and their neural correlates are lacking. Here the authors report a pipeline enabling fully automatic multi-animal tracking during social encounters, together with simultaneous electrophysiological recordings, and show this works in low-light settings.

Systems for automatic detection of a single disease may miss other important conditions. Here, the authors show a deep learning platform can detect 39 common retinal diseases and conditions.

Existing high-performance deep learning methods typically rely on large training datasets with high-quality manual annotations, which are difficult to obtain in many clinical applications. Here, the authors introduce an open-source framework to handle imperfect training datasets.

Off-line neuro-evolution produces robot swarms whose good performance in simulation does not often transfer to the real word. With an extensive empirical study, Hasselmann et al. substantiate overfitting as the dominant cause.

The requirement for external electric supplies has significantly limited the application of electrochromic devices in modulating light absorption as smart windows. Here, the authors report automatic switching perovskite solar cells-powered all-in-one gel electrochromic device in response to surrounding light intensity in real-time.

It is difficult to synchronize the cell cycle in a population of yeast cells for extended periods of time. Here the authors use a cybergenetic system with inbuilt feedback to synchronize a population of modified yeast.

Patch clamp recording of neurons is slow and labor-intensive. Here the authors present a method for automated deep learning driven label-free image guided patch clamp physiology to perform measurements on hundreds of human and rodent neurons.

Efficient strategies to optimize metamaterial design for specific applications are urgently needed despite the rapid progress in this area. Here the authors propose a computational method combining an optimization algorithm with discrete element simulations for the automatic design of mechanical metamaterial actuators.

The role of automatic electrocardiogram (ECG) analysis in clinical practice is limited by the accuracy of existing models. In that context, the authors present a Deep Neural Network (DNN) that recognizes different abnormalities in ECG recordings which matches or outperform cardiology and emergency resident medical doctors.

Traces from single-molecule fluorescence microscopy (SMFM) experiments exhibit photophysical artifacts that typically make analysis time-consuming. Here, the authors have developed an easily accessible software, AutoSiM, for two distinct applications of deep learning to the efficient processing of SMFM time traces.

To optimize operating conditions of large scale semiconductor quantum devices, a large parameter space has to be explored. Here, the authors report a machine learning algorithm to navigate the entire parameter space of gate-defined quantum dot devices, showing about 180 times faster than a pure random search.

In this study, the authors investigate the predictability of sudden eruptions, motivated by the 2019 eruption at Whakaari (White Island), New Zealand. The paper proposes a machine learning approach that is able to identify eruption precursors in data streaming from a single seismic station at Whakaari.