Browse Articles
Discover research articles across all indexed journals
Deterministic and stochastic free vibration analysis of CNT reinforced functionally graded cantilever plates
Enhanced PCB defect detection via HSA-RTDETR on RT-DETR
n-Type thermoelectric elastomers
Research on optimal deep learning modeling in HaiNan dialect recognition
CRISPR/dCas9-TET1–mediated epigenetic editing reactivates miR-200c in breast cancer cells
Historic insights and future potential in wheat elaborated using a diverse cultivars collection and extended phenotyping
Abstract Wheat is one of the most important staple crops worldwide. Wheat breeding mainly focused on improving agronomy and techno-functionality for bread or pasta production, but nutrient content is becoming more important to fight malnutrition. We therefore investigated 282 bread wheat cultivars from seven decades of wheat breeding in Central Europe on 63 different traits related to agronomy, quality and nutrients in multiple field environments. Our results showed that wheat breeding has tremendously increased grain yield, resistance against diseases and lodging as well as baking quality across last decades. By contrast, mineral content slightly decreased without selection on it, probably due to its negative correlation with grain yield. The significant genetic variances determined for almost all traits show the potential for further improvement but significant negative correlations among grain yield and baking quality as well as grain yield and mineral content complicate their combined improvement. Thus, compromises in improvement of these traits are necessary to feed a growing global population.
Discovering cognitive strategies with tiny recurrent neural networks
Abstract Understanding how animals and humans learn from experience to make adaptive decisions is a fundamental goal of neuroscience and psychology. Normative modelling frameworks such as Bayesian inference 1 and reinforcement learning 2 provide valuable insights into the principles governing adaptive behaviour. However, the simplicity of these frameworks often limits their ability to capture realistic biological behaviour, leading to cycles of handcrafted adjustments that are prone to researcher subjectivity. Here we present a novel modelling approach that leverages recurrent neural networks to discover the cognitive algorithms governing biological decision-making. We show that neural networks with just one to four units often outperform classical cognitive models and match larger neural networks in predicting the choices of individual animals and humans, across six well-studied reward-learning tasks. Critically, we can interpret the trained networks using dynamical systems concepts, enabling a unified comparison of cognitive models and revealing detailed mechanisms underlying choice behaviour. Our approach also estimates the dimensionality of behaviour 3 and offers insights into algorithms learned by meta-reinforcement learning artificial intelligence agents. Overall, we present a systematic approach for discovering interpretable cognitive strategies in decision-making, offering insights into neural mechanisms and a foundation for studying healthy and dysfunctional cognition.
Optimizing emergency response services in urban areas through the fault-tolerant metric dimension of hexagonal nanosheet
Comparative analysis of deep learning architectures in solar power prediction
Abstract Integrating renewable energy sources into the electricity grid requires accurate forecasts of solar power production. With the aim of enhancing the accuracy and reliability of forecasts, this study presents a comprehensive comparative analysis of eight state-of-the-art Deep Learning (DL) architectures—Autoencoder, Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Simple Recurrent Neural Network (SimpleRNN), Convolutional Neural Network (CNN), Temporal Convolutional Network (TCN), Transformer, and Lightweight Informer for Long Sequence Time-Series Forecasting (InformerLite)—applied to solar power prediction using a dataset with 4,200 historical records and 20 meteorological and astronomical features. A comprehensive assessment of Root Mean Squared Error $$\:\left(\varvec{R}\varvec{M}\varvec{S}\varvec{E}\right)$$ , Mean Absolute Error $$\:\left(\varvec{M}\varvec{A}\varvec{E}\right)$$ , Mean Absolute Percentage Error $$\:\left(\varvec{M}\varvec{A}\varvec{P}\varvec{E}\right)$$ , and Coefficient of Determination $$\:\left({\varvec{R}}^{2}\right)$$ metrics was performed on the training, validation, and test datasets. The TCN model had the greatest performance across all models, achieving a test R² of 0.7786, an $$\:\varvec{R}\varvec{M}\varvec{S}\varvec{E}$$ of 429.4863, and a balanced relative standard deviation ( $$\:\varvec{R}\varvec{S}\varvec{D}$$ ) of 0.6827, so exhibiting an exceptional capacity to capture temporal patterns. The Autoencoder achieved a $$\:{\varvec{R}}^{2}$$ of 0.7648 and had the greatest overall performance on the entire dataset, resulting in a Whole $$\:{\varvec{R}}^{2}$$ of 0.8437. In contrast, the Transformer model demonstrated significantly poorer performance (Test $$\:{\varvec{R}}^{2}$$ = 0.0714), underscoring its limitations in this context without any architectural modifications. This study not only demonstrates the best DL models for solar power forecasting as qualified by useful statistical metrics, but also provides a scalable, interpretable, and extensible forecasting framework for real-world energy systems. The findings verify the informed DL integration to smart grid scenarios, laying the foundations for further developments in hybrid modeling, multi-horizon prediction, and deployment in resource-constrained environments with limited computational power and resources.
The geologic history of marine dissolved organic carbon from iron oxides
Abstract Dissolved organic carbon (DOC) is the largest reduced carbon reservoir in modern oceans1,2. Its dynamics regulate marine communities and atmospheric CO2 levels3,4, whereas 13C compositions track ecosystem structure and autotrophic metabolism5. However, the geologic history of marine DOC remains largely unconstrained6,7, limiting our ability to mechanistically reconstruct coupled ecological and biogeochemical evolution. Here we develop and validate a direct proxy for past DOC signatures using co-precipitated organic carbon in iron ooids. We apply this to 26 marine iron ooid-containing formations deposited over the past 1,650 million years to generate a data-based reconstruction of marine DOC signals since the Palaeoproterozoic. Our predicted DOC concentrations were near modern levels in the Palaeoproterozoic, then decreased by 90−99% in the Neoproterozoic before sharply rising in the Cambrian. We interpret these dynamics to reflect three distinct states. The occurrence of mostly small, single-celled organisms combined with severely hypoxic deep oceans, followed by larger, more complex organisms and little change in ocean oxygenation and finally continued organism growth and a transition to fully oxygenated oceans8,9. Furthermore, modern DOC is 13C-enriched relative to the Proterozoic, possibly because of changing autotrophic carbon-isotope fractionation driven by biological innovation. Our findings reflect connections between the carbon cycle, ocean oxygenation and the evolution of complex life.
Association between modified mediterranean diet score and menopause-specific quality of life and symptoms: a cross-sectional study
Microbially induced smectite to illite transformation in natural sediments during laboratory compression
ASL reveals regional brain perfusion impairment in neonates with mild hypoxic ischemic encephalopathy
Abstract There is a lack of neuroimaging data and effective biomarkers in infants with mild hypoxic ischaemic encephalopathy (HIE). Cerebral reperfusion injury has shown potential as marker of neurodevelopmental outcome in moderate and severe HIE. We examined cerebral perfusion by using arterial spin labelling (ASL) in infants with mild HIE and its associations with adverse outcomes. We also studied the presence of any potential regional sensitivity of cerebral blood flow (CBF) on the effects of HIE severity. This prospective cohort study included term and near-term neonates admitted for HIE across 3 neonatal intensive care units in Italy between October 2019–2022. Magnetic resonance imaging (MRI)-ASL was performed between 4 and 10 days after birth. Neurodevelopmental outcome was assessed at 24–28 months. Of the 94 infants included in the analysis, 74 neonates had mild [79%], 15 moderate [16%], 5 severe encephalopathy [5%]. Of the 71 neonates with mild HIE and neurodevelopmental outcome available, 15 (21%) showed mild disability. Basal ganglia CBF was the only region significantly associated with cognitive, motor and language Bayley scores (false-discovery-rate < 0.05). HIE severity had a regional dependent effect with involvement of Heschl, Rolandic operculum, limbic lobe, subcortical gray nuclei followed by frontal lobes. Basal ganglia CBF in infants with mild HIE was associated with adverse outcomes even without any MRI visible deep brain nuclei injury.
Cardiac myosin-binding protein-C levels are associated with severity and prognosis in stable coronary artery disease
Watch rappelling robots dive into a lava tube — for science
Breaking point: mechanical stress helps NINJ1 protein to rupture membranes
Gal4 drivers of the geosmin receptor Or56a exhibit ectopic expression in the labral sense organ of Drosophila
Abstract The fruit fly, Drosophila melanogaster, is a valuable model for studying the mechanisms of chemosensation. The odorant receptor Or56a has been shown to be narrowly tuned to geosmin—a chemical that flies use as a proxy for toxic molds and bacteria—and its activation drives olfactory avoidance behavior. Here, I find that existing Gal4 drivers using cloned promoter fragments of the Or56a gene drive unexpected expression in the labral sense organ (LSO), an internal taste sensory organ within the fly pharynx, in addition to their reported expression in the olfactory antennae. However, the presence of geosmin in sucrose solution does not elicit taste aversion or reduce consumption. Furthermore, a knock-in Or56a-T2A-Gal4 line newly generated in this study does not drive expression in the LSO. These results suggest that the LSO expression likely reflects ectopic expression from the existing Or56a-Gal4 drivers rather than the endogenous Or56a expression pattern. This study adds to the growing evidence that genetic drivers constructed using cloned promoters may not always faithfully recapitulate endogenous gene expression patterns, which should be taken into consideration when interpreting experimental results.
Audio signal analysis using a modified forward–forward algorithm with enhanced segmentation for soil pest detection
Abstract The presence of pests in soil costs the agriculture industry billions of dollars every year since it reduces crop yields and raises preventive costs. The pest detection in soil is vital for maintaining healthy crops, optimizing pest management, and ensuring economic and ecological sustainability. There are several invasive and non-invasive methods available for pest detection, where invasive methods are costly as well as time-consuming compared to the non-invasive methods. From various non-invasive methods, audio-based pest detection in the soil is one of the effective, low-cost tools. The generation of pest sounds is random in nature and contains a lot of inactive and background noisy portions in the recorded sound signals. To reduce the unnecessary computations in analyzing the inactive portions, an improved audio activity detection algorithm has been designed in this paper using Short Time Energy features for segmentation, which provides an average of 20% less computational requirements as compared to the baseline models. In the second step, the Forward Forward Algorithm has been used for its benefits in enhanced numerical stability, simplified computations, and enhanced precision over traditional back propagation-based algorithms. For improved performance in the detection of pests in soil, the traditional FF algorithm has been further updated by using root mean square in the goodness and loss function calculation. Through the comparative analysis with several baseline models, it has been observed that the proposed method consistently provides an average of 5% enhanced performance.