哈尔科夫大学

The intricate computer-aided analysis has been employed to investigate the different aspects of insulin cooperation with polyphenols, including quercetin, resveratrol, sesamin, gallic and salicylic acids. Molecular docking and molecular dynamics simulations revealed that quercetin, resveratrol and sesamin binding to the insulin via hotspot amino acid residues Phe24-Tyr26 is predicted to be associated with opening of protein structure and stabilization of Phe24, the steps which are considered critical for initiating the insulin binding to its receptor. Network pharmacology analysis coupled with bioactivity and toxicity predictions underscored the potential of quercetin, resveratrol and gallic acid to interact with pivotal insulin metabolic pathways, namely, AMPK and mTOR. The revealed propensity of quercetin and resveratrol to alter the structure of insulin, coupled with their predicted ability to interact with pivotal insulin metabolic pathways, suggests a possible dual mechanistic role of these polyphenols that may be of relevance in the context of insulin-centric impairments, and warrants further experimental validation.

For the first time, a statistical study has been conducted on the geomagnetic bay-like and quasi-periodic disturbances based on the datasets collected at 19 recording stations participating in INTERMAGNET Magnetic Observatories. In order to identify the disturbances from the volcanic explosion, a preliminary analysis has been used of the state of space weather during the catastrophic Tonga volcanic explosion of January 15, 2022. We summarize the main results as follows: The non-monotony of the variations in the strength of Y and Z geomagnetic field components increased appreciably on the day of the explosion as compared to the variations observed during the days used as a quiet time reference, while the eastward component of the geomagnetic field exhibited up to a 70 nT increase in variability. The duration and time delay of the bay-like disturbances increased with distance from the volcano, while their amplitude decreased. The propagation speeds of the bay-like disturbances at various observatories were determined to be in the 732–1325 m/s range. Four groups of time delays of quasi-periodic disturbances have been identified in a simultaneous analysis for the first time; they correspond to the apparent speeds of a few hundred km/s, 287–378 m/s, 212–233 m/s, and 130 m/s. The time delay in each group increased with distance away from the volcano. The agreement between theoretical estimates and the observational data testifies to the adequacy of the mechanism adopted for the generation of the disturbances.

The prediction of quantum dynamical responses lies at the heart of modern physics. Yet, modeling these time-dependent behaviors remains a formidable challenge because quantum systems evolve in high-dimensional Hilbert spaces, often rendering traditional numerical methods computationally prohibitive. While large language models have achieved remarkable success in sequential prediction, quantum dynamics presents a fundamentally different challenge: forecasting the entire temporal evolution of quantum systems rather than merely the next element in a sequence. Existing neural architectures such as recurrent and convolutional networks often require vast training datasets and suffer from spurious oscillations that compromise physical interpretability. In this work, we introduce an approach: Kolmogorov-Arnold networks (KANs) augmented with physics-informed loss functions that enforce the Ehrenfest theorems. Our method achieves superior accuracy with significantly less training data: It requires only 5.4% of the samples (200) compared to temporal convolution networks (3700). We further introduce the chain of KANs, an architecture that embeds temporal causality directly into the model design, making it particularly well suited for time series modeling. Our results demonstrate that physics-informed KANs offer a compelling advantage over conventional black-box models, maintaining both mathematical rigor and physical consistency while dramatically reducing data requirements.

Under the current global biodiversity crisis, there is a need for automated and noninvasive monitoring techniques that can gather large amounts of data cost-effectively at various ecological scales, from local to large spatial scales. These data can then be analyzed to inform stakeholders and decision-makers. One such technique is passive acoustic monitoring, which is commonly coupled with automatic identification of animal species based on their sound. Automated sound analyses usually require the training of sound detection and identification algorithms. These algorithms are based on annotated acoustic datasets which mark the occurrence of sounds of species inside sound recordings. However, compiling large annotated acoustic datasets is time-consuming and requires experts, and therefore, they normally cover reduced spatial, temporal, and taxonomic scales. This data paper presents WABAD, the World Annotated Bird Acoustic Dataset for passive acoustic monitoring. WABAD is designed to provide the public, the research community, and conservation managers with a novel and globally representative annotated acoustic dataset. This database includes 5047 min of audio files annotated to species-level by local experts with the start and end time and the upper and lower frequencies of each identified bird vocalization in the recordings. The database has a wide taxonomic and spatial coverage, including information on 91,931 vocalizations from 1192 bird species recorded at 72 recording sites in 29 recording locations (mainly countries) and distributed across 13 biomes. WABAD can be used, for example, for developing and/or validating automatic species detection algorithms, answering ecological questions, such as assessing geographical variations on bird vocalizations, or comparing acoustic diversity indices with species-based diversity indices. The dataset is published under a Creative Commons Attribution 4.0 International license that permits redistribution and reuse on the condition that the original work is properly credited.

Alzheimer’s disease (AD) is a progressive neurodegenerative disease and the current diagnosis tools, which are used, including clinical examination and neuroimaging, are costly, time consuming and complicated to generalize across demographics. Traditional machine learning (ML) methods such as Support Vector Machines (SVM), LightGBM, and Random Forest (RF) are known to be having weakness in terms of feature selection, class imbalance, and generalization. This paper presents a new Enhanced TabTransformer with Self-Supervised Learning (ETT-SSL) framework of AD classification. The structure combines the transformer-based architecture, self-supervised learning (SSL) for feature representation, SHAP-based feature selection, and the use of SMOTE to balance classes. Experimental results also indicate that ETT-SSL reaches a high accuracy of 95.8%, which is a much better result than baseline models (SVM: 72.1, RF: 78.3, LightGBM: 80.5) and even the standard TabTransformer (85.2%). Additionally, ETT-SSL offers greater accuracy and improved recall, overcoming the issue of false negatives in AD diagnosis. SHAP analysis can be used to improve transparency in clinical decision-making by making use of the model interpretability. The proposed framework recommends a salivary, interpretability and clinically practical methodology that can be used to expand to multimodal data source (e.g. MRI scans, genomic markers and electronic health records) to improve accuracy and generalisability.