Browse Articles

K-M LLM-pro: Physics-guided cross-modal adaptation for fine-grained spatiotemporal trajectory classification

Spatiotemporal trajectory classification is essential for intelligent perception systems but faces challenges including weak separability of dynamic features, representation collapse under limited samples, and heterogeneous conflicts in multimodal data. To address these issues, we propose K-M LLM-pro, a physics-guided cross-modal adaptation framework that integrates statistical mechanics with large language models (LLMs) to improve trajectory understanding. Our approach incorporates: (1) physics-informed prompt engineering based on Kramers-Moyal coefficients, embedding physical constraints via reproducing kernel Hilbert space projection; (2) a dynamic patching optimization mechanism combining variance maximization and Lyapunov stability criteria for unified modeling of heterogeneous trajectories; and (3) dual spatiotemporal adapters with a parameter-efficient expansion strategy, injecting domain knowledge while optimizing only 3.8% of new parameters. Experimental results on public datasets such as Geolife and AIS show that K-M LLM-pro outperforms state-of-the-art models in classification accuracy, demonstrating strong performance even in few-shot scenarios with only 1% of training data. To our knowledge, this is the first work to integrate K-M coefficients as interpretable statistical priors into LLMs, offering a lightweight and effective solution for modeling complex spatiotemporal dynamics.

Simulation of the fruit and vegetable intakes meeting the dietary reference intakes of Japanese adults from the National Health and Nutrition Survey

The genome of Gallaecimonas pentaromativorans strain 10A, isolated from a Pacific oyster, sheds light on an environmentally widespread genus with remarkable metabolic potential

Bacteria in the genus Gallaecimonas are known for their ability to breakdown complex hydrocarbons, making them of particular ecological and biotechnological significance. However, few species have been isolated to date, and their ecological distribution has yet to be examined. Here, we report a novel strain of G. pentaromativorans , designated as strain 10A, which was isolated from a Pacific oyster ( Magallana gigas , a.k.a. Crassostrea gigas ) collected from a farm experiencing a mass mortality event in British Columbia (BC), Canada. Gallaecimonas pentaromativorans strain 10A is a rod-shaped, motile bacterium and has a circular genome of 4,322,156 bp encoding 3,928 protein-coding sequences (CDS). Phylogenetic analysis showed that strain 10A is closely related to members of G. pentaromativorans. Like other Gallaecimonas members, strain 10A is predicted to harbor specific pathways involved in degrading xenobiotic compounds including polycyclic aromatic hydrocarbons (PAHs), producing biosurfactants, and assimilating nitrate and sulfate; however, it is uniquely equipped with an additional 166 genes belonging to 147 protein families, including a putative higB - higA that likely contributes to enhanced stress response. Strain 10A also possesses Clustered Regularly Interspaced Short Palindromic Repeat (CRISPR) and CRISPR-associated (Cas) system (CRISPR-Cas), prevalent in Gallaecimonas (detected in three out of four species), implying a potential defense mechanism against exogenous mobile genetic elements such as plasmids and viruses. We also mined publicly available databases to establish the widespread distribution of bacteria in the genus Gallaecimonas in seawater, sediments, and freshwater across latitude, suggesting its versatility and importance to environmental processes. Ultimately, this study demonstrates that the genome of G. pentaromativorans strain 10A, isolated from a Pacific oyster, may encode a suite of putative functions, including xenobiotic breakdown, biosurfactant production, and CRISPR-Cas defense. This plasticity and breadth in metabolic function help to explain the cosmopolitan distribution of members of this genus.

Modeling the rise and demise of Classic Maya cities: Climate, conflict, and economies of scale

Urbanization was one of the most significant transitions in human history, yet explanations for the rise and expansion of early cities remain contentious. Here, we propose that simple models from population ecology can integrate existing theories for the development of early cities. Using newly synthesized paleoclimatological, paleoecological, demographic, and historical data from across the Classic Period Maya Lowlands (250–1000 CE) integrated with piece-wise structural equation models, we show that climate downturns, intergroup conflict, and strong economies of scale interact to promote the coevolution of urbanism and patron–client relationships, fueling city expansion, urban institutions, and systemic inequality. In addition, we elucidate how these nonlinear pathways structure the persistence or dissolution of cities. This study underscores the importance of robust economies of scale in the development of early cities and provides a comprehensive framework for understanding the conditions that promote or hinder urbanization, offering insights applicable to both ancient and contemporary urban dynamics.

Molecules interacting with CasL-Like 2 enhances tumor angiogenesis and progression by activating mTOR/HIF1α/VEGF pathway in kidney renal clear cell carcinoma

Machine learning detects hidden treatment response patterns only in the presence of comprehensive clinical phenotyping

Inferential statistics traditionally used in clinical trials can miss relationships between clinical phenotypes and treatment responses. We simulated a randomised clinical trial to explore how gradient boosting (XGBoost) machine learning compares with traditional analysis when ‘ground truth’ treatment responsiveness depends on the interaction of multiple phenotypic variables. As expected, traditional analysis detected a significant treatment benefit (outcome measure change from baseline = 4.23; 95% CI 3.64–4.82). However, recommending treatment based upon this evidence would lead to 56.3% of patients failing to respond. In contrast, machine learning correctly predicted treatment response in 97.8% (95% CI 96.6–99.1) of patients, with model interrogation showing the critical phenotypic variables and the values determining treatment response had been identified. Importantly, when a single variable was omitted, accuracy dropped to 69.4% (95% CI 65.3–73.4). This proof of principle underscores the significant potential of machine learning to maximise the insights derived from clinical research studies. However, the effectiveness of machine learning in this context is highly dependent on the comprehensive capture of phenotypic data.

Integrated molecular and ADME-toxicity profiling identifies PGV-5 and HGV-5 as potential agents to counteract multidrug-resistant (MDR) cancer

Abstract Curcumin, a pharmacological agent found in turmeric’s rhizome, has been studied for its various therapeutic properties. However, its clinical development is hindered by its instability and low solubility in water, resulting in inadequate oral bioavailability. Two potential curcumin analogs, 2,5-bis(4’-hydroxy-3’,5’-dimethoxybenylidene)cyclopentanone (PGV-5) and 2,6-bis(4’-hydroxy-3’,5’-dimethoxybenylidene)cyclohexanone (HGV-5), are being developed to address this issue and enhance their therapeutic efficacy. The study aims to screen novel curcumin analog compounds by integrating in silico assessment of ADME properties, acute toxicity studies, and computational analysis. PGV-5 and HGV-5 are classified as Global Harmonized System of Classification and Labeling of Chemicals (GHS) class 4 and class 5, respectively, in acute toxicity assessment, as they cause histopathological changes in the heart and lungs. Their ADME profile indicates they serve as effective P-glycoprotein (P-gp) inhibitors, making them potential candidates for development as anti-multidrug resistance agents, particularly in cancer cells. Molecular docking on P-gp revealed significant inhibitory capability relative to curcumin, exhibiting comparable binding characteristics to the native ligand, as evidenced by superior docking scores. Subsequent molecular dynamics simulations confirmed the stable interaction of both compounds with P-gp, with HGV-5 showing the most favorable binding free energy. Target gene mapping revealed several pivotal targets including AKT1, STAT3, EGFR, and NF-κB1. These findings suggest that PGV-5 and HGV-5 merit further research as agents against multidrug-resistant in cancer, regardless of their toxicity profiles. Further confirmation of their effects requires more laboratory studies and clinical trials.

Enhancement of solar cell efficiency through tailored electrodeposited seed layers and CdS: O surface texturing

Malaria knowledge and preventive practices among caregivers of under-five children in Southwest Nigeria

Understanding individual neurodegenerative progression in Parkinson’s disease through normative modelling

Abstract Parkinson’s disease (PD) is a neurodegenerative disorder with motor symptoms (e.g., bradykinesia, tremors) and non-motor symptoms (e.g., cognitive deficits). Symptom progression varies across individuals, possibly due to differences in the spread of disease pathology. This study investigates individual-level gray matter atrophy in PD patients compared to a reference cohort, modeling neurobiological trajectories to understand symptom progression. Using normative modeling, we mapped individual deviations in gray matter atrophy in PD patients (Personalized Parkinson Project, PPP; N = 408; 42% female) against a reference model (N = 58, 836) of non-diagnosed individuals. Gray matter atrophy was defined as negative deviations from the normative model in cortical thickness and subcortical volume at baseline and two-year follow-up. We correlated the deviations with clinical motor and cognitive symptoms at an individual level and compared changes across PD subtypes (mild-motor predominant, intermediate, and diffuse-malignant). Cross-sectionally, PD patients showed significant gray matter atrophy, which correlated with cognitive impairment. Longitudinally, cortical thinning and subcortical atrophy patterns showed variation amongst subtypes. Specifically, the diffuse-malignant subtype, which is characterized by more diffuse symptoms and faster clinical progression, exhibited pronounced cortical thinning and subcortical atrophy over time. In this paper, we have considered the deviation scores at three levels of granularity: cases vs. control, subtypes, and the individual level. While our findings show subgroup-level patterns of variability, they also provide a method for exploring individual-level metrics of disease progression, acknowledging that individuals may deviate from the predefined categories or groups and can exhibit large variability over time.

Prediction of the short-term prognosis of acute ischaemic stroke in patients with high treatment platelet reactivity using explainable machine learning

Methylated 1,2-naphthoquinone derivative SJ006 as an inhibitor of human glucose 6-phosphate dehydrogenase in non-small cell lung cancer cell lines

Abstract Glucose 6-phosphate dehydrogenase (G6PD) is crucial for redox balance and biosynthesis via the pentose phosphate pathway (PPP), driving non-small cell lung cancer (NSCLC) proliferation. This study assessed the cytotoxic and enzymatic effects of five 1,2-naphthoquinone (NQ) derivatives, including SJ006 derived from Usnea barbata , in NSCLC cell lines (A549 and NCI-H292) compared to traditional inhibitors (DHEA and 6AN). All 1,2-NQs demonstrated concentration-dependent cytotoxicity against NSCLC cells. Among them, NN02 exhibited the highest cytotoxicity comparable to 6-AN, followed by NN01, NN04, SJ006, and SJ007, which showed moderate effects comparable to DHEA. SJ006 uniquely inhibited G6PD activity without altering its mRNA or protein expression. Unlike DHEA and 6AN, SJ006 functioned as an uncompetitive inhibitor, decreasing both K m and V max , with molecular docking confirming strong G6PD binding. Additionally, SJ006 increased reactive oxygen species (ROS) levels, induced G2/M cell cycle arrest, and triggered late apoptosis in NSCLC cells. Its effects were reversed by D-(−)-ribose, confirming PPP disruption as the mechanism. These results highlight SJ006 as a novel G6PD inhibitor that disrupts redox homeostasis and biosynthesis-driven cell proliferation, showing promise as an anticancer agent for NSCLC.

Incidence and predictors of mortality among persons with rifampicin-resistant tuberculosis and HIV in Mozambique

Triangular reentrant honeycomb metamaterial structure for broadband sound attenuation using shape memory polymers

Implication of dopamine transporter and electroencephalography biomarkers in dementia with lewy bodies

Isometric representations in neural networks improve robustness

Abstract Artificial and biological agents are unable to learn given completely random and unstructured data. The structure of data is encoded in the distance or similarity relationships between data points. In the context of neural networks, the neuronal activity within a layer forms a representation reflecting the transformation that the layer implements on its inputs. In order to utilize the structure in the data in a truthful manner, such representations should reflect the input distances and thus be continuous and isometric. Supporting this statement, findings in neuroscience propose that generalization and robustness are tied to neural representations being continuously differentiable. Furthermore, representations of objects have the capacity of being hierarchical. Combined together, these two conditions imply that neural networks need to both preserve the distances between inputs as well as have the capacity to apply cuts at different resolutions, corresponding to different levels of a hierarchy. During cross-entropy classification, the metric and structural properties of network representations are usually broken both between and within classes. To achieve and study this behavior, we train neural networks to perform classification while simultaneously maintaining the metric structure within each class at potentially different levels of a hierarchy, leading to continuous and isometric within-class representations. We show that such network representations turn out to be a beneficial component for making accurate and robust inferences about the world. We come up with a network architecture that facilitates hierarchical manipulation of internal neural representations. We verify that our isometric regularization term improves the robustness to adversarial attacks on MNIST and CIFAR10. Finally, we use toy datasets and show that the learned map is isometric everywhere, except around decision boundaries.

Immunological phenotype in asthma and its impact on long-term renal outcomes

Abstract Asthma is associated with both airway and systemic inflammation as well as non-respiratory adverse outcomes. However, data regarding its impact on long-term renal outcomes is lacking. We classified all asthma patients who were followed at Queen Mary Hospital in 2017 into eosinophilic or non-eosinophilic phenotypes based on their highest blood eosinophil counts (BEC) during stable state in the year (≥ 300 or < 300 cells/mm3 respectively) and prospectively evaluated their clinical outcomes in the subsequent 5 years. The relationship between patient phenotypes and the long-term renal outcomes were assessed. Five hundred and four asthma patients with baseline Stage 1 to 3 chronic kidney disease were included [296 (58.7%) and 208 (41.3%) in eosinophilic and non-eosinophilic groups respectively]. Among patients with baseline renal function at CKD stage 1 to 3, one hundred and four patients (20.6%) had renal progression in this cohort (56 patients (26.9%) vs. 48 patients (16.2%) in the non-eosinophilic and eosinophilic groups respectively). Patients with non-eosinophilic asthma showed increased risks of renal progression over 5 years of follow-up [adjusted odds ratio (aOR) 2.615, 95% CI 1.151–5.942 p = 0.022] and more rapid eGFR decline (−4.29 ± 3.48 mL/min/1.73m2/year vs. −3.48 ± 3.07 mL/min/1.73m2/year, p = 0.007) than those with eosinophilic phenotype. Patients who developed renal progression had higher risk of death [adjusted hazard ratio (aHR) 1.614 (95% CI 1.041–2.502); p = 0.032]. Progressive renal function deterioration is prevalent amongst asthma patients, and those with non-eosinophilic phenotype are at risk of renal progression.

Selection of heat stress tolerant wheat genotypes for desert environments

Abstract High temperature is a critical abiotic stress that severely impacts agricultural productivity, especially in semi-arid and arid regions. This study assesses the phenotypic performance and genetic diversity of twenty advanced wheat genotypes and checks under the field conditions of heat stress for two years. Heat stress led to significant reductions in grain yield and related traits, with an average yield decline of 53.8%. Path analysis revealed a negative impact of heading date on grain yield under stress conditions. Stress indices indicated strong heat tolerance in the genotype YR × Ksu110-240, which showed only a 10.4% reduction in grain yield, whereas DHH3-26 exhibited high sensitivity with a 53.6% reduction. Genetic diversity analysis using 30 Simple Sequence Repeat (SSR) polymorphic markers identified significant marker-trait associations, particularly Xgwm 285 and Xgwm 577, which were strongly linked to heat tolerance related traits. These markers provide valuable tools for marker-assisted selection (MAS), facilitating the breeding of heat-resilient wheat varieties. This study highlights the significance of combining molecular markers with phenotypic assessments to improve wheat adaptation to challenging environmental conditions. The wide genetic diversity offers opportunities for introducing novel alleles into breeding programs, which will be critical for developing wheat varieties that can sustain productivity under increasingly variable and extreme environmental conditions. By using the genetic and phenotypic diversities, breeders can target specific traits and markers to develop heat-resilient wheat varieties, ensuring food security in regions threatened by rising global temperatures.

Material properties of recycled PET composites for structural anchoring and other civil engineering applications

Phylogenomics redefines the evolutionary history of mosquitoes

Mosquitoes have a substantial impact on human and animal health, but their deeper evolutionary relationships have been difficult to resolve. We inferred a time-calibrated phylogenetic history of mosquitoes using conserved genome-wide markers from representatives of major lineages. Our analyses revealed that codon bias and positive selection in subfamily Anophelinae contributed to a substantial level of branch attraction bias between Anophelinae and outgroup taxa, which in our view has misled previous phylogenetic analyses of mosquitoes. Accounting for this systematic phylogenetic bias led to a revised view of mosquito evolution, including the nonmonophyly of subfamily Culicinae. Similarly, we dated the origin of mosquitoes to the mid-Cretaceous (~106 Mya) and most extant genera to after the KPg boundary <66 Mya, 100 My younger than previous estimates and coincident with the origin of Plasmodium parasites. Our study provides a foundation for future analyses of the evolution of mosquito-borne disease.