Fine tuning the pore structure of Aluminum-based metal-organic framework via mixed-linker strategy for enhancing ethane/ethylene separation

Structural basis and functional mining of active site loop for fine tuning substrate affinity of L-asparaginase from Bacillus licheniformis.

Self-supervised learning (SSL) has demonstrated its potential to reduce reliance on labeled data and improve generalization to independent test sets compared to traditional supervised learning (SL). However, current SSL paradigms are predominantly designed as domain-agnostic and are often evaluated on benchmark datasets like ImageNet. As already shown in other studies, SSL paradigms tailored for the medical domain can outperform domain-agnostic SSL paradigms. This study introduces NeuroBooster, a paradigm tailored for brain MRI analysis. Its core idea is to leverage automated feature extraction tools to generate anatomical features, which are then used as targets in a regression pretext task. The paradigms were evaluated and compared on a regression task, i.e., age prediction, and a classification task, i.e., the discrimination between patients with Alzheimer's disease and cognitively normal subjects. Each pre-training and fine tuning experiment was repeated across 30 randomized seeds to estimate the uncertainty, while enabling full reproducibility. Our results revealed NeuroBooster's superiority across various scenarios. Notably, with only 1% of labeled data available, NeuroBooster achieved an average mean absolute error of 5.67 years smaller than SL. These findings demonstrate NeuroBooster's efficacy and highlight its potential within brain MRI analysis, suggesting a promising direction for developing specialized SSL paradigms in this domain.

SNF1/AMPK controls its own localization by phosphorylating its activating kinase Sak1

Unimolecular multiagonists integrating GLP-1R/GIPR, and increasingly GCGR agonism, have altered the landscape of peptide therapeutics for metabolic syndrome. This progress has shifted the field from optimizing individual ligands to engineering defined receptor selectivity (balance) and, in some cases, pathway-selective signaling (bias) within a single-peptide scaffold. Despite this momentum, tuning 'balance' and 'bias' remains difficult because small sequence changes can have receptor-dependent, nonintuitive effects, and meaningful retuning often requires wholesale scaffold redesign. Here, we implement a systematic strategy based on simple N-terminal chemical modifications across dual- and triagonist templates to decouple these variables. Because the peptide N-terminus is buried within the membrane-embedded transmembrane core of class B GPCRs, localized chemical edits at this site provide a sensitive lever for reweighting receptor activation. These minimal changes can alter signaling preference without loss of efficacy, expanding the design grammar and offer rapid tuning of potency, efficacy, and bias.

Large language models (LLMs) show growing potential for decision support. However, integrating domain-specific medical knowledge while maintaining accuracy, safety, and interpretability remains challenging for postoperative discharge instructions and patient education. Fine-tuning (FT), retrieval-augmented generation (RAG), and hybrid FT+RAG approaches are prominent strategies for knowledge integration, but their comparative performance in postoperative care has not been systematically evaluated. We aimed to compare the performance, reliability, and safety characteristics of baseline, FT, RAG, and hybrid FT+RAG LLM configurations for postoperative decision support. We conducted a comparative evaluation of four LLM configurations using Google Gemini 2.5 Flash. A total of 600 postoperative question-answer pairs were used for model adaptation and validation, while 150 queries were reserved for final evaluation. Queries included routine postoperative questions, emergency escalation scenarios, and deliberately out-of-scope prompts. Outputs were independently assessed by 3 blinded clinical experts for clinical medical accuracy, safety/refusal accuracy, completeness, and relevance. Automated metrics evaluated readability, faithfulness, and hallucination propensity. All knowledge-enhanced models significantly outperformed baseline in overall accuracy (baseline 68.0% vs FT 92.7%, RAG 91.3%, FT+RAG 97.3%; P<.001). For in-scope clinical queries, FT+RAG achieved the highest clinical medical accuracy (96.7%) and was the only configuration to significantly outperform baseline in pairwise comparisons. Enhanced models also demonstrated higher safety/refusal accuracy than baseline; however, the baseline configuration did not receive equivalent safety or deferral instructions, which likely influenced these findings. FT+RAG achieved the strongest composite classification performance, including 100% precision, 96.7% recall, and 98.3% F1 score. FT and RAG showed broadly comparable performance across most secondary outcomes. Although knowledge-enhanced models demonstrated lower readability than baseline, restricted analysis of 100 routine in-scope postoperative queries suggested that part of this difference was attributable to standardized safety boilerplate. Incorporating domain-specific knowledge through FT, RAG, or both improved postoperative decision-support performance compared with the baseline LLM. All knowledge-enhanced approaches demonstrated strong performance, with the hybrid FT+RAG configuration achieving the most favorable overall point estimates across several outcomes. However, differences among the enhanced configurations were generally modest and less evident in sensitivity analyses restricted to unanimously rated queries. These findings support knowledge-enhanced LLMs as promising tools for postoperative education and decision support, while highlighting the need for further validation, readability optimization, transparent governance, and sustained human oversight before patient-facing deployment.

We study the prediction of T-cell response for specific given peptides, which could, among other applications, be a crucial step towards the development of personalized cancer vaccines. It is a challenging task due to limited, heterogeneous training data featuring a multi-domain structure; such data entail the danger of shortcut learning, where models learn general characteristics of peptide sources, such as the source organism, rather than specific peptide characteristics associated with T-cell response. Using a transformer model for T-cell response prediction, we show that the danger of inflated predictive performance is not merely theoretical but occurs in practice. Consequently, we propose a domain-aware evaluation scheme. We then study different transfer learning techniques to deal with the multi-domain structure and shortcut learning. We demonstrate a per-source fine tuning approach to be effective across a wide range of peptide sources and further show that our final model is competitive with existing state-of-the-art approaches for predicting T-cell responses for human peptides.

ABSTRACT The structure of a cluster plays a decisive role in determining its physical and chemical properties. However, as cluster size increases, the number of possible isomers grows exponentially, and first-principles calculations become computationally demanding, posing significant challenges for structure prediction. To address this challenge, we examined an efficient method for searching low-energy structures of atomic clusters integrating machine learning interatomic potentials (MLIPs) and a Comprehensive Genetic Algorithm (CGA). We constructed training datasets for Cu, Ag, and Au clusters and trained the Orb-v2 potential using two strategies: fine-tuning a pre-trained model and training from scratch. Our benchmarking results demonstrate that the fine-tuned model achieves significantly lower energy prediction mean absolute errors (5 ∼ 7 meV/atom) compared to the pre-trained model (36 ∼ 368 meV/atom). Compared to training from scratch, fine tuning a pre-trained model requires 50% less epochs to train and demonstrates better accuracy. Integrating the fine-tuned MLIP with CGA enables efficient and effective searching for low-energy structures of atomic clusters, successfully reproducing known low-energy configurations of small clusters with less than 55 atoms.

The radical SAM enzyme Elp3 and eukaryotic Elongator complex catalyze formation of a key intermediate transfer RNA (tRNA) modification, 5-carboxymethyluridine (cm5U), in the anticodons of tRNAs across all domains of life. cm5U-derived modifications are important for fine tuning codon-anticodon interactions and efficient protein translation, and defects in this modification are linked to development of neurodegenerative disease in humans. Here we reconstitute tRNA modification activity with a model Elp3 enzyme and combine structural analyses, enzymology, and isotope incorporation experiments to show Elp3 harbors a conserved molecular tunnel that shuttles free acetate molecules from the acetyl-CoA binding domain to the radical SAM active site over 20 Å away, where acetate undergoes radical-mediated reaction and addition to tRNA U34. Our model explains how Elp3 bridges a large distance between active sites to catalyze tRNA carboxymethylation and illustrates a unique mechanism for intermediate transport in radical SAM enzymes.

The N1 component of event-related potentials (ERP) reflects print tuning and lexicality effects. Previous studies have shown that for a second learned writing system, tuning and lexical effects are associated with age and ability. However, the developmental trajectory of these tuning and lexicality effects, and how language skills influence them, remains unclear. The present study investigated how English reading abilities contribute to longitudinal changes in N1 amplitude and print tuning among Chinese children in Hong Kong. Forty-three children performed a repetition detection task while EEG was recorded to examine three types of tuning: coarse tuning (real word versus false font symbol), fine tuning (real word versus nonword), and lexicality effect (real word versus pseudoword). Children's English word reading accuracy (EWR) was assessed. Results revealed significant coarse tuning and lexicality effects but no significant fine tuning effect. Only coarse tuning showed longitudinal change, with the coarse tuning effect decreasing over time. Furthermore, the N1 coarse tuning effect increased with EWR in the first assessment but decreased with EWR in the second assessment. This modulation was not found for fine tuning or lexicality effects. These findings support the visual perceptual expertise account in the second language (L2) context, demonstrating that the coarse tuning effect, but not fine tuning and lexicality effects, undergoes developmental changes that are modulated by reading skills.

ABSTRACTObjectiveWe investigated the role of GLI3, a transcription factor highly expressed in the pathogenic THY1+CD34− sublining subset of rheumatoid arthritis synovial fibroblasts (RASFs), in regulating their pathogenic behavior.MethodsGLI3 protein levels were quantified in freshly isolated RASF subsets by Western blotting. Bulk RASFs were subjected to siRNA‐mediated knockdown (KD) of GLI3 or GLI1, followed by RNA sequencing. The effects of GANT61 on RASF proliferation, cell‐cycle progression, migration, viability, and apoptosis were assessed using EdU/PI analysis, scratch assays, CCK‐8 assays, and Annexin V/PI flow cytometry.ResultsGLI3 expression was enriched in THY1+CD34− RASFs at both mRNA and protein levels. GLI3 KD increased GLI1 expression and upregulated genes involved in inflammation, matrix remodeling, and cell cycle regulation, including IL6, IL11, IL24, IL33, MMP3, PLAU, CCNA2, and E2F1. Pathway enrichment analysis revealed activation of ECM–receptor interaction, PI3K–Akt, and TNF signaling. Co‐silencing GLI1 with GLI3 blunted the induction of IL11, IL24, IL33, CCNA2, E2F1, and PLAU observed with GLI3 KD alone, indicating that GLI1 mediates a subset of the transcriptional effects induced by GLI3 loss. GANT61 suppressed CCNA2 and E2F1 expression, inhibited RASF proliferation and migration, and did not markedly increase apoptosis.ConclusionGLI3 functions as a negative regulator of GLI1 and its downstream targets that drive the pathogenic behavior of RASFs. Targeting the GLI1–GLI3 axis may represent a promising therapeutic strategy to modulate fibroblast‐driven inflammation and joint destruction in RA.

Background/Objectives: Disorganization of retinal inner layers (DRIL) is an important and supportive biomarker in optical coherence tomography (OCT) imaging for diagnosing the extent of diabetic macular edema (DME) in patients and anticipating visual outcomes. But the manual DRIL identification is subject to interobserver bias and requires a lot of time and effort from the experts. This research presents a novel, computerized, and clinically guided approach for the classification of DRIL that leverages the central 1 mm foveal region extracted through the annotations provided by the expert ophthalmologists and investigates the effectiveness of a transformer and Masked Auto Encoder (MAE) based foundation model (RETFound) as the primary approach. Methods: We fine-tuned and validated the RETFound model, utilizing accurate foveal center coordinates provided by the experienced ophthalmologists. Our approach emphasizes the macular region that is significant diagnostically, where DME biomarkers manifest more predominantly. To guarantee robust evaluation, the dataset was divided into 85% training and 15% held-out test sets. We performed 5-fold cross-validation exclusively on the training dataset with baseline, conservative, and moderate fine-tuning strategies, and the final model was evaluated on the independent, unseen test set. Convolutional neural network (CNN)-based transfer learning (TL) models (MobileNetV2, EfficientNetB0, InceptionV3, DenseNet121, and DenseNet169) were also assessed for comparative evaluation. Results: The RETFound model yielded the best outcomes under the conservative fine-tuning strategy, achieving a mean test accuracy (AC) of 0.9339 ± 0.0036 and an area under the curve (AUC) of 0.9660 ± 0.0028 on the independent held-out test set across the five fold-trained models. The moderate and baseline evaluations achieved comparatively lower outcomes, highlighting the effectiveness of the conservative approach. The RETFound model consistently outperformed CNN models, exhibiting stability and superior generalization for DRIL classification. We performed statistical validation using the Wilcoxon signed-rank test and 95% confidence intervals to confirm the robustness of the proposed method, and an ablation analysis showed that the fovea-centered region of interest (ROI) guidance consistently improved results when compared with whole OCT analysis. Conclusions: This research demonstrates that the deep-learning (DL) methods assisted by expert clinical knowledge with an anatomically aligned ROI could provide remarkable results in DRIL detection applications. This work attempts to establish an anatomically relevant framework for computerized DRIL identification that focuses on the highly crucial macular region, possibly helping in faster intervention and improved diagnosis in the management of DME.

Elastin–collagen nanoparticles (ECnPs) have been shown in our previous studies to self-assemble into different morphologies, including nanoplates and nanovesicles, by manipulating the sequence length of the elastin-like peptide (ELPs) and collagen-like peptide (CLPs) of a given conjugate. In this work, we demonstrate that the morphologies of ECnPs can also be modulated, for a given ECnP sequence, with variations in solution pH and/or the amount of encapsulated drug. Specifically, the peptide (VPGYG)6-(GPO)8 preferentially formed nanovesicles under basic conditions but assembled into nanoplates under acidic conditions. Another sequence, (VPGWG)2(VPGFG)2-(GPO)8, produced nanovesicles when loaded with a high concentration of dexamethasone-carboxyfluorescein (Dex-CF), but transitioned to nanoplates at lower drug loading. Furthermore, in addition to the different morphologies observed for a given set of initial solution conditions, our studies also illustrate the possibility of triggering vesicle-to-plate transformations for a given ECnP with release of Dex-CF over time. These results highlight multiple avenues for controlling ECnP morphology, expanding their applicability as a flexible and efficient drug delivery platform.

C4 grasses are crucial for food and biofuel production. Originating from warm regions of the world, C4-photosynthesizing plants typically exhibit poor chilling tolerance. Some C4 grasses of Miscanthus are recognized for their exceptional chilling tolerance, however, the mechanism behind it is not fully understood. Here, we hypothesize that the rapid adjustment of leaf pigment composition contributes to mechanisms that protect photosynthesis during the initial short-term response to chilling. Miscanthus accessions with documented contrasting levels of chilling tolerance were subjected to chilling under dark and light conditions with or without nutrient limitations. The changes in pigment composition were assessed by hyperspectral indexes and molecularly validated. Our results showed that high-chilling-tolerant accession accumulates zeaxanthin and anthocyanins while reducing chlorophyll content at the end of chilling night when grown on the low-fertility soil. Interestingly, at the end of the night, low soil fertility alone was able to induce a significant difference in zeaxanthin accumulation between accessions. Night-accumulated zeaxanthin led to 38% faster NPQ following morning. Transcriptional differential regulation of enzymes involved in pigment anabolism and catabolism supports the dynamic adjustment in leaf pigment composition. The investigated changes in pigment composition can inspire new strategies to engineer crops for better stress resistance.

Computer vision presents a great opportunity for improving pest monitoring in agriculture, particularly for yellow sticky traps, a critical component in IPM. However, despite the growing interest in applying object detection models for insect identification, insect datasets present unique challenges, and approaches for fine-tuning model parameters to achieve reliable performance remain limited. This study explores the influence of fine tuning three key hyperparameters (learning rate, optimizer type and batch size) on the performance of two popular object detection models (YOLO and RT-DETR), in detecting pests on yellow sticky traps, with a particular emphasis on identifying cucumber beetles. Results showed that higher learning rates reduced performance across precision, recall, and mAP50 for both models. In contrast, SGD improved outcomes, particularly for RT-DETR, while YOLO proved more robust to high learning rates. Our study also showed that both models achieved comparable accuracy levels, once optimal settings were determined for each model. These findings highlight the importance of hyperparameter tuning for reliable pest detection systems and support the development of scalable AI workflows for precision agriculture.

Gata3 is an essential transcription factor for the development of several distinct immune cell lineages such as T cells, natural killer (NK) cells, and innate lymphoid cells (ILCs). As such, the levels and timing of Gata3 expression are critical for directing lineage fate decisions. The Gata3 locus has a complex and dynamic distal regulatory enhancer landscape. Recently, we identified a non-coding RNA, Dreg1, located immediately upstream of the classic +280 kb T/NK cell enhancer (Tce1). To test its function, we excised the Dreg1 locus in mice and observed a selective reduction of group 2 ILCs (ILC2) across multiple tissues, but mature T, NK, and other ILC lineages remained unchanged. In bone marrow, common innate lymphoid cell progenitors (ILCPs) increased while ILC2 progenitors (ILC2P) decreased, with a modest reduction of Gata3 in upstream progenitors consistent with an early developmental bottleneck. Chromatin profiling showed the Dreg1 locus is accessible in early lymphoid progenitors and became decorated with H3K27ac in ILCP in a Tcf1-dependent manner. Furthermore, Tcf1-deficient cells did not express Dreg1 and showed alterations in the epigenetic landscape of the Dreg1 locus. Finally, we discovered that potential homologues of Dreg1 harboured in a syntenic enhancer of GATA3 are also highly expressed in human ILC2. Taken together, we conclude that Dreg1 is a Tcf1-dependent non-coding RNA critical for fine tuning the high level of Gata3 required for the optimal development of the ILC2 lineage.

Omnidirectional cameras are a suitable and cost-effective choice for Visual Place Recognition (VPR), as they provide comprehensive information from the scene regardless of the robot orientation. However, vision sensors are vulnerable to environmental appearance changes (e.g., illumination, weather, season or moving objects). While multi-modal sensing approaches can overcome these challenges, they introduce significant cost and system complexity. This paper introduces PDPR (Panoramic-Depth Place Recognition), a novel fusion framework that enhances the robustness of VPR methods by integrating visual data with geometric features derived from monocular depth estimation techniques, while using a single-camera setup. In the ablation study, both early and late fusion strategies are evaluated to optimally combine appearance-based and depth-derived features. The extensive evaluation on challenging, indoor and outdoor datasets demonstrates that PDPR consistently boosts retrieval performance across multiple state-of-the-art VPR models. Furthermore, this improvement is achieved without requiring any fine tuning, allowing our method to function as a pluggable module for pretrained models. Consequently, this work presents a powerful, practical and low-cost solution for robust VPR, with high potential to scale as monocular depth estimation and VPR models continue to improve. The project website can be found at https://marcosalfaro.github.io/projects-PDPR/ . • Monocular depth estimation is used to enhance place recognition. • A thorough evaluation of preprocessing techniques to enhance the depth maps. • Fusion techniques are designed to leverage visual and geometric data. • A model-agnostic approach that improves the performance even with no fine tuning. • A robust method across different scenarios and lighting conditions.

The Golgi apparatus is a highly dynamic organelle and central to the regulation of protein glycosylation, cargo sorting and additional cellular processes such as mitosis, stress responses, autophagy and inflammation. There have been major advances in understanding the dynamics of the Golgi apparatus and the relationship between remodelling the Golgi architecture and function. Membrane structural/scaffold proteins of the Golgi, including golgins, GRASPs and adaptors interact with a diverse range of cytoskeletal and signalling molecules and play a major role in the regulation of the Golgi morphology and function. Modulation of the higher-order Golgi ribbon architecture in mammalian cells is directly associated with physiological and pathological responses, including neurological diseases and cancer. An important question is the influence of morphological states of the Golgi architecture on the fine tuning of glycosylation. Here we review the relationship between the morphology and function of this organelle, in physiology and disease, especially to in relation to the impact of the fragmentation of the Golgi ribbon on the steady state location of glycosyltransferases and glycan synthesis. The current unresolved issues relevant to changes in Golgi morphology on protein glycosylation are highlighted.

The essential elements of an ecological approach to Pavlovian conditioning.

With the rapid development of the industrial Internet, multi-machine collaborative control has become a key technology for enhancing the efficiency and stability of industrial systems. Traditional optimization methods are difficult to meet its requirements of high dimension, nonlinearity and real-time performance. Therefore, this study proposes an intelligent algorithm-based optimization strategy for multi-machine collaborative control, implemented through a PSO-DDPG hybrid framework combining Particle Swarm Optimization (PSO) and Deep Deterministic Policy Gradient (DDPG). This method combines the global search ability of PSO and the local fine tuning of DDPG, and solves the optimization problem in the dynamic control environment through adaptive strategy adjustment. The experimental results show that PSO-DDPG performs outstandingly in control accuracy, system stability and computational efficiency compared with traditional algorithms (such as PSO, DQN and GA) in multi-machine cooperative control. Specifically, PSO-DDPG has improved control accuracy by 9.8%, reduced response time by 12.3%, reduced energy consumption by 15.2%, and improved computational efficiency by 22%, respectively. Especially in complex environments such as load fluctuations and equipment failures, PSO-DDPG can adjust the control strategy in real time, significantly improving resource utilization and the adaptability of the system. In addition, PSO-DDPG also has significant advantages in reducing computing resource consumption and ensuring system stability, providing an efficient and reliable solution for dynamic optimization tasks in the industrial Internet.