A comprehensive analysis of viral mutations is essential for understanding viral evolution, disease epidemiology, diagnosis, drug resistance, etc. However, challenges remain in capturing complex mutation patterns and supporting diverse viral families with varying genome architectures. To address these needs, we present VARIANT, an web server for mutational analysis of RNA viral genomes and associated viral products across both single- and multi-segment virus genomes. The server takes as input a viral reference genome, a reference protein sequence, and/or multiple sequence alignment, and automatically provides full annotation of mutation types, including standard categories such as point mutations (missense, silent, and nonsense), insertions, deletions, or frameshift events in both coding and non-coding regions. In addition, VARIANT detects three biologically significant mutation patterns that are overlooked by conventional software/packages: ``row mutations'' (consecutive substitutions within a window of 3 nts), ``hot mutations'' (two non-consecutive substitutions within a window of 3 nts), and potential programmed ribosomal frameshifting (PRF) regions. The server currently contains automatic analysis of major viral pathogens, including SARS-CoV-2, HIV-1, Influenza H3N2, Ebola virus, and Chikungunya virus. It also allows users to analyze customized viruses. Users can track VARIANT analysis progress in real time, visualize mutation distributions, and download structured results in ZIP format. VARIANT also incorporates dual graph topology analysis to classify frameshifting element structures from dot-bracket notation input. This feature enables systematic comparison of RNA secondary structure motifs across viral families by mapping structures to a comprehensive library of dual graph topologies. The web server is freely available at https://variant.up.railway.app.
Co-translational protein folding is shaped by the vectorial nature of translation, which causes residues to emerge sequentially from the ribosome. As a result, residues whose native interaction partners lie downstream in sequence cannot immediately form their native contacts and remain transiently unsatisfied until those partners are synthesized. These unsatisfied residues are vulnerable to non-native interactions and often require the engagement of co-translational chaperones. We previously developed the Native Fold Delay (NFD) metric to quantify the time lag between the synthesis of a residue and the point at which it can form all its native contacts. Here, we present the FoldDelay web server, a freely accessible platform that extends the NFD concept into a more comprehensive framework for analyzing native residue-residue contact formation during translation. Starting from user-submitted AlphaFold or PDB structures, the site identifies all N- to C-terminal residue-residue contacts, estimates their earliest possible formation times, and integrates domain annotations to distinguish between intra- and inter-domain contacts. The server provides a suite of linked interactive visualizations that allows users to explore native contact formation dynamics and detect transiently unsatisfied regions. The FoldDelay web server is freely accessible at https://folddelay.switchlab.org.
RNA secondary structure plays a critical role in gene regulation, yet existing computational and experimental tools for structure analysis are often fragmented across prediction, ensemble modeling, and functional interpretation workflows. Here, we present ShapeRNA, a user-friendly web server for integrated RNA secondary structure prediction, ensemble inference, and structure-aware regulatory annotation. ShapeRNA supports three complementary analytical workflows, including sequence-based structure prediction, reactivity-guided modeling using SHAPE or DMS data, and sequencing-guided ensemble inference from high-throughput probing experiments. The platform integrates multiple established prediction algorithms and provides standardized data processing, ensemble clustering, and visualization. In addition, ShapeRNA enables mapping of RNA modification sites, microRNA target regions, and RNA-binding protein interaction motifs onto predicted RNA structures and representative ensemble conformations. We demonstrate the utility of ShapeRNA through applications including analysis of mutation-associated structural changes in MAPT exon 10, characterization of conformational heterogeneity in the HIV-1 Rev Response Element, and regulatory annotation of the oncogenic long non-coding RNA HULC. ShapeRNA provides an accessible and extensible platform for investigating RNA structural heterogeneity and regulatory mechanisms. This website is free and open to all users, and there is no login requirement. The server is accessible at https://shaperna.com.
The HMMER web server, available at https://www.ebi.ac.uk/Tools/hmmer, provides online access to tools from the HMMER software suite (http://hmmer.org/) for protein analysis using profile hidden Markov models. Users can perform sequence similarity searches against a range of regularly updated protein sequence databases or annotate protein sequences with domains and families using profile HMM libraries from protein family databases. Since the 2018 update, the continued exponential growth of sequence databases has necessitated substantial infrastructural improvements to maintain search performance speed and service reliability. To achieve this, the web interface has been completely reengineered using modern web technologies (JavaScript and React), providing users with an enhanced experience, including session-based search history and streamlined results visualization. The web application programming interface has been rewritten to better support programmatic access with updated endpoints and JSON-based responses. The infrastructure has been redesigned to efficiently handle searches against much larger databases through horizontal scaling and asynchronous job processing. Target database offerings have been updated to reflect current usage patterns and data availability. The HMMER web server is free and open to all users, and there is no login requirement.
Protein-protein and protein-nucleic acid interactions are fundamental to numerous cellular functions, yet only a small fraction have been experimentally characterized. Although modern computational methods have been developed for predicting interacting residues in proteins, they are challenging to use due to individual installation and execution requirements, lack of a standardized input or output format, inability to cover multiple target biomolecules and absence of support for result analysis. Moreover, the performance of many methods varies across different proteins. For instance, algorithms trained on complexes or intrinsically disordered regions may not perform well on other types. To overcome these challenges, we have developed PROBind, a web server for predicting, analyzing and interactively visualizing protein, DNA and RNA binding residues. PROBind integrates 12 predictors trained on structural or disordered proteins. It supports protein sequences and structures as input for predicting binding residues and allows for the integration of prediction results from external predictors. By normalizing and averaging predictions from multiple predictors targeting the same ligand, PROBind generates meta-predictions that balance discrepancies among different methods. Furthermore, it provides interactive graphical tools for result analysis and contextualization. Overall, PROBind accommodates diverse ligand types and supports predictions and analysis based on both structure and sequence data, overcoming the limitations of existing tools. PROBind is freely accessible at https://www.csuligroup.com/PROBind.
Metagenomic sequencing has transformed virus discovery; however, downstream bioinformatic analyses for viral identification, classification, and host prediction remain fragmented across multiple tools. Here, we present PhaBOX2, a major upgrade that extends the platform from a specialized bacteriophage identification tool to a comprehensive and integrated suite for viral sequence analysis. PhaBOX2 broadens its detection, taxonomic, and host prediction scope beyond phages to enable the characterization of archaeal and eukaryotic viruses. The updated workflow incorporates rigorous quality control and quantitative analyses, automatically removes host contamination, clusters sequences into viral operational taxonomic units, and performs phylogenetic analysis based on marker genes. In contrast to traditional "black-box" deep learning approaches, PhaBOX2 combines alignment-based strategies with machine-learning models under a "glass-box" design philosophy, providing interpretable intermediate evidence alongside final predictions to improve transparency and biological interpretability. Powered by a dedicated high-performance computing infrastructure, the server delivers a fully automated, end-to-end workflow, while achieving an ~80% reduction in processing time. PhaBOX2 thus provides a robust and user-friendly ecosystem for viral metagenomic analysis and is freely available at https://phage.ee.cityu.edu.hk/.
暂无摘要（点击查看详情）
暂无摘要（点击查看详情）
This paper focuses on Bayesian inference of an M/M/1 queuing model with balking, a phenomenon in which customers choose not to join a queue due to long waiting line. In this paper, the balking probability is considered as a function of number of customers and their impatience level. The degree of impatience of customers plays a crucial role effecting the balking probability. Higher threshold of impatience implies that customers are more sensitive to queue length, i.e. they are less willing to join a queue when the queue is even slightly longer. Conversely, lower threshold of impatience indicates that customers will less balk. In this scenario, there is a higher probability that customers will opt to join the queue, even when it extends to a considerable length. This paper provides the Bayesian estimates for traffic intensity (ρ), employing various prior distributions such as beta, truncated gamma, and uniform prior distributions under a squared error loss function. Through the sampling importance re-sampling (SIR) technique, we obtained the posterior estimates, risk, and credible intervals that showcase the effectiveness of our methodology. Furthermore, simulation studies demonstrate the convergence of estimators, and our findings are further validated through analysis of real-life data.
Fairness-aware federated graph neural networks (FedGNNs) necessitate consideration of both the server and the clients. However, fairness-aware methods struggle to enhance dual-perspective (i.e., server and clients) fairness without sacrificing utility due to the distributed learning framework. As a consequence, the utility sacrifices of fairness-aware graph learning methods are even exacerbated in federated frameworks. In this work we propose F3GL, a dual-perspective fairness federated graph learning method that enhances both global (for the server) and local fairness (for clients) while preserving utility. Through theoretical analysis, we delineate the similarity between original sensitive features and those after convolution under different spectra. Our findings reveal that only the principal eigenvalue contributes to enhancing this similarity. Moreover, our theoretical analysis applies universally to both clients and servers. Specifically, employing a specialized eigenvalue selection strategy allows for effective optimization of both local and global fairness. Drawing on these insights, we improve dual-perspective fairness through the lens of spectral graph theory without sacrificing utility. Experimental results on two real-world datasets show the superiority of F3GL over existing baselines.
PURPOSE: To share first experience from the Swiss RetinAI Consortium in applying artificial intelligence (AI)-based optical coherence tomography (OCT) analysis across multiple centres, to highlight practical barriers encountered during implementation, and to outline practical recommendations for future collaborative multicentre AI-based OCT research. METHODS: The Swiss RetinAI Consortium, consisting of six ophthalmology centres, implemented the FDA-cleared Discovery platform (RetinAI, Bern, Switzerland) for collaborative OCT analysis. Challenges encountered during protocol development for OCT export, anonymisation, data upload, AI-based analysis, and data sharing were systematically collected from all sites involved. These included regulatory, technical, and methodological aspects. The findings were discussed at expert meetings and consolidated into shared guidance for data acquisition and processing. The aim was to provide practical recommendations to support standardised workflows in future multicentre AI-driven research. IDENTIFIED CHALLENGES: Key regulatory, technical, and methodological challenges were identified during multicentre implementation. These included the legal requirement for Swiss-based server hosting, incomplete anonymisation due to heterogeneous export protocols, and non-standardised OCT protocols treated as being equivalent, despite covering different retinal areas. Additional barriers were inter-device variability, ETDRS grid misalignment without the option for manual correction, major segmentation errors requiring extensive review, and unfiltered data exports that often exceeded one million entries, thereby increasing the risk of readout errors. Moreover, the absence of automated quality screening and the lack of cross-centre reproducibility data were identified as important methodological gaps. Together, these challenges shaped our practical recommendations for reliable multicentre AI-based OCT analysis. PRACTICAL RECOMMENDATIONS: Based on the identified challenges, we derive practical recommendations focusing on standardised anonymised data export procedures, harmonised OCT acquisition protocols, mandatory quality control steps for grid placement and segmentation, and structured strategies for handling large-scale output data. Our recommendations offer a practical roadmap to support future collaborative multicentre AI-based OCT research. ZIEL: Ziel dieser Arbeit ist es, erste Erfahrungen des Swiss RetinAI Consortiums bei der multizentrischen Anwendung einer KI-basierten Analyse der optischen Kohärenztomografie (OCT) darzustellen, praxisrelevante Hürden während der Implementierung zu identifizieren und konkrete Empfehlungen für zukünftige multizentrische KI-basierte OCT-Forschungsprojekte abzuleiten. Das Swiss RetinAI Consortium, bestehend aus 6 ophthalmologischen Zentren, führte im Rahmen eines multizentrischen Implementierungsprozesses eine KI-basierte OCT-Analyse mit der FDA-zugelassenen Discovery-Plattform (RetinAI, Bern, CH) durch. Herausforderungen entlang des OCT-Exports, der Anonymisierung, des Daten-Uploads, der KI-basierten Analyse und des Datenaustauschs wurden multizentrisch systematisch erhoben. Diese umfassten regulatorische, technische und methodische Aspekte. Die Ergebnisse wurden in Expertenrunden diskutiert und in gemeinsame Handlungsempfehlungen überführt. Zu den identifizierten Herausforderungen zählten die Anforderung an den Serverstandort in der Schweiz, unvollständige Anonymisierung infolge heterogener Exportprozesse sowie nicht standardisierte OCT-Aufnahmeprotokolle, die trotz unterschiedlicher erfasster Netzhautareale als gleichwertig behandelt wurden. Weitere Hürden waren Inter-Device-Variabilität, eine fehlerhafte automatische Platzierung des ETDRS-Gitters ohne manuelle Korrekturmöglichkeit, ausgeprägte Segmentierungsfehler mit hohem manuellem Korrekturaufwand sowie unstrukturierte Datenexporte mit teils über einer Million Einträgen, die das Risiko von Auswertungsfehlern erhöhen. Darüber hinaus stellten fehlende automatisierte Qualitätskontrollen und begrenzte multizentrische Reproduzierbarkeitsdaten relevante methodische Barrieren dar. Diese Erkenntnisse bildeten die Grundlage für die Ableitung praxisorientierter Empfehlungen. Auf Basis der identifizierten Herausforderungen haben wir praxisnahe Empfehlungen formuliert, die sich auf standardisierte und vollständig anonymisierte Datenexporte, harmonisierte OCT-Akquisitionsprotokolle, verpflichtende Qualitätskontrollschritte zur Überprüfung von ETDRS-Gitterplatzierung und (Miss-)Segmentierung sowie strukturierte Strategien zur Verarbeitung großer Datenoutputs konzentrieren. Diese Empfehlungen werden in einer praxisorientierten Roadmap zusammengefasst und sollen zukünftige multizentrische KI-basierte OCT-Forschungsprojekte unterstützen.
Recent developments in immersive tools for radiation-related training have been notable. However, these training programmes have primarily been developed from a first-person perspective, often limiting the sharing of the training experience. This study developed a tool that facilitates real-time spatial sharing of training using mixed reality (MR) and augmented reality (AR). Our MR system, which wirelessly connected HoloLens 2 with a survey meter mock-up, enabled the visualisation of radiation and simulated a detector's response. It also shared training information every second via a WebSocket server. Concurrently, the AR system allowed observers, equipped with an AR tablet, to view the MR operator's status in real time based on this shared data. Post-training, the shared data were saved as text, enabling reflective learning through visualisation. We anticipate that this tool would significantly enhance the efficiency of survey meter training without radiation exposure.
In IIoT edge computing, multi-edge server collaborative scheduling faces two core issues due to random task arrivals, heterogeneous resources, and complex topology: traditional model-driven methods cannot make dynamic decisions in dynamic environments, and conventional MARL fails to characterize inter-node topological dependencies and load correlations. To address this, this paper investigates the joint optimization of task offloading, computing resource allocation, and SFC orchestration in IIoT, constructs a cloud-edge-end collaborative architecture, and models the problem as a POMDP to minimize the overall system cost under multiple constraints. A graph-guided value-decomposition MARL method is proposed, which extracts spatial topology and neighborhood-load features of edge nodes via a GNN and combines them with the QMIX framework to realize multi-agent centralized training and distributed execution. Simulations show that the algorithm converges stably under different server scales and task loads, significantly outperforms benchmark algorithms, and can suppress performance degradation in high-load scenarios, demonstrating its robustness and scalability in complex industrial environments.
Digital Pathology (DP) is a fast-emerging branch of pathology focused on digitizing pathology data. A key challenge of DP usage for pathology laboratories, especially mid- to small-sized clinical labs, are the upfront costs associated with instrumentation and the logistical challenges of implementation. In the current project, we built an end-to-end DP solution using low-cost, open-source components that is user-friendly at a small scale. We repurposed readily available microscopy components in a pathology lab to assemble a fully functional DP pipeline for translational research applications. We tested multiple low-cost complementary metal-oxide semiconductor (CMOS) cameras in this project and chose a user-friendly Canon camera for image acquisition. An open-source DP server solution, OMERO v.5.6.4, was used as the image management system (IMS) to host and serve the WSIs on an Ubuntu 22.04 operating system. The server-hosted WSI images were evaluated remotely and asynchronously by multiple pathologists physically situated in Albuquerque, NM; Salt Lake City, UT; and Palo Alto, CA. Each pathologist assessed the quality of the WSI pipeline, image quality, and WSI interaction experience using a 23-question survey. Overall, the custom, low-cost WSI pipeline was noted to be a robust and user-friendly experience by the pathologists. The current DP setup is unlikely to be useful as a commercial, scalable DP pipeline for large-scale clinical applications. However, it demonstrates the feasibility of creating customized, small-scale DP solutions (at a low price point) for asynchronous translational pathology research applications. Additionally, building customized DP pipelines provides excellent educational opportunities for pathology residents to gain in-depth knowledge of the various technical elements of a DP workflow. In summary, we have established a low-cost, end-to-end WSI DP pipeline useful for spatiotemporally asynchronous translational pathology research, in an academic setting.
In vehicular networks, federated learning faces significant challenges due to resource heterogeneity, dynamic participation patterns, and intermittent connectivity among vehicles. Traditional client selection mechanisms often fail to consider the two-tier decision-making process inherent in vehicular network environments, where both central servers and individual vehicles must make participation decisions based on their respective constraints. Moreover, existing model aggregation algorithms typically assume fixed client participation and cannot adapt to the highly variable participation patterns unique to vehicular networks. This paper proposes a comprehensive vehicular federated learning framework with three key innovations. First, we introduce a strategy-driven adaptive client participation mechanism with a two-tier decision-making process that combines server-side reinforcement learning-based client selection with client-side autonomous participation decisions based on local resource thresholds. Second, we develop an incremental online policy learning algorithm based on Proximal Policy Optimization (IO-PPO) to address the data scarcity challenge in federated learning environments by enabling continuous learning from limited trajectory data. Third, we propose a dynamic client size-adaptive optimized model aggregation algorithm that adapts to different participation patterns while considering both current and historical client contributions. Our approach leverages a synergistic combination of reinforcement learning for adaptive decision-making, asynchronous federated learning principles for flexible participation, and graph-based modeling for capturing network topology effects. Extensive experimental results demonstrate that compared to existing methods, the proposed framework significantly improves learning efficiency, convergence stability, and model performance in realistic vehicular network scenarios.
Monkeypox virus (MPXV) is emerging as a global public health concern due to its nature of spread. There are limited treatment options, as the sole drug for treatment is lacking, highlighting the need for new therapeutic options. The use of computer-aided drugs discovery such as molecular docking, molecular dynamic (MD) simulations and post-simulation analysis are important tools in identifying potential compounds that can target specific proteins of the virus, such as DNA polymerase to stop virus replication. This study employed molecular docking and molecular simulation with the aim to identify potential inhibitors for MPXV treatment from the ZINC Database. Molecular docking was performed using PyRx 0.8 version after virtual screening of the ZINC database using the Tranches tool; then, toxicity prediction of the selected compounds was performed using the ProTox-3.0 web server. Molecular dynamics simulation was conducted using GROMACS version 4.5 to evaluate the structural stability and dynamic behavior of the protein-ligand complex for the best interacting compound. Furthermore, post-simulation analysis was conducted using standard GROMACS utilities for visualizing time-dependent properties from MD simulations. A total of 16 compounds were shortlisted based on their molecular docking scores and interaction profiles with the monkeypox virus DNA polymerase (PDB ID: 8HG1). The leading compound, ZINC000019418450, demonstrated strong binding affinity (-7.4 kcal/mol). According to post-simulation analysis, all top compounds formed between one and five hydrogen bonds and up to eleven hydrophobic contacts with residues within the active site, thus providing strong geometric and energetic evidence for binding stability. Notably, our identification of ZINC000104288636 as a Class 6 compound with an LD50 of 23,000 mg/kg adds translational value by highlighting candidates with low predicted acute toxicity. Overall, this study lays a solid foundation for the rational design of next-generation monkeypox antiviral therapeutics. Future work is needed for experimental validation of prioritized compounds to assess their biochemical efficacy and pharmacological potential.
Protein sequence alignment is one of the most fundamental procedures in bioinformatics. Due to its many downstream applications, improvements to this procedure are of great importance. We consider two revolutionary concepts that emerged recently as candidates for improving the state-of-the-art alignment methods: AlphaFold and protein language models such as Ankh, ProtT5, or ESM-C. Alignment improvements can come from the structural alignment of AlphaFold-predicted structures or the scoring based on the similarity of protein embeddings produced by the protein language models. Thorough comparison on many domains from BAliBASE and CDD demonstrates that the Ankh-score method produces much better sequence alignments than the structural alignments using US-align of AlphaFold3-predicted structures. Both are better than the traditional method using BLOSUM matrices. This suggests that Ankh embeddings may possess certain information that is not available in the AlphaFold3-predicted structures. The alignment software is freely available as a web server at e-score.csd.uwo.ca and as source code at github.com/lucian-ilie/E-score.
The present dataset contains three years of continuous, high-frequency air-quality monitoring of particulate matter (PM2.5 and PM10) in the metropolitan area of Asunción and the Western Region of Paraguay. Data were collected from eleven monitoring stations: nine in urban Asunción, one at the National University of Asunción campus in San Lorenzo, and one in the semi-rural Cerrito, Paraguayan Chaco. The monitoring period covers from April 2019 to December 2021. The monitoring network was established to record spatio-temporal variations in particulate matter concentrations throughout urban and semi-rural environments. Key meteorological variables, including temperature, relative humidity, atmospheric pressure, and wind direction, were continuously recorded to enable analysis of dispersion trends. Particulate matter concentrations were measured at five-minute intervals using Optical Particle Counters (OPC-N2), with data automatically transmitted to a central server every thirty minutes for quality control, storage, and processing. This dataset allows direct comparison with national air-quality standards established by the National Resolution No 259/2015 of the Ministry of Environment and Sustainable Development. Its full coverage supports policy evaluation, cross-sector assessments, and studies on environmental and public health impacts. Full technical documentation is included to secure transparency, traceability, and reproducibility, providing the basis for management and scientific research in Paraguay.
Limosilactobacillus reuteri, formerly Lactobacillus reuteri, is a rod-shaped, Gram-positive, facultative anaerobe that colonizes the gastrointestinal tract of most vertebrates, including humans. We report the first isolation of L. reuteri strain HDB from the stool of a healthy Indian infant. Species assignment using the Type (Strain) Genome Server (TYGS) placed HDB within the L. reuteri clade, showing closest affinity to L. reuteri subspecies porcinus (dDDH 69.7%) yet clustering phylogenomically with L. reuteri DSM 17938. Hybrid de novo assembly (Illumina + Oxford Nanopore GridION MK1) generated a single circular 2,226,956 bp chromosome (GC 39.04%) encoding 2,160 CDS. Functional annotation identified genes involved in vitamin B12 biosynthesis, reuterin production, and probiotic functions, along with enriched carbohydrate and cofactor metabolic pathways. Comparative analysis with 59 L. reuteri genomes revealed a pangenome of 11,725 gene families, including 944 core gene families, 171 soft-core gene families, 1912 shell gene families, and 8698 cloud gene families, highlighting notable diversity. Core-genome phylogeny aligns HDB closely with the reference strain DSM 17938, confirming its identity as a human-associated lineage. dN/dS analysis indicated strong purifying selection across host niches, with no evidence of widespread positive selection. Genome-scale modeling predicts expanded carbohydrate flux in HDB against global references. The genetic background, along with its conserved metabolic features, suggests that HDB carries genomic characteristics commonly associated with human-derived L. reuteri strains. These observations support its consideration for further evaluation as a regionally sourced probiotic candidate. These conclusions are based on genomic and computational predictions and require experimental validation through adhesion, colonization, and safety studies.
Nanobody binding is largely governed by the HCDR3 loop, which adopts distinct placement regimes relative to the framework: compact, framework-contacting (kinked blueprint) and solvent-exposed (extended blueprint). Many nanobodies also contain additional cysteines that form non-canonical disulphide bonds, imposing covalent constraints on binding-loop conformations. Current structure predictors are typically trained and benchmarked with smooth coordinate-based objectives, so models may appear reasonable under root-mean-square deviation (RMSD), while adopting an incorrect HCDR3 blueprint or failing to recover the native disulphide connectivity, impacting paratope geometry and functional interpretation. Here, we show that the HCDR3 blueprint is predictable from sequence alone, allowing for explicit constraints during modelling. We implement these principles into NbForge, a lightweight nanobody folding model that incorporates blueprint- and disulphide-aware inductive biases and is trained with filtered self-distillation. NbForge improves recovery of HCDR3 blueprint and non-canonical disulphide formation over previous lightweight models and achieves coordinate accuracy at par to state-of-the-art, large, resource-intensive predictors, while running at sub-second inference speed. We show that using NbForge monomer models as templates further improves the success rate of predicting nanobody-antigen complexes. Together, these results motivate blueprint- and disulphide-aware benchmarks for nanobody modelling beyond RMSD, and show that appropriate inductive biases can close the performance gap to heavyweight predictors. We make the sequence classifier (NbFrame) and NbForge available for download and via a user-friendly web server.