AutoWAFuzzer: An adaptive framework for web application firewall penetration testing with multi-agent system and RAG-enabled reinforcement learning

Antibiotic resistance prediction with an attention-based bi-LSTM clinical decision support system.

Arachnoid cysts are extra-axial cerebrospinal fluid collections within the arachnoid membrane. Ruptured or hemorrhagic arachnoid cysts, although rare, present significant controversies in management. The present study is an attempt to analyze the factors contributing to management decision of ruptured/hemorrhagic arachnoid cysts using patient-level data from the literature. A literature search was conducted on PubMed and EMBASE to identify case reports and series of ruptured arachnoid cysts. Tree-augmented naïve Bayes (TAN) classifiers were implemented to analyze factors influencing surgical decision. The dataset was split into training and testing sets (0.75:0.25) and augmented using data augmentation techniques to address class imbalance. TAN classifiers were evaluated for accuracy and area under the curve, and a web application was developed to explore the networks. The dataset included 254 unique cases after exclusion of missing data. Middle cranial fossa cysts accounted for 95% of cases, with a male predominance (M:F ratio 4.29:1). Management was predominantly surgical (89.8%), with craniotomy being the most common procedure. TAN classifiers for surgery and type of surgery were validated internally with accuracies of 90.48 and 75%, respectively. Cyst location, presence and type of hemorrhage, patient age group, Galassi classification were key influencing variables. The choice of surgical modality was influenced by additional variables like head injury, seizure, and macrocrania. TAN models highlighted the interrelated factors influencing management decision but do not propose definitive strategies. The generalizability of the findings are limited by heterogenous data, imbalance of various management strategies, particularly conservative management, and evolution of surgical techniques over time. The complexity of decision-making underscores the need for multicenter registries to improve data quality and to formulate optimal management strategy.

NeuroBACE-ML: A reliability-aware screening framework for high-throughput prioritization of potent BACE1 inhibitors.

Integrating Shiny Web Applications into Pediatric Transplant Research: Development, Implementation, and Case Application

Histology education relies on the visual interpretation of microscopic structures and requires repeated practice for effective learning. Advances in educational technology, particularly virtual microscopy, have created new opportunities for improving histology education. During the COVID-19 pandemic, online learning approaches became essential alternatives to traditional laboratory-based teaching. This study aimed to compare the effects of the eHistoLab web application and conventional e-lecturing on students' learning outcomes and perceptions in histology education among medical sciences students in Iran. This quasi-experimental study used a two-group pre-test/post-test design. A total of 100 medical, dentistry, and pharmacy students enrolled in a histology course at Smart University of Medical Sciences (SMUMS), Iran, during the summer semester of 2023 participated in the study. Participants were randomly assigned into two groups. One group received instruction through e-lecturing, while the other group was taught using the eHistoLab web application, a bilingual virtual microscopy platform providing access to high-resolution histological slides, image magnification, search functions, and interactive communication with instructors. Learning outcomes were assessed using pre- and post-tests, and data were analyzed using paired and independent t-tests in SPSS version 24. The findings revealed that 50% of students preferred virtual microscopy alone for learning histology, while 30% preferred a combination of methods, and 20% preferred e-lecturing alone. Regarding educational resources, 58% of students preferred the eHistoLab application, whereas 26% preferred textbooks and 16% preferred recorded e-lectures. Meantime, 62% of students preferred the eHistoLab tool in terms of motivation for learning histology, whereas 30% preferred the combined approach and 8% favored recorded e-lectures. Moreover, the results indicated that the post-test scores in the group taught with the eHistoLab web application were significantly higher than those in the e-lecturing group (P < 0.001). The findings of this study suggest that the eHistoLab web application improved students' learning outcomes and was positively perceived as a learning tool in histology education. Virtual microscopy may therefore serve as a useful complementary approach to lecture-based histology teaching.

Cross-site scripting (XSS) poses a significant threat to web application security. While Deep Learning (DL) has shown remarkable success in detecting XSS attacks, it remains vulnerable to adversarial attacks due to the discontinuous nature of the mapping between the input (i.e., the attack) and the output (i.e., the prediction of the model whether an input is classified as XSS or benign). These adversarial attacks employ mutation-based strategies for different components of XSS attack vectors, allowing adversarial agents to iteratively select mutations to evade detection. Our work replicates a state-of-the-art XSS adversarial attack, highlighting threats to validity in the reference work and extending it towards a more effective evaluation strategy. Moreover, we introduce an XSS Oracle to mitigate these threats. The experimental results show that our approach achieves an escape rate above 96% when the threats to validity of the replicated technique are addressed.

The accurate parameterization of drug-like molecules is a fundamental prerequisite for molecular dynamics simulations in structure-based drug design. While the AM1-BCC charge model has served as the de facto "standard" for GAFF2 parameterization for two decades, the recently introduced ABCG2 model offers substantially improved accuracy in predicting hydration free energies and other key physicochemical properties. However, accessing ABCG2 parameters traditionally requires downloading and installing the full AmberTools suite-a multi-gigabyte software package with complex dependencies-presenting a significant barrier for many practitioners, particularly experimental collaborators and researchers new to the field. Here we present a major upgrade to the PrimaDORAC web interface that makes ABCG2 parameterization readily accessible to the entire drug design community. Through a minimalistic integration of essential AmberTools components directly into the web application framework, users can now obtain GAFF2 parameters with ABCG2 charges for any drug-like molecule in seconds, without any software installation. The interface accepts a single SMILES string or structure file and returns a complete archive containing GROMACS-compatible topology files and a ready-to-use PDB structure. By removing all technical barriers to accessing state-of-the-art ABCG2 parameters, the upgraded PrimaDORAC interface is aimed at empowering medicinal chemists, pharmacologists, and computational researchers alike to conduct more accurate MD simulations with minimal effort. The service is freely available at www1.chim.unifi.it/orac.

idney stone disease is a common urological condition where delayed detection can lead to severe complications. Diagnosis using ultrasound imaging depends on expert interpretation, which may not always be available in resource-limited settings. Manual analysis is time-consuming and pronetovariability,reducingthereliabilityofearly screening. Although artificial intelligence has shownpotentialinmedicalimaging,manyexisting systems lack robustness in handling irrelevant inputs. To address this, this work presents an AI-based kidney stone detection system using ultrasound imaging. The framework utilizes a MobileNetV2-based convolutional neural network with transfer learning to classify images into Normal and Kidney Stone Detected categories. A deep feature-based validation mechanism ensures that only relevant kidney ultrasound images are processed, preventing incorrect predictions. Grad-CAMvisualizationsareusedtohighlightimportant regions influencing the model’s decision. Experimental results demonstrate that the system provides accurate and reliable predictions. The model is deployed as a Flaskbased web application, enabling real-time analysis with confidence scores and visual explanations. This work highlights the effectiveness of combining deep learning and explainable AI for kidney stone screening.

Chemical toxicity assessment commonly includes in vivo rat exposure experiments, with transcriptomic measurements collected at various exposure times and chemical doses. The mechanisms underlying chemical-induced toxicity are then inferred by analyzing changes in gene expression. Recently, genome-scale metabolic models (GSMs), which represent the metabolic network of a cell/organism and contain metabolites, reactions, genes, and the relationship between the genes and reactions, have been used to provide a systems-level understanding of gene expression. However, most of the algorithms that integrate gene expression with GSMs require familiarity with MATLAB or Python programming, making them less accessible for users without computational experience. Here, we introduce ToxMet (https://toxmet.bhsai.org), an open-access, user-friendly web application that provides tabular and graph-based network views to visualize the latest rat GSM (iRno v4.2) and predicts chemical-induced metabolic perturbations in rat tissues by integrating toxicogenomic measurements with the rat GSM. ToxMet uses two well-validated computational algorithms, TIMBR and Pheflux, to predict metabolic perturbations and provides the prediction results as interactive and downloadable tables, scatter plots, and network visualizations. As such, the web tool can process a maximum of 10 conditions for a single job, and the results can be used for dose-response studies to monitor organ metabolism at the subsystem level. We evaluated ToxMet's ability to predict toxicity mechanisms by applying it to publicly available toxicogenomic data for two exemplar toxicants: gentamicin and thioacetamide, which are known to induce kidney and liver injury, respectively. ToxMet predicted known toxicity mechanisms for both chemicals, thus demonstrating its ability to provide novel insights into the metabolic mechanisms of chemical-induced toxicity and aid in the discovery of biomarkers and therapeutics using gene expression data.

Kidney stone disease (KSD) is increasingly prevalent among patients with diabetes mellitus and hypertension. Obesity-related metabolic abnormalities may be associated with stone formation, yet their combined association with KSD has not been fully explored. Using data from the National Health and Nutrition Examination Survey (NHANES) 2007-2018, we conducted a cross-sectional study including adults with both diabetes and hypertension. Eight obesity-related composite indices were used as predictors, and self-reported KSD history was defined as the outcome. Nine supervised machine learning algorithms were developed and compared using cross-validation within the training set. Model hyperparameters were tuned using cross-validation within the training set, with the area under the receiver operating characteristic curve (AUC) specified a priori as the primary performance metric. Final model performance was evaluated on an independent test set. To interpret the model, we employed the SHapley Additive exPlanations (SHAP) method, which quantifies both the importance and marginal association of each feature. Subsequently, the final selected model was interpreted using SHAP and deployed as an interactive web application using Shiny. Among the 9 models, the Random Forest classifier demonstrated the best discriminative performance based on cross-validated training performance and achieved an AUC of 0.895 (0.864-0.926) in the independent test set, along with acceptable calibration. Feature importance and SHAP analyses consistently identified the roundness fat mass (RFM) and lipid accumulation product (LAP) as the features most strongly associated with KSD. Obesity-related composite indices were significantly associated with KSD in diabetic-hypertensive adults. The Random Forest model showed superior and discriminative performance among the evaluated algorithms, supporting its potential utility in risk stratification. An interactive web-based Shiny app was developed to enhance clinical applicability: https://obesityrelated.shinyapps.io/apps2/.

A quick, noninvasive method based on 31P NMR spectroscopy to measure the pH of samples containing phosphate buffer is presented. By taking advantage of all four phosphate species populated at acidic, mildly acidic, mildly alkaline and alkaline pH, the method is applicable over a wide range of pH values, from 1 to 13. This tool is most precise at pH values near the pKa values, namely from 5.8 to 8.0 and 10.6 to 13.0. To facilitate its use, a web application is presented that calculates the pH value directly from phosphate 31P chemical shift (ppm) when sodium or potassium is the cation: ( https://rmni.iqf.csic.es/software/31phnmr/ ). In addition, the potential of this approach to monitor reactions which generate or consume H+ is illustrated by following the hydrolysis of GTP catalyzed by Ras-like protein in brain 1a (Rab1a), an essential protein linked to Parkinson's disease and tuberculosis. The intrinsic GTPase rate of Rab1a is found to be 3.3 ± 0.8·10- 3 min at 37°C, which places Rab1a amoung the Rabs with slow GTPase rates and the longer lived "active" GTP-bound states.

The rapid digital transformation of educational institutions necessitates intelligent systems that enhance visitor experience and streamline administrative processes. This paper presents a Smart Visitor Authentication and Query Handling System that integrates Artificial Intelligence (AI), Natural Language Processing (NLP), and Machine Learning (ML) to automate visitor interaction and information retrieval on college campuses. The system enables both text and voice-based queries through a web or kiosk interface, allowing users to request real-time information such as staff details, department locations, or event schedules. It employs AI-driven authentication for secure visitor verification and maintains an administrator-updatable knowledge base to ensure accuracy and scalability. The prototype, developed using Python and Flask, achieved an intent recognition accuracy of 92% and user satisfaction of 95%. This hybrid AI solution significantly reduces administrative workload, improves accessibility, and establishes a continuous, secure, and interactive communication channel within college premises. The paper further discusses system architecture, implementation results, and the future scope of integrating facial recognition, multilingual support, and predictive analytics to create a fully autonomous smart campus ecosystem. Keywords—visitor management system; QR code; campus security; appointment scheduling; real-time communication; MERN stack; digital visitor pass; web application

MOrbVis is an open-source web application that visualizes molecular orbitals directly in the browser without precomputed Gaussian Cube files. Reading only a Molden file, it evaluates Gaussian-type basis functions (s through g shells) on a three-dimensional grid through WebGPU compute shaders. Benchmarks on five devicesfrom a smartphone to a desktop with an NVIDIA RTX 5090show that single-orbital evaluation on grids exceeding 106 points finishes within 100 ms, up to 3 orders of magnitude faster than the single-threaded CPU path. The tool requires no installation and is available at https://yasuaki-ito.github.io/morbvis/.

Laparoscopy is increasingly recognized as a valuable tool in pediatric trauma management, offering the benefits of reduced morbidity compared to traditional open laparotomy. However, its precise role in blunt abdominal trauma in children remains controversial due to concerns about diagnostic accuracy, technical complexity, and patient safety. This systematic review aimed to synthesize current evidence regarding the indications, clinical outcomes, and safety profile of laparoscopic management in pediatric blunt abdominal trauma. A systematic literature search was conducted in PubMed, Web of Science, Lilacs, Scielo, and Scopus databases for studies published between January 2010 and December 2025. Eligible studies included pediatric patients (≤ 18 years old) with blunt abdominal trauma who underwent diagnostic or therapeutic laparoscopy. Screening and data extraction were performed independently by two reviewers using the Rayyan web application, following PRISMA guidelines. A descriptive analysis summarized patient characteristics, injury patterns, operative details, conversions, complications, and mortality. Twenty studies comprising 255 pediatric patients were included, with a median age of 9 years. The available evidence was predominantly derived from case reports and small case series. Conversion from laparoscopy to open surgery occurred in 39.2% of cases, reflecting both the diagnostic role of laparoscopy and intraoperative findings requiring definitive management. The most commonly reported injuries involved the bowel, pancreas, spleen, and liver. Laparoscopy was frequently used as a diagnostic tool and, in selected cases, enabled therapeutic intervention, with generally low reported complication rates and no mortality in the included studies. Median length of hospital stay was 5 days. Laparoscopy appears to be a feasible and potentially valuable adjunct in the management of pediatric blunt abdominal trauma in carefully selected, hemodynamically stable patients. It may contribute to reducing non-therapeutic laparotomies and provide both diagnostic and therapeutic benefits. However, given that current evidence is largely limited to low-level studies, these findings should be interpreted with caution and considered hypothesis-generating. Further prospective and comparative studies are required to better define its role and optimize patient selection.

Background: Confident chemical annotation in nontarget small-molecule mass spectrometry critically depends on the availability of high-quality tandem mass spectral (MS2) reference libraries. While community efforts have driven significant expansion of open-access repositories, technical challenges in assembling standardized, metadata-rich records continue to limit broader participation, underscoring the need for improved computational tools to assist contributors. Methods: To promote the creation and sharing of standardized reference MS2 spectral records, we have developed Librarian, a free, open-access web application designed for rapid and scalable assembly of high-resolution MS2 libraries. Librarian integrates automated retrieval and harmonization of chemical identifiers and metadata from PubChem, compound mixture design for high-resolution mass spectrometry (HRMS) acquisition, and assembly of curated MS2 spectra into repository-ready records compatible with public spectral databases. Results: Through a simple in-browser interface, Librarian offers a flexible end-to-end workflow compatible with popular open-source pre-processing tools to lower technical barriers and facilitate broader community participation in library development. As a demonstration, we used Librarian to create and deposit a spectral library comprising over 1500 new MS2 records into MassBank, which was further applied in retrospective analysis of environmental datasets. Conclusions: Librarian streamlines the creation of standardized, metadata-rich and repository-ready MS2 reference records. Addressing a key bottleneck in community spectral library development and sharing, Librarian supports the continued growth of open-access resources for metabolomics, exposomics, and environmental mass spectrometry. The Librarian web application is publicly accessible via the SciLifeLab Serve platform.

MindGrab: A Spectrally-Motivated Architecture for Accessible Deep Learning in Neuroimaging.

In addition to increasing employment prospects, the quick expansion of internet job portals has led to a surge in fraudulent job ads. This project introduces a Fake Job Detection System based on Artificial Intelligence (AI) that automatically detects bogus job postings. Optical Character Recognition (OCR)-based picture extraction, Uniform Resource Locator (URL), and text input are all accepted by the system. Text preprocessing and feature extraction are done using Natural Language Processing (NLP) methods and Term Frequency–Inverse Document Frequency (TF-IDF). Additionally, fraud indications including money demands, exaggerated salary, and urgent language are identified by a rule-based engine. Long Short-Term Memory (LSTM), Random Forest, and Logistic Regression models are used to analyze the generated data. To increase accuracy and dependability, an ensemble learning technique is used to provide the final forecast. Through a Flask web application, the system gives real-time prediction with an accuracy of about 98%. Additionally, Streamlit is used to deploy the solution on Snowflake for scalable cloud accessibility. This study shows how deep learning (DL), machine learning (ML), and artificial intelligence (AI) may successfully shield people from online recruiting frauds.

Background: Deaf individuals who use sign language (SL) as their primary language often encounter communication barriers in everyday service interactions dominated by spoken or written language. These challenges are particularly evident in street-food settings, where ordering requires rapid and precise communication. This study developed and evaluated a visual-first, multimodal web application to support street-food ordering for Deaf users. Methods: The application integrated food images, structured menu selection, SL videos, text, and audio output to facilitate communication with hearing vendors. Development followed an iterative User-Centered Design (UCD) process involving expert review, pilot testing, and final usability evaluation with 60 Deaf participants. Quantitative and qualitative data were collected to assess usability and user experience. Results: Participants reported high levels of agreement regarding accessibility, visual clarity, and comprehension of SL content. Structured menus, realistic food images, concise SL videos, and audio output were perceived as helpful for constructing and communicating orders. However, lower ratings were observed for learnability and button operation. Qualitative findings further identified challenges related to interaction flow, visibility of interactive elements, and interface complexity. Discussion: The findings suggest that visual-first, multimodal interfaces can improve perceived communication efficiency and support more accessible food-ordering interactions for Deaf users. The study highlights the importance of Deaf-informed, culturally grounded design and demonstrates how iterative UCD processes can identify both usability benefits and design trade-offs. These findings provide practical guidance for the development of accessible communication technologies in everyday service contexts. The application is available at: https://supachan.github.io/street_food_ordering/index.html.

Geographic Information Science (GIScience) offers promising data and tools to innovatively capture and analyze human circadian rhythms across space and time at unprecedented scales, yet its potential remains underexplored. This systematic investigation examines 60 studies published between 2001 and 2024 that applied GIScience to investigate circadian rhythm patterns and their associations with social-environmental factors. Most studies relied on survey or sensor-based data, with limited use of emerging sources such as social media, mobility, web applications, and remote sensing. We found that previous research largely emphasized individual- and population-level monitoring, particularly at the city scale with minute-level temporal resolutions. Traditional statistical approaches dominated, while spatial analysis and advanced spatial-temporal modeling techniques were rarely applied. Findings highlight that both geographic attributes (e.g., latitude, longitude, altitude) and social-environmental conditions (e.g., light, noise, socioeconomic status) exert spatially heterogeneous influences on circadian rhythms, underscoring the need for greater integration of GIScience data and methods into circadian research.