搜索结果：JavaScript

Uchimata is a toolkit for visualization of 3D structures of genomes. It consists of two packages: a Javascript library facilitating the rendering of 3D models of genomes, and a Python widget for visualization in Jupyter Notebooks. Main features include an expressive way to specify visual encodings, and filtering of 3D genome structures based on genomic semantics and spatial aspects. Uchimata is designed to be highly integratable with biological tooling available in Python. Uchimata is released under the MIT License. The Javascript library is available on NPM, while the widget is available as a Python package hosted on PyPI. The source code for both is available publicly on Github (https://github.com/hms-dbmi/uchimata and https://github.com/hms-dbmi/uchimata-py) and Zenodo (https://doi.org/10.5281/zenodo.17831959 and https://doi.org/10.5281/zenodo.17832045). The documentation with examples is hosted at https://hms-dbmi.github.io/uchimata/.

Public domain or access-controlled web apps provide a simple and relatively cheap way to increase the accessibility of information within data resources. Modern data environments, especially those subject to Māori Data Sovereignty (MDSov) constraints, face technical challenges balancing data security and privacy with accessibility and collective benefit. We are exploring how modern web technologies, including open-source tools such as JavaScript, can be combined with the data analytics software R to build modern data-driven web applications with embedded data sovereignty capability. The Rserve software allows JavaScript to communicate with R, and we are developing new technology that enables deployment within diverse data environments in ways that align with MDSov. This approach embeds technical solutions to the challenges of implementing MDSov principles within a data environment, creating data systems that are engineered to enable data owners to control data access, sharing and use.

Champuru is a web-based software tool that helps determine the two sequences present in mixed Sanger chromatograms obtained by simultaneously sequencing two DNA templates of unequal lengths. A previous version (Champuru 1.0) was published as a simple Perl CGI (Common Gateway Interface) application, but the server hosting it was decommissioned, which prompted us to update Champuru and develop it further. The new Champuru 2, implemented in Haxe and hosted at GitHub Pages, offers an improved graphical user interface as well as more sophisticated algorithms to compute alignment scores, making it more efficient at detecting the most likely alignment positions between forward and reverse traces. It also compares the distribution of alignment scores to the theoretical expectation for the comparison of two random sequences and uses this comparison to calculate p-values for the offset pairs it detects. Moreover, Champuru 2 now makes it possible to analyse other offset pairs than the one detected as most likely by the selected algorithm. Champuru 2 is freely accessible at https://eeg-ebe.github.io/Champuru/, including both a graphical user interface (running a JavaScript version transpiled from the Haxe source code) and a compiled command-line version (obtained by transpiling the Haxe source code into C++).

With the expansion of higher education, the uncertainty of students' academic completion and the diversity of academic crises have posed new challenges to the management of higher education. This study aims to design and implement a dynamic academic early warning system based on machine learning to predict and intervene in students' academic crises. By analyzing the causes of academic crisis, the fuzzy comprehensive evaluation with analytic hierarchy process method is used to construct an academic early warning indicator system comprising 10 key indicators. This ensures the scientificity and rationality of the indicator system through expert scoring and consistency test. On this basis, a radial basis function neural network was used to construct an academic early warning model, which outperforms the recurrent neural network and Softmax regression model in terms of prediction accuracy and convergence speed. The system was developed using hypertext markup language, cascading style sheets, JavaScript, and Python to achieve a user-friendly human-computer interaction interface and provide personalized academic alert services. The experimental results demonstrate that the system exhibits high sensitivity and accurate recognition capabilities when dealing with large-scale student datasets, achieving an accuracy rate of 96.32% and a root mean square error of 0.2926, which meets the practical requirements of academic early warning. The results of this study not only provide a new academic early warning tool for colleges, but also have important practical value for promoting the construction of smart campus and digital campus. The current findings, drawn from a sample limited to engineering with an overrepresentation of males, require future validation in multi-disciplinary and gender-balanced cohorts to establish broader applicability.

Recent advances in protein structure prediction have created high-confidence candidate structures for nearly every known protein-coding gene. At the same time, many software packages have been created to visualize protein structures, protein multiple sequence alignments (MSAs), and protein annotations. However, few software tools can highlight the direct relationship between nucleotide variation of protein-coding genes in genome space and the evolutionary and structural context of that variation in protein space. To help address these needs, we created a suite of robust and reusable JavaScript components to show protein structures, MSAs, phylogenies, and their relationship to protein-coding gene regions using the JBrowse 2 genome browser. This software allows users to interface with web services such as AlphaFoldDB and Foldseek to access pre-computed structures, or to upload protein structures from sources such as ColabFold or PDB. Our resources are available at https://github.com/GMOD/proteinbrowser.

To benchmark zero-shot generative pre-trained transformer (GPT)-based multimodal large language models (MLLMs) for pressure injury (PI) staging from photographs and quantify the effects of prompt strategy, structured outputs, and clinically meaningful label granularity. We performed a retrospective observational benchmark using 1,091 public, de-identified PI photographs labeled Stage I-IV. In the standardized analysis, all 10 model/prompt conditions were evaluated using a standardized, resume-safe application programming interface pipeline with per-image logging. We evaluated exact four-class staging, three-class staging (I/II/III-IV), skin-break screening (I vs. II-IV), and an advanced-intervention threshold (I-II vs. III-IV). Reporting followed Strengthening the Reporting of Observational Studies in Epidemiology and artificial intelligence/machine learning guidance; metrics included accuracy, Wilson 95% confidence intervals, F1 scores, weighted kappa, and threshold sensitivity/specificity. All conditions yielded parsed predictions for all 1,091 images. The best accuracy was 93.77% for skin-break screening (GPT-5.2 structured JavaScript Object Notation [JSON]), 90.28% for the advanced-intervention threshold (GPT-5.1 full prompt, low reasoning), 83.78% for three-class staging (GPT-5.1 full prompt, low reasoning), and 65.08% for exact four-class staging (GPT-5.2 structured JSON). The best advanced-intervention condition achieved sensitivity 96.72%, specificity 83.79%, and negative predictive value 96.19%. Stage IV undercalling remained safety-relevant; GPT-5.2-pro had Stage IV recall 8.79% and false-negative rate 91.21%. This prompt-transparent benchmark shows how outcome granularity, prompting, and structured outputs affect GPT-based PI staging and adds ordinal, threshold-based, and safety metrics. GPT-based MLLMs may support clinician-supervised triage and prioritization, but image-only autonomous exact staging is not clinically ready.[Figure: see text].

The rise of harsh weather conditions has made the crop yield vulnerable to fire and variable losses. To address this issue, this article proposes a real-time fire monitoring system which is appropriate for modern smart agriculture. This system utilizes cloud computing technology, Internet of Things (IoT) sensors, telemetry technology, and embedded systems technology to monitor the status in the fields in real time every second. The system has three layers. The first is the IoT device layer, which is composed of flame and smoke sensors, a raspberry pi 3 B+, and a network gateway. The second is the ThingsBoard cloud layer, which is used for efficient processing of large amounts of information. The third is the telemetry layer, which is used for aggregation of the information collected. One of the advantages of the system is the use of a customized aggregation algorithm, which uses sensor information and sends the results in JavaScript object notation (JSON) format using the message queuing telemetry transport (MQTT) protocol. The system was tested in the fields and performed well. It recorded an accuracy of 96.1% in detecting fire in 50 tests, with the rate of false alarms being below 2.8%. It is also clear from the tests that the system can differentiate between true and false alarms. The proposed system sends information to the cloud every two seconds, with an average response time of below 300 milliseconds. The results show that the monitoring system has a reliability rate of over 98%. The results also demonstrate that the Raspberry Pi used in this study had a stable and reasonable central processing unit (CPU) and memory usage rate. Compared to other previously proposed prototypes, the proposed system for monitoring has the advantage of incorporating resource telemetry and fire detection within the same IoT and cloud computing environment, a notable improvement towards sustainable agriculture and food security, as well as the mitigation of agricultural risks.

Implementing target trial emulation (TTE) studies as standardized, reproducible analytic workflows is technically demanding. We developed Text-guided Health-study Estimation and Specification Engine Using Strategus (THESEUS), which uses large language models (LLMs) to translate free-text study descriptions into structured analytic specifications and Strategus R scripts within the Observational Health Data Sciences and Informatics (OHDSI) ecosystem. THESEUS executes 2 steps: an LLM maps study descriptions to a JavaScript Object Notation (JSON) schema, and validated specifications are converted into Strategus R scripts through rule-based logic. For standardization evaluation, we compared specifications generated by 8 LLMs using 15 OHDSI-based TTE studies and 15 non-OHDSI studies under primary-analysis and full-analyses settings. Under the primary-analysis setting, overall standardization accuracy ranged from 0.93 to 0.97 across models in OHDSI studies and from 0.82 to 0.95 in non-OHDSI studies. Gemini-3.1-Pro achieved the highest overall accuracy in OHDSI studies, while Gemini-3.1-Pro and Gpt-5.5 jointly achieved the highest overall accuracy in non-OHDSI studies. Under the full-analyses setting, field-level sensitivity ranged from 0.83 to 0.97 in OHDSI studies, with 0.07-0.80 false positives (FPs) per study, and from 0.77 to 0.89 in non-OHDSI studies, with 0.53-1.20 FPs per study. Gpt-5.5 performed best at the field level. THESEUS was implemented as a web application and coding-agent tools. Pairing a standardized data model with a structured analysis framework enables reliable LLM-assisted interpretation of study descriptions and deterministic workflow construction in observational research. THESEUS supports translation of natural language study descriptions into executable, shareable code in standardized observational research settings.

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.

This research introduces a study of a domain-specific intelligent assistant, TBAid, that is a conversational chatbot to assist with tuberculosis (TB) awareness and health advice. A structured rule-based system integrated with the Hugging Face Inference API using the Qwen/Qwen2.5-72B-Instruct large language model provides TB-focused responses to structured user queries. TBAid is designed to increase public awareness in low-resource and rural areas. It specifically targets communities with poor access to specialist consultations and medical report interpretation. A key novelty of the assistant is its dual-explanation capability, which can frame responses for a non-expert user (e.g., a patient) or provide a medically precise version for healthcare workers. This ensures answers are both accessible and clinically safe by remaining strictly domain-relevant. While the chatbot does not currently analyze images directly, its architecture is designed for future integration. It can accept predictive outputs from a separate, pre-existing CT-based TB classification model. It has a user interface written in HTML, CSS, and JavaScript, and can be deployed on GitHub as a static web app or a local Flask server. This framework enables real-time access and secure decision-making. It is modular, scalable, and can be integrated with AI-based medical diagnostics in the future.•Combines rule-based logic and conversational AI for domain-specific TB support.•Enhances accessibility through lightweight, local, and online deployments.•Supports modular expansion for integration with CT-based diagnostic outputs.

Intracerebral hemorrhage (ICH) remains associated with high mortality and treatment variability. Current workflows rely on fragmented imaging interpretation and operator-dependent surgical planning. The objective was to develop and validate an agentic artificial intelligence (AI) framework integrating automated imaging analysis, guideline-based reasoning, and trajectory optimization for ICH treatment. Fifty consecutive computed tomography (CT) and computed tomography angiography (CTA) datasets from patients with spontaneous ICH were retrospectively analyzed. The system performed multi-class anatomical segmentation of skin, skull, brain, ventricles, and hematoma, followed by volumetric quantification and JavaScript Object Notation (JSON) based structured encoding of imaging biomarkers. A knowledge-based module incorporating international ICH guidelines generated risk stratification and treatment recommendations. When evacuation was indicated, an automated trajectory modeling module proposed a patient-specific minimally invasive surgical corridor. Overall agreement between AI-generated and expert treatment recommendations was 82% (41/50 cases), with substantial agreement beyond chance (Cohen's κ = 0.71). Discrepancies occurred primarily in borderline surgical indication scenarios. In evacuation candidates, the automated planner generated feasible trajectories in all 50 cases. Median angular deviation between AI-generated and expert-defined trajectories was 7.6°, interquartile range (IQR) 5.1-9.8°. AI-generated trajectories demonstrated equal or greater safety margins relative to expert planning in the majority of cases. End-to-end processing has a potential to substantially reduce simulated decision-support time compared with manual workflow. The proposed agentic AI framework enables structured, explainable, and workflow-integrated decision support for ICH management. This system may reduce operator variability and enhance precision in minimally invasive evacuation planning.

Persona prompting is widely used to steer large language models (LLMs), but its effects on safety-critical clinical reasoning are not well characterized. We performed a two-by-two factorial in silico experiment crossing time-pressure framing (high versus low) with optimization target (safety-first versus lean-efficiency). We used 28 Japanese-language synthetic emergency department vignettes covering chest pain, abdominal pain, headache, and dyspnea. Four trap cases contained prespecified contraindication or sequencing rules. Each persona evaluated each vignette twice, yielding 224 independent runs. Outputs followed a fixed JavaScript Object Notation (JSON) schema and were scored for the number of proposed tests, entropy of the probability distribution across the top five differential diagnoses, discharge decisions, safety-net specificity, and contraindication or sequencing violations, with severity grading. High time-pressure framing reduced the number of proposed tests (beta = -1.05, p < 0.001) and diagnostic breadth (beta = -0.246, p < 0.001). Safety-first prompting increased proposed testing (beta = 1.32, p < 0.001) and diagnostic breadth (beta = 0.247, p < 0.001), with no significant interaction. Among discharge plans (36 of 224 runs), safety-first prompting improved safety-net specificity (mean 4.5 versus 2.6 on a five-point scale). Contraindication or sequencing violations occurred only in the high/lean condition (eight of 56 runs, 14.3%); in trap cases, violations were eight of eight under high/lean and zero of 24 in the other three conditions. Persona components predictably shifted simulated clinical reasoning. Time-pressure framing narrowed diagnostic search and reduced proposed testing, whereas safety-first prompting improved safety-netting and prevented severe trap-case violations outside the high/lean condition. Prompt-aware stress testing may help identify unsafe prompt configurations before clinical deployment.

Here, we describe a collection of genomic database portals, SoyBase (https://soybase.org), Legume Information System (https://legumeinfo.org), and PeanutBase (https://peanutbase.org), that support breeding and research work in the legume plant family. The legume family includes important crops such as soybean, peanut, common bean, lentils, chickpeas, as well as approximately 20,000 other species that are important in all terrestrial ecosystems. Beyond the value of the portals for species in this large clade (as well as for plant biology more generally), the database and site architecture of these portals will be of interest to developers of similar genomic sites, as the data management and software solutions are generic and should be applicable to a wide variety of organisms. The architecture for these sites has been designed for rapid, modular, flexible development well suited to genomic data and to rapid incorporation of new data. Website content is handled with a static site generator (Jekyll). Interactive applications are developed using javascript encapsulated as Web Components that access back-end data via APIs for stability and flexibility. This architecture allows for both code portability and for customization to serve the unique needs of each research community.

Subjective comparisons of aspects of experience provides reliable and powerful numerical data, which can provide us means to characterize structures of consciousness. Yet, an exhaustive set of comparative and pairwise judgements among N stimuli requires N2 trials, which is costly for in-lab face-to-face data collection from a participant. By utilizing an online experimental platform, it is easy to recruit many participants, randomly distributing a small proportion of all possible pairs. However, random assignment is not efficient in obtaining data uniformly across all pairs of stimuli. Here, we introduce a new method for minimizing variance in trial counts across stimulus pairs, by integrating PsychoPy with GitHub Gist, which records the frequency with which each pair has been presented. We provide JavaScript code that can be incorporated into customized code chunks in PsychoPy. The program can be run on Pavlovia for online participants, and we show the effectiveness of our method. &#x2022; The frequencies that each stimulus pair has been shown are stored on GitHub Gist. &#x2022; When a new participant starts on Pavlovia, our methodology reads the frequencies, selects the least presented stimuli for the participant, and updates the frequencies. &#x2022; The frequencies get dynamically balanced for efficient data collection.

Across biomedical research and care, many conversations transmit information with profound practical, ethical, and legal consequences. The process of informed consent, where individuals decide to join a study or accept clinical care, is perhaps the most consequential, yet it is also complex, labor-intensive, and variable across sites. Existing platforms for information transmission in the informed consent context largely reproduce static documents and lack reproducibility or auditability, while generative chatbots offer flexibility at the cost of stochasticity, hallucination, and regulatory risk. We present Kauro, an open-source, graph-based chatbot that encodes scripted conversations as version-controlled JavaScript Object Notation (JSON) structures, enabling deterministic traversal (ie, paths through the graph), complete audit logging, and IRB-verifiable oversight. Its modular separation of client, server, and script ensures portability across institutions. By operationalizing constraint rather than flexibility, Kauro reframes deployment of machine intelligence in biomedical communication with reproducibility and auditability, offering a scalable platform generalizable to any domain where conversations demand safety, precision, and trust.

The secondary use of clinical data has been widely studied, addressing many challenges in conducting observational studies. However, the complexity of dataset structures and detailed data requirements has led researchers to develop user-friendly query builders, enabling medical researchers to define cohorts more efficiently across the datasets. Although these tools simplify the workflow, their learning curve can sometimes be steep. Motivated by improving the usability to conducted observational studies, we propose a modular conversational assistant framework that would address these limitations. It can be integrated in any web application as an javascript component, improving the system usability. Additionally, the proposed framework would employ deterministic algorithms to reduce computational overhead. The system would enable integration into existing medical information systems through configuration files rather than code modifications. Validation within the OHDSI ecosystem would demonstrate practical applicability for real-world observational research scenarios.

High-stakes licensing exams such as the United States Medical Licensing Examination (USMLE) play a critical role in medical education, influencing both trainee progression and patient outcomes. Access to high-quality board preparation resources is uneven and often cost-prohibitive, disproportionately affecting students from underrepresented or financially disadvantaged backgrounds. An artificial intelligence (AI)-driven system to generate USMLE-style practice questions aligned with National Board of Medical Examiners (NBME) item-writing guidelines using a Large Language Model (LLM) enhanced with retrieval augmented generation, chain-of-thought and few-shot prompting, and JavaScript Object Notation schema validation was developed and piloted at the University of Cincinnati College of Medicine between November and December 2023. Five lectures from a preclinical hematology course were selected, and 565 questions were generated for 177 first-year medical students. A human-in-the-loop process, led by a faculty course director, ensured content validity and adherence to educational standards. Validated questions were deployed via a mobile app, allowing students to practice, receive performance feedback, and access an AI tutor. Of the 565 questions, 490 (87%) were deemed accurate and NBME-compliant. Eighty students used the question bank, completing up to 220 questions each. Although not statistically significant, increased use trended toward improved performance on related exam questions. Qualitative feedback highlighted enthusiasm for AI-assisted study tools, with calls for broader content coverage. This pilot demonstrates that LLMs can generate high-quality, guideline-aligned practice questions. To improve scalability and reduce faculty workload, future iterations will incorporate AI-based review agents for pre-screening content. The platform is intended to be expanded to additional courses, training phases, and health professions. Ongoing refinement will focus on improving content specificity and maintaining accuracy, especially in advanced and subspecialty education.