This video is a compilation of several presentations from a Galaxy community conference, showcasing a wide range of recent advancements and new applications for the platform. The talks cover significant improvements to the user experience, such as a unified workflow run form and more powerful data collection tools that simplify complex data structures. Several presentations explore the integration of cutting-edge technologies, including an AI-powered natural language agent for data analysis, the application of quantum machine learning for complex datasets, and the use of deep learning for creating epigenetic clocks directly in R.
Further topics address the scalability and automation of the Galaxy ecosystem. These include a framework for creating custom, interactive user interfaces for specialized scientific domains (neutron science), automated pipelines for integrating large tool suites like Anvio, and an intelligent job scheduling broker (TPV Broker) for optimizing resource use across a distributed network. The presentations collectively demonstrate the community's focus on enhancing Galaxy's power, accessibility, and adaptability for modern, large-scale bioinformatics research.
A Showcase of Innovations and Applications in the Galaxy Platform
00:00:05 Improving the Workflow Run Form: A presenter named Ahmed details major enhancements to the workflow run form, creating a unified, single-pane interface for analyses. Key improvements include integrating the data uploader and collection creation tools directly into the form, eliminating the need for users to navigate to different parts of Galaxy. This streamlines the setup process, making it more intuitive, especially for new users.
00:09:24 Advanced Data Collection Management: John Chilton presents new features for handling complex data collections. This includes a new collection type that can handle both paired-end and single-end reads within a single workflow, reducing maintenance overhead. He also introduces advanced collection builders and "fetch workbooks" (spreadsheets) that allow users to define and import highly complex, nested data structures with simple tables, a task that previously required a complex rule builder.
00:18:57 An AI-Powered Data Analysis Agent: This talk, presented on behalf of Junha Chu, introduces an interactive data analysis agent inside Galaxy. This tool uses large language models (LLMs) to allow users to analyze datasets, particularly tabular data, using natural language queries. The presentation covers the agent's evolution through three versions, its use of prompt optimization (DSPI) for better performance, and its ability to bridge the gap for flexible data manipulation and visualization.
00:29:15 Integrating Quantum Machine Learning (QML): Brian Robinold discusses the integration of quantum computing into Galaxy. He explains the fundamentals of Qubits and quantum circuits and introduces Q-Sage, a codebase designed to automate QML experiments. The tool evaluates the complexity of a dataset to help researchers determine if quantum machine learning would be advantageous over classical methods for their specific analysis.
00:39:49 Deep-Learned Epigenetic Clocks in R: Jamie Park demonstrates how to use the torch for R package within Galaxy to build deep-learning-based epigenetic clocks. She explains that traditional clocks use linear models (elastic net), while her deep learning approach captures non-linear aging patterns more effectively. Her simple three-layer neural network achieved a mean absolute error of 2.25 years, outperforming existing elastic net-based models.
00:48:20 Glim: A New Generation of ML Tools: This presentation introduces Glim (Galaxy Learning and Modeling), a suite of tools designed to embed best practices and accessibility into machine learning. Glim offers three main tools—Tabular Learner, Image Learner, and Multimodal Learner—built on powerful libraries like PyCaret, Ludwig, and AutoGluon. The goal is to simplify complex ML experiments into a few steps while providing robust, informative reports to eliminate the "black box" nature of ML.
00:57:30 NOVA: Custom UIs for Neutron Science: Greg Watson introduces the Neutron's Open Visualization Analysis (NOVA) framework, designed to build custom, single-page web applications that leverage the Galaxy backend. This addresses the need for specialized, interactive user interfaces for neutron data analysis, which are not feasible with the standard Galaxy UI. The framework uses Python libraries to handle data binding and interaction with Galaxy's infrastructure.
01:08:04 Automating Tool Integration in Galaxy: Fabio Bumb presents two approaches to automate the integration of new tools. The first is a pipeline that successfully wrapped the entire Anvio suite (over 100 tools) by programmatically parsing its argument structure. The second is an experimental AI-driven approach using a large language model (DeepSeeker) trained on existing Galaxy wrappers to generate new XML wrappers from natural language descriptions of a tool.
01:18:30 TPV Broker for Intelligent Job Scheduling: This talk details the TPV Broker, a system designed to improve job scheduling across Galaxy's distributed Pulsar network. Unlike the default scheduler, the TPV Broker is an API that collects real-time metrics (e.g., queue size, running jobs) from all compute endpoints. It uses this information to make smarter decisions, routing jobs to the most optimal location to reduce wait times and improve resource utilization.
01:29:04 Analyzing a Million Heart Cells with Galaxy: PhD student Mairin shows how she used the public EU Galaxy server to analyze a massive combined dataset of over one million single heart cells. Her research, focused on atrial fibrillation, relied on Galaxy's ability to handle large-scale data and its single-cell tools (Scanpy, Harmony) to identify cell clusters and gene expression changes associated with the disease, demonstrating the power of public data and public compute resources.
01:39:39 Gentoics: A Package for Multi-omics Integration: A presenter introduces Gentoics, a Bioconductor R package for integrating multi-omics data at the gene level. The package uses statistical models to analyze how genomic alterations (like copy number variations and methylation) and transcriptional regulators impact gene expression. It includes a comprehensive Shiny app for interactive visualization and exploration of the results.
01:49:15 Integrating R Shiny Apps into Galaxy: This final presentation outlines a standardized framework for integrating interactive R Shiny applications as tools within Galaxy. By using a "Bioconductor R Shiny container" with best practices, developers can wrap their Shiny apps so they can be run from the Galaxy tool panel. This allows for interactive data exploration and visualization as a step within complex analysis workflows, with results being sent back to the user's Galaxy history.