Galaxy Training Material

A production-proven monitoring stack for Galaxy infrastructure consists of the following components:

• Telegraf: Used for base monitoring. It is easy to extend by adding scripts or commands to your servers that generate output in the InfluxDB line protocol format.
• InfluxDB: The time-series database used to store monitoring data.
• Grafana: The visualization layer. It allows you to import dashboards developed by UseGalaxy.* and extend them for your own specific needs.
• Galaxy Job Radar: An experimental solution currently under development that provides an overview of Galaxy job traffic globally.

The gtn-downloader is a Python script designed to automate the retrieval and organization of training data from the Galaxy Training Network (GTN).

Key Functionalities:

• Automated Retrieval: Scans the GTN repository for data-library.yaml files and downloads the associated datasets.
• Structured Organization: Automatically creates a directory hierarchy based on tutorial names and the metadata found in data-library.yaml.
• Reporting: Generates a download-summary.tsv file containing metadata, a download report (success, error, or already downloaded), and total file sizes.

The ToolFactory is an automated, form-driven tool generator designed for scientists and developers who write command-line scripts (e.g., R, Bash, Python). It allows users to quickly wrap working scripts into Galaxy tools by providing information through a familiar Galaxy UI form.

Key capabilities include:

• Automated XML Generation: Uses galaxyml to generate tool XML from form settings.
• Instant Feedback Loop: Generated tools are immediately installed in the appliance, allowing for immediate testing and exploration.
• Iterative Development: Tools can be updated by re-running the generation job and modifying the form settings (e.g., updating labels, help text, or parameters).
• Toolshed Readiness: Produces tools as Toolshed-ready archives, which can be exported for sharing after adding a test.

Sentry is an error tracking software used by admins and developers to monitor and diagnose application issues. It provides real-time alerts and captures context for errors, such as stack traces and user feedback. In the context of Galaxy and Pulsar, Sentry can be used to log issues and failing tool runs, allowing administrators to identify and fix errors before they are reported by users.

This tutorial covers:

• Installing Sentry (locally via docker-compose or via the hosted service at https://sentry.io/).
• Configuring Galaxy and Pulsar to send errors to Sentry.
• Monitoring performance using Sentry.

The DOME (Data, Optimization, Model, and Evaluation) recommendations provide a community-driven consensus list of minimal requirements for reporting supervised machine learning (ML) analyses in biological research. Developed by the ELIXIR Machine Learning Focus Group, these guidelines aim to enhance transparency, reproducibility, and the ability to critically evaluate ML-based studies where direct experimental validation may be absent.

The framework focuses on four core components:

1. Data: Ensuring data integrity, provenance, and proper partitioning.
2. Optimization: Documenting algorithms, parameters, and tuning.
3. Model: Describing model architecture and interpretability.
4. Evaluation: Employing rigorous performance measures and statistical comparisons.

These recommendations are primarily aimed at supervised learning in biological applications.

The Galaxy interface is divided into three primary functional areas:

• Tools: Contains the bioinformatics tool packages available for analysis.
• Main Dashboard: The central interface area used for interacting with data and viewing system messages.
• History: Displays the documents and datasets associated with your current project.

Note on History Management: Galaxy histories can become very busy. Because Galaxy's default naming conventions are not always intuitive, it is a best practice to manually label your files with descriptive names to keep your project organized.

The workflow produces several key datasets required for ENA submission:

1. Raw FASTQ files: Converted from input AB1 files.
2. Cleaned FASTQ files: Renamed and filtered files ready for submission.
3. FASTA files: Converted from FASTQ for BLAST alignments.
4. Quality Control Results: Falco and MultiQC results (both raw statistics and HTML reports) generated before and after cleaning.
5. Blast Results:
   • Raw Blast Results: Alignment results containing detailed metrics.
   • Filtered Blast Results: Files containing only the closest homologous sequences.
6. Filtered Fastq files: The final collection of sequences used for submission.