GLEAM: Galaxy Learning and Modeling

GLEAM (Galaxy Learning and Modeling) is a suite of machine learning tools for the Galaxy platform. Developed by the Goecks Lab, GLEAM empowers researchers to train models, generate predictions, and produce reproducible reports—all from a user-friendly interface without writing code.

Features

• Modern best practices for machine learning
• Reproducible and scalable workflows
• Machine learning support for diverse data types: tabular, image, text, categorical, and more
• Deep learning via Ludwig and automated ML via PyCaret
• Easy installation in Galaxy via XML wrappers
• Auto-generated visual reports

Available Tools

1. TabularLearner

Machine learning for structured tabular datasets using PyCaret.

• Train classification and regression models
• Evaluate performance and extract feature importance
• Generate predictions on new datasets
• Create interactive HTML reports

2. ImageLearner

Deep learning-based image classification using Ludwig.

• input files: Zip file with images and csv with metadata
• Tasks: classification
• Models available: ResNet, EfficientNet, VGG, Shufflenet, Vit, AlexNet and More...
• Output: Ludwig_model file, a report in the form of an HTML file (with learning curves, confusion matrices, and etc...), and a collection of CSV/json/png files containing the predictions, experiment stats and visualizations.

3. Galaxy-Ludwig

General-purpose interface to Ludwig's full machine learning capabilities.

• Train and evaluate models on structured input (tabular, image, text, etc.)
• Expose Ludwig’s flexible configuration system
• Ideal for users needing advanced model customization

4. Galaxy-Digital Pathology Processing

Set of three specialized tools designed to transforms raw, large pathology images into a structured format, enabling the application of best practices for model development and ensuring data readiness for robust and efficient training.

• Image Tiler: Accepts .svs image format, which is the most common proprietary format for digital pathology whole slide images.
• Embedding Extractor: Leverages pre-trained models from the TorchVision foundation models for feature extraction (for example, ResNet50, EfficientNet_B0, DenseNet121).
• Multiple Instance Learning (MIL) Bag Processor: Facilitates the aggregation of embeddings from individual image tiles into "bags" using various pooling techniques (such as Max Pooling or Attention Pooling).

Installation

1. Clone the repository:

   git clone https://github.com/goeckslab/gleam.git

2. Add entries for each tool in your tool_conf.xml of your galaxy instance:

   <tool file="<path-to-your-local-tabularlearner/tabular_learner.xml>" />
   <tool file="<path-to-your-local-imagelearner/image_learner_train.xml>" />
   <tool file="<path-to-your-local-galaxy-ludwig/ludwig_train.xml>" />

Contributing

We welcome contributions. To propose new tools, report bugs, or suggest improvements:

1. Fork the repository

2. Create a feature branch

3. Commit and test your changes

4. Submit a pull request