torchseed

A modern graphical interface for configuring and generating JSON configuration files for torchtree using Streamlit.

torchseed provides an interactive workflow inspired by tools such as BEAUTI (BEAST), allowing users to build phylogenetic analyses visually without manually editing JSON configuration files.

---

Features

• Interactive GUI for building torchtree analyses
• Automatic JSON configuration generation
• Dynamic parameter panels that update according to selected models
• Support for common phylogenetic components:
  • Site models: Discretized Weibull, proportion of invariant sites
  • Substitution models
    • Nucleotide: JC69, HKY, K80, GTR
    • Protein: LG, WAG
  • Clock models: Strict, one rate per branch
  • Tree priors
    • Unrooted tree: Exponential distribution
    • Coalescent: constant, exponential, skyride, skygrid, skyglide
  • Parametric priors: editable
  • Variational inference configuration
• Real-time parameter validation
• Interactive visualisations with Plotly
• Table and summary displays using Pandas
• Extensible architecture for adding new models and plugins
• Streamlit-based responsive interface

---

Technology Stack

Built with:

• Streamlit
• NumPy
• SciPy
• Pandas
• Plotly

---

Installation

Clone the repository

git clone https://github.com/4ment/torchseed.git
cd torchseed

Create the Conda environment

conda env create -f environment.yml
conda activate torchseed

---

Running the Application

Start the Streamlit app with:

streamlit run torchseed/app.py

The application will automatically open in your browser.

---

Example Workflow

1. Import sequence alignments
2. Select and configure tip dates
3. Import tree
4. Select a site and substitution models
5. Select clock models if needed
6. Specify priors
7. Configure ADVI
8. Preview the generated JSON configuration
9. Export the configuration file
10. Run the analysis using torchtree

---

Project Structure

torchseed/
│
├──torchseed/
│  ├── app.py                  # Streamlit entry point
│  ├── tabs/                   # Multi-page application views
│  └── builders/               # Build JSON components
│
├── requirements.txt
├── environment.yml
└── README.md

---

Roadmap

• Alignment import wizard
• Tree visualisation
• Session save/load support
• Prior distribution visualisation
• Plugin architecture
• YAML export support
• Analysis templates
• Direct integration with torchtree execution

---

Contributing

Contributions and feature requests are welcome.

1. Fork the repository
2. Create a feature branch
3. Commit your changes
4. Open a pull request

---

License

Distributed under the GPLv3 License. See LICENSE for more information.

---

Acknowledgements

• torchtree
• BEAST 2
• Streamlit