A tool for generating synthetic data from dbGaP variable reports.

# Clone the repository
git clone https://github.com/yourusername/synTOPia-data-gen.git
cd synTOPia-data-gen

# Create and activate a virtual environment
python -m venv .venv
source .venv/bin/activate  # On Windows, use `.venv\Scripts\activate`

# Install dependencies
pip install -e .

The tool provides a command-line interface for parsing dbGaP variable reports and generating synthetic data.

You can parse variable reports from either a local XML file or directly from a dbGaP URL:

# Parse from a local XML file
syntopia parse path/to/your/variable_report.xml

# Parse from a dbGaP URL
syntopia parse https://ftp.ncbi.nlm.nih.gov/dbgap/studies/phs001402/phs001402.v3.p1/pheno_variable_summaries/phs001402.v3.pht008239.v1.p1.TOPMed_WGS_Mayo_VTE_Subject_Phenotypes.var_report.xml

The parsed schema will be saved as a YAML file in the syntopia/data/topmed/schemas directory.

After parsing the variable report, you can generate synthetic data:

# Generate 1000 synthetic records
syntopia generate syntopia/data/topmed/schemas/your_schema.yaml --n 1000 --output syntopia/data/topmed/synthetic

Both commands support additional options:

# Enable debug logging
syntopia parse your_file.xml --debug
syntopia generate your_schema.yaml --debug

# Set a random seed for reproducibility
syntopia generate your_schema.yaml --seed 42

# Apply custom rules during generation
syntopia generate your_schema.yaml --rules path/to/rules.yaml

syntopia/
├── data/
│   └── topmed/
│       ├── raw/          # Raw XML files
│       ├── schemas/      # Parsed YAML schemas
│       └── synthetic/    # Generated synthetic data
├── parsing/             # XML parsing and schema generation
├── synth/              # Synthetic data generation
└── cli.py             # Command-line interface

SynTOPia includes a Model Context Protocol (MCP) server that enables integration with Claude Desktop for natural language interaction with synthetic data generation capabilities.

1. Install dependencies:

   pip install -e .

2. Test the MCP server:

   python test_mcp_server.py

3. Configure Claude Desktop: Add this to your Claude Desktop configuration file:

   {
     "mcpServers": {
       "syntopia": {
         "command": "python",
         "args": ["-m", "syntopia.mcp_server"],
         "cwd": "/path/to/your/syntopia-data-gen"
       }
     }
   }

4. Restart Claude Desktop and start using natural language commands like:

   • "Parse this dbGaP variable report: [URL]"
   • "Generate 500 synthetic samples from the schema file"
   • "Show me available schemas"
   • "Preview the synthetic data file"

For detailed MCP server documentation, see MCP_SERVER_README.md.

To run tests:

pytest

MIT License