DRHIP: Data Reduction for HyPhy with Inference Processing

A Python toolkit for analyzing HyPhy results. Process and summarize evolutionary selection analysis results across multiple methods and genes.

Installation

You can install DRHIP using one of the following methods:

From PyPI

pip install drhip

From Bioconda

conda install -c bioconda drhip

From Source (Development)

git clone https://github.com/veg/DRHIP.git
cd DRHIP
pip install -e .

Features

Analysis Methods Support

• RELAX: Test for relaxation of selection
• BUSTED: Branch-site unrestricted statistical test for episodic diversification
• CFEL: Contrast-FEL for comparing selection between groups
• FEL: Fixed effects likelihood test for selection
• MEME: Mixed effects model of evolution
• PRIME: Property Informed Models of Evolution

Key Features

• Process HyPhy results for multiple genes in parallel
• Generate comprehensive summary statistics
• Site-specific analysis with conservation tracking
• Thread-safe output handling
• Combined output files across all genes for easy analysis
• Backward compatibility with older HyPhy versions (graceful handling of missing fields)

Usage

# Run analysis on HyPhy results from CAPHEINE workflow
drhip -i /path/to/results/hyphy -o /path/to/output

Input File Structure

The toolkit is designed to work with HyPhy results from the CAPHEINE workflow, which are organized in the following structure:

hyphy/
├── BUSTED/
│   ├── gene1.BUSTED.json
│   └── ...
├── FEL/
│   ├── gene1.FEL.json
│   └── ...
├── MEME/
│   ├── gene1.MEME.json
│   └── ...
├── PRIME/
│   ├── gene1.PRIME.json
│   └── ...
├── RELAX/                  # Optional - presence determines if comparison files are generated
│   ├── gene1.RELAX.json
│   └── ...
└── CONTRASTFEL/           # Optional - presence determines if comparison files are generated
    ├── gene1.CONTRASTFEL.json
    └── ...

Output Files

The toolkit produces combined files that aggregate data across all genes:

• combined_summary.csv: Gene-level summary statistics including:

  • RELAX: K parameter, p-values, and LRT statistics (if RELAX results are available)
  • BUSTED: Evidence of selection, omega distributions
  • Conservation metrics
  • Number of tested branches (N)
  • Total length of tested branches (T)
  • Number of sites (sites)

• combined_sites.csv: Site-specific analysis including:

  • Beta values per comparison group
  • Inferred substitutions at each site
  • Site amino acid composition

• combined_comparison_summary.csv: Comparison group-specific summary statistics (only generated when RELAX or Contrast-FEL results are present):

  • Total number of branches analyzed per group (N)
  • Total branch lengths analyzed per group (T)
  • Group-specific dN/dS ratios
  • Conservation metrics per group

• combined_comparison_site.csv: Comparison group-specific site analysis (only generated when RELAX or Contrast-FEL results are present):

  • Site-specific metrics for each comparison group
  • Inferred substitutions at each site per group
  • Site amino acid composition per group
  • Group-specific selection indicators

The combined files automatically include the superset of all fields found across all genes, with missing values marked as ‘NA’. This makes it easy to analyze patterns across the entire dataset.

Data Representation

The toolkit uses consistent markers to represent different types of data:

• Significant Results: Actual values (p-values, property names, etc.)
• Non-significant Results: “-“ marker for most fields
• Missing or Malformed Data: “NA” marker

If data for a particular method is missing for a gene, or if the data is malformed or cannot be processed, the toolkit will output “NA” in the corresponding fields. This allows for easy identification of missing data versus non-significant results.

Extending the Toolkit

Adding New Fields for Specific Methods

To add new fields to the output for an existing method:

1. Modify the method’s class: Each method has its own class in drhip/methods/ (e.g., busted.py, relax.py).

   • Add the new field to the process_results() method for summary data
   • Add the new field to the process_site_data() method for site-specific data

2. Update validation fields (optional): If the field should be validated as a required field:

   • Add it to SUMMARY_FIELDNAMES or SITES_FIELDNAMES in config.py

Example for adding a new summary field to BUSTED:

def process_results(self, results):
    summary_data = super().process_results(results)

    # Add your new field
    summary_data['new_field_name'] = self._extract_new_field(results)

    return summary_data

def _extract_new_field(self, results):
    # Logic to extract the new field from results
    # Return 'NA' if data is missing or invalid
    try:
        # Extract data from results
        return extracted_value
    except:
        return 'NA'

Field Validation

The toolkit validates that all expected fields are present in the output:

1. Required fields are defined in config.py:

   • SUMMARY_FIELDNAMES: Required fields for summary files
   • SITES_FIELDNAMES: Required fields for site files

2. Field validation occurs in process_gene.py:

   • Missing fields are reported as warnings
   • This ensures consistent output structure across all genes

Adding a New Analysis Method

To add support for a new HyPhy analysis method:

1. Create a new method class in drhip/methods/:

   • Inherit from HyPhyMethod in base.py
   • Implement the required methods: process_results() and optionally process_site_data()

2. Register the method in registry.py:

   • Import your new method class
   • Add it to the __init__ method of HyPhyMethodRegistry

3. Update the method paths in config.py:

   • Add an entry to METHOD_PATHS if your method uses a different directory name

Example for a new method:

# new_method.py
from .base import HyPhyMethod

class NewMethod(HyPhyMethod):
    def __init__(self):
        super().__init__(name='NEW_METHOD', file_suffix='NEW_METHOD.json')

    def process_results(self, results):
        # Process and return summary data
        return {...}

    def process_site_data(self, results):
        # Process and return site-specific data
        return {...}

Then in registry.py:

from .new_method import NewMethod
# ...
self.register(NewMethod())

Comparison Groups Detection

Comparison groups represent different sets of branches in the phylogenetic tree (e.g., ‘foreground’ vs ‘background’, or ‘test’ vs ‘reference’) that are being compared for evolutionary selection differences:

1. What are comparison groups:

   • Groups of branches in the phylogenetic tree labeled for comparative analysis
   • Used by methods like Contrast-FEL and RELAX to compare selection pressures between different lineages
   • Typically labeled as ‘test’/‘reference’, ‘foreground’/‘background’, or custom labels

2. File requirements:

   • RELAX: Files must be named [gene].RELAX.json and placed in the RELAX/ directory (optional)
   • Contrast-FEL: Files must be named [gene].CONTRASTFEL.json (not [gene].CFEL.json) and placed in the CONTRASTFEL/ directory (optional)
   • Note: The presence of either RELAX or Contrast-FEL results will trigger the generation of comparison output files

3. Detection logic is in utils/result_helpers.py:

   • detect_comparison_groups() attempts to extract group labels from CONTRASTFEL and RELAX results files
   • The toolkit examines these results to find branch labels assigned to different groups
   • Falls back to default groups (‘test’/‘reference’) if no explicit labels are detected

4. Adding detection for a new method:

   • Add a detection function in detect_comparison_groups()
   • Add your method to the detection_functions dictionary
   • Add your method to the methods_to_check list

5. Consistency validation:

   • The toolkit ensures that groups are consistent across methods
   • Raises an error if inconsistent groups are detected

6. Method-specific requirements:

   • Some methods (like RELAX) require comparison groups to be set before returning fields
   • The toolkit gracefully handles these requirements with proper error handling
   • If a method can’t provide fields without comparison groups, those fields are skipped

7. CSV combining logic:

   • The toolkit combines CSV files with similar suffixes across all genes
   • Handles various output types: summary, sites, comparison_summary, comparison_site, and comparison
   • Ensures proper field ordering for each output type

Development and Testing

For detailed information about contributing to DRHIP, including development setup, testing guidelines, and code quality standards, please see CONTRIBUTING.md.

Quick Start for Developers

Install the package with development dependencies:

# Install with dev dependencies
pip install -e ".[dev]"

# Install pre-commit hooks for automatic code quality checks
make pre-commit-install

Running Tests

# Run all tests
make test

# Run tests with coverage
make test-cov

# Or use pytest directly
pytest tests/ -v

Code Quality

# Check code quality (runs ruff, black, isort)
make lint

# Auto-format code
make format

Continuous Integration

This project uses GitHub Actions for:

• Automated testing on Python 3.8, 3.9, 3.10, 3.11 (Ubuntu and macOS)
• Code quality checks (linting and formatting)
• Coverage reporting

All tests must pass before merging pull requests.

Test Coverage

The test suite includes:

• Unit tests for all HyPhy analysis methods (BUSTED, FEL, MEME, PRIME, RELAX, CFEL)
• Backward compatibility tests for older HyPhy versions
• Integration tests for gene processing
• Thread safety validation
• Input validation tests
• Minimal actual HyPhy output files for reliable testing

Requirements

• Python 3.7+
• numpy
• pandas
• scipy

License

This project is licensed under the MIT License - see the LICENSE file for details.

Authors

• Sergei L Kosakovsky Pond (spond@temple.edu)
• Hannah Verdonk (hannah.verdonk@temple.edu)
• Danielle Callan (dcallan@temple.edu)