PhosphoRS Algorithm Documentation

June 7, 2026 · View on GitHub

Overview

PhosphoRS (Phosphorylation site Ranking and Scoring) is a comprehensive algorithm for phosphorylation site localization that implements a Compomics-inspired scoring method. It provides detailed site-specific probability calculations and isomer analysis for confident phosphorylation site assignment.

Algorithm Description

Core Principles

  1. Isomer Generation: Generates all possible phosphorylation site combinations
  2. Theoretical Spectrum Matching: Compares experimental spectra with theoretical predictions
  3. Probability Scoring: Calculates site-specific probabilities using statistical models
  4. Confidence Assessment: Provides detailed confidence metrics for each potential site

Mathematical Foundation

PhosphoRS uses a sophisticated statistical approach:

  1. Site-Determining Ions (SDI): Identifies ions that distinguish between different site assignments

  2. Binomial Probability: Uses cumulative binomial probability for scoring:

    P(X ≥ k) = Σ(i=k to n) C(n,i) * p^i * (1-p)^(n-i)

    Where:

    • n = number of theoretical ions
    • k = number of matched ions
    • p = probability of random match

  3. Delta Selection: Optimizes spectrum reduction using site-determining ion differences

  4. Normalization: Converts raw scores to site-specific probabilities

Key Features

  • Fragment Tolerance: 0.05 Da (default)
  • Site-Specific Probabilities: Individual confidence scores for each site
  • Isomer Analysis: Detailed analysis of all possible site combinations
  • Neutral Loss Support: Handles phospho-specific neutral losses
  • Decoy Validation: Optional decoy site analysis

Implementation Details

Parameters

ParameterDefaultDescription
fragment_tolerance0.05Fragment mass tolerance in Da
fragment_method_ppmFalseUse ppm tolerance instead of Da
add_precursor_peakFalseInclude precursor peaks in analysis
add_ion_types("b", "y")Ion types to consider
max_ion_charge2Maximum fragment charge state
add_neutral_lossesTrueInclude neutral losses
add_decoysFalseInclude decoy sites for validation
window_size100.0Window size for spectrum reduction
max_depth8Maximum depth for intensity thresholds
min_depth2Minimum depth for intensity thresholds

Workflow

  1. Site Identification: Identify potential phosphorylation sites (S, T, Y)
  2. Isomer Generation: Create all possible site combinations
  3. Spectrum Filtering: Reduce spectrum using window-based peak selection
  4. SDI Analysis: Identify site-determining ions across isomers
  5. Delta Selection: Optimize spectrum reduction using SDI differences
  6. Theoretical Matching: Generate and match theoretical spectra
  7. Probability Calculation: Calculate site-specific probabilities
  8. Result Assignment: Assign final localization scores

Advanced Features

Site-Determining Ion Analysis

The algorithm identifies ions that are unique to specific site assignments:

def _site_determining_ions(profiles, precursor_charge, add_neutral_losses):
    """
    Identify site-determining ions that distinguish between different
    phosphorylation site assignments.
    """
    # Implementation details...

Delta Selection

Optimizes spectrum reduction by selecting the best depth based on site-determining ion differences:

def _reduce_by_delta_selection(filtered_spec, profiles, fragment_tolerance, fragment_method_ppm):
    """
    Reduce spectrum using delta-based depth selection with site-determining ions.
    """
    # Implementation details...

Binomial Probability Calculation

Uses sophisticated statistical methods for probability calculation:

def binomial_tail_probability(k: int, n: int, p: float) -> float:
    """
    Calculate cumulative binomial probability P(X >= k) for X~Bin(n,p).
    """
    # Implementation details...

Usage Examples

Command Line Interface

# Basic usage
onsite phosphors -in spectra.mzML -id identifications.idparquet -out results.idparquet

# With custom parameters
onsite phosphors -in spectra.mzML -id identifications.idparquet -out results.idparquet \
    --fragment-mass-tolerance 0.05 \
    --fragment-mass-unit Da \
    --threads 1 \
    --add-decoys

Python API

from onsite import calculate_phospho_localization_compomics_style

# Calculate phosphorylation site probabilities
site_probs, isomer_details = calculate_phospho_localization_compomics_style(
    peptide_hit, 
    spectrum,
    fragment_tolerance=0.05,
    fragment_method_ppm=False,
    add_neutral_losses=True
)

Advanced Usage

# Custom configuration
site_probs, isomer_details = calculate_phospho_localization_compomics_style(
    peptide_hit,
    spectrum,
    modification_name="Phospho",
    potential_sites={"S", "T", "Y"},
    fragment_tolerance=0.05,
    fragment_method_ppm=False,
    add_precursor_peak=False,
    add_ion_types=("b", "y"),
    max_ion_charge=2,
    add_neutral_losses=True,
    add_decoys=False
)

# Process results
if site_probs is not None:
    print("Site Probabilities:")
    for site_index, probability in site_probs.items():
        print(f"  Site {site_index}: {probability:.4f}")
    
    print("\nIsomer Details:")
    for seq_str, score in isomer_details:
        print(f"  {seq_str}: {score:.4f}")

Performance Characteristics

Computational Complexity

  • Time Complexity: O(n²) where n is the number of potential sites
  • Space Complexity: O(n) for storing isomers and spectra
  • Optimization: Uses caching for probability calculations

Performance Metrics

  • Processing Speed: ~50-200 PSMs/second (depending on complexity)
  • Memory Usage: ~2-4 GB for typical datasets
  • Accuracy: High accuracy for peptides with clear site-determining ions

Optimization Features

  1. Distribution Caching: Caches binomial probability calculations
  2. Window Reduction: Efficient spectrum filtering
  3. SDI Optimization: Focuses on site-determining ions
  4. Memory Management: Efficient memory usage for large datasets

Output Format

Site Probabilities

Returns a dictionary mapping site indices to probability scores:

{
    0: 85.2,  # Site 0 has 85.2% probability
    3: 14.8,  # Site 3 has 14.8% probability
    7: 0.0    # Site 7 has 0% probability
}

Isomer Details

Returns a list of tuples with sequence and score:

[
    ("PEPTIDE(Phospho)SEQUENCE", 0.95),
    ("PEPTIDESEQUENCE(Phospho)", 0.05)
]

Limitations

  1. Computational Cost: Can be expensive for peptides with many potential sites
  2. Fragment Quality: Requires high-quality MS/MS spectra
  3. Site Ambiguity: May struggle with highly ambiguous localizations
  4. Memory Usage: Can be memory-intensive for large datasets

Troubleshooting

Common Issues

  1. Low Probabilities: Check fragment tolerance and spectrum quality
  2. Memory Errors: Reduce window size or max depth
  3. Poor Localization: Ensure sufficient site-determining ions
  4. Slow Processing: Consider reducing max depth or window size

Optimization Tips

  1. Fragment Tolerance: Use appropriate tolerance for your instrument
  2. Window Size: Adjust based on spectrum complexity
  3. Depth Settings: Balance between accuracy and performance
  4. Neutral Losses: Enable for phospho-specific analysis

References

Original Publication

Taus, T., Köcher, T., Pichler, P., Paschke, C., Schmidt, A., Henrich, C., & Mechtler, K. (2011). Universal and confident phosphorylation site localization using phosphoRS. Journal of Proteome Research, 10(12), 5354-5362.

DOI: 10.1021/pr200611n

Abstract: We present a new approach for confident phosphorylation site localization using phosphoRS, a method that combines the intensity information of site-determining fragment ions with a probability-based scoring scheme. The method is universally applicable to any type of mass spectrometric data and provides confident phosphorylation site localization with high accuracy.

Key Features of Original Implementation

  1. Universal Applicability: Works with any type of MS data
  2. Confident Localization: High accuracy for site assignment
  3. Probability-Based Scoring: Statistical framework for confidence assessment
  4. Site-Determining Ions: Focus on ions that distinguish between sites
  • AScore: Alternative probability-based approach
  • LuciPHOr: FLR-based approach for site localization
  • Mascot Delta Score: Similar concept in database search engines

Implementation Notes

Differences from Original

  1. Compomics Integration: Enhanced with Compomics-style scoring
  2. PyOpenMS Integration: Uses PyOpenMS for spectrum handling
  3. Advanced SDI Analysis: Improved site-determining ion identification
  4. Delta Selection: Optimized spectrum reduction

Future Improvements

  1. Machine Learning: Integration of ML-based scoring
  2. Real-time Processing: Streaming analysis capabilities
  3. Cross-linking: Extension to cross-linked peptides
  4. Quantification: Integration with quantification methods