Protein Interaction Network Analysis

Comprehensive protein interaction network analysis using ToolUniverse tools. Analyzes protein networks through a 4-phase workflow: identifier mapping, network retrieval, enrichment analysis, and optional structural data.

Features

✅ Identifier Mapping - Convert protein names to database IDs (STRING, UniProt, Ensembl) ✅ Network Retrieval - Get interaction networks with confidence scores (0-1.0) ✅ Functional Enrichment - GO terms, KEGG pathways, Reactome pathways ✅ PPI Enrichment - Test if proteins form functional modules ✅ Structural Data - Optional SAXS/SANS solution structures (SASBDB) ✅ Fallback Strategy - STRING primary (no API key) → BioGRID secondary (if key available)

Databases Used

| Database | Coverage | API Key | Purpose | |----------|----------|---------|---------| | STRING | 14M+ proteins, 5,000+ organisms | ❌ Not required | Primary interaction source | | BioGRID | 2.3M+ interactions, 80+ organisms | ✅ Required | Fallback, curated data | | SASBDB | 2,000+ SAXS/SANS entries | ❌ Not required | Solution structures |

Quick Start

Basic Usage

python
from tooluniverse import ToolUniverse
from python_implementation import analyze_protein_network

# Initialize ToolUniverse
tu = ToolUniverse()

# Analyze protein network
result = analyze_protein_network(
    tu=tu,
    proteins=["TP53", "MDM2", "ATM", "CHEK2"],
    species=9606,  # Human
    confidence_score=0.7  # High confidence
)

# Access results
print(f"Mapped: {len(result.mapped_proteins)} proteins")
print(f"Network: {result.total_interactions} interactions")
print(f"Enrichment: {len(result.enriched_terms)} GO terms")
print(f"PPI p-value: {result.ppi_enrichment.get('p_value', 1.0):.2e}")

Expected Output

🔍 Phase 1: Mapping 4 protein identifiers...
✅ Mapped 4/4 proteins (100.0%)

🕸️  Phase 2: Retrieving interaction network...
✅ STRING: Retrieved 6 interactions

🧬 Phase 3: Performing enrichment analysis...
✅ Found 245 enriched GO terms (FDR < 0.05)
✅ PPI enrichment significant (p=3.45e-05)

✅ Analysis complete!

Use Cases

1. Single Protein Analysis

Discover interaction partners for a protein of interest:

python
result = analyze_protein_network(
    tu=tu,
    proteins=["TP53"],  # Single protein
    species=9606,
    confidence_score=0.7
)

# Top 5 partners will be in the network
for edge in result.network_edges[:5]:
    print(f"{edge['preferredName_A']} ↔ {edge['preferredName_B']} "
          f"(score: {edge['score']})")

2. Protein Complex Validation

Test if proteins form a functional complex:

python
# DNA damage response proteins
proteins = ["TP53", "ATM", "CHEK2", "BRCA1", "BRCA2"]

result = analyze_protein_network(tu=tu, proteins=proteins)

# Check PPI enrichment
if result.ppi_enrichment.get("p_value", 1.0) < 0.05:
    print("✅ Proteins form functional module!")
    print(f"   Expected edges: {result.ppi_enrichment['expected_number_of_edges']:.1f}")
    print(f"   Observed edges: {result.ppi_enrichment['number_of_edges']}")
else:
    print("⚠️  Proteins may be unrelated")

3. Pathway Discovery

Find enriched pathways for a protein set:

python
result = analyze_protein_network(
    tu=tu,
    proteins=["MAPK1", "MAPK3", "RAF1", "MAP2K1"],  # MAPK pathway
    confidence_score=0.7
)

# Show top enriched processes
print("\nTop Enriched Pathways:")
for term in result.enriched_terms[:10]:
    print(f"  {term['term']}: p={term['p_value']:.2e}, FDR={term['fdr']:.2e}")

4. Multi-Protein Network Analysis

Build complete interaction network for multiple proteins:

python
# Apoptosis regulators
proteins = ["TP53", "BCL2", "BAX", "CASP3", "CASP9"]

result = analyze_protein_network(
    tu=tu,
    proteins=proteins,
    confidence_score=0.7
)

# Export network for Cytoscape
import pandas as pd
df = pd.DataFrame(result.network_edges)
df.to_csv("apoptosis_network.tsv", sep="\t", index=False)

5. With BioGRID Validation

Use BioGRID for experimentally validated interactions:

python
# Requires BIOGRID_API_KEY in environment
result = analyze_protein_network(
    tu=tu,
    proteins=["TP53", "MDM2"],
    include_biogrid=True  # Enable BioGRID fallback
)

print(f"Primary source: {result.primary_source}")  # "STRING" or "BioGRID"

6. Including Structural Data

Add SAXS/SANS solution structures:

python
result = analyze_protein_network(
    tu=tu,
    proteins=["TP53"],
    include_structure=True  # Query SASBDB
)

if result.structural_data:
    print(f"\nFound {len(result.structural_data)} SAXS/SANS entries:")
    for entry in result.structural_data:
        print(f"  {entry.get('sasbdb_id')}: {entry.get('title')}")

Parameters

analyze_protein_network() Parameters

| Parameter | Type | Default | Description | |-----------|------|---------|-------------| | tu | ToolUniverse | Required | ToolUniverse instance | | proteins | liststr] | Required | Protein identifiers (gene symbols, UniProt IDs) | | species | int | 9606 | NCBI taxonomy ID (9606=human, 10090=mouse) | | confidence_score | float | 0.7 | Min interaction confidence (0-1). 0.4=low, 0.7=high, 0.9=very high | | include_biogrid | bool | False | Use BioGRID if STRING fails (requires API key) | | include_structure | bool | False | Include SASBDB structural data (slower) | | suppress_warnings | bool | True | Suppress ToolUniverse loading warnings |

Species IDs (Common)

• 9606 - Homo sapiens (human)
• 10090 - Mus musculus (mouse)
• 10116 - Rattus norvegicus (rat)
• 7227 - Drosophila melanogaster (fruit fly)
• 6239 - Caenorhabditis elegans (worm)
• 7955 - Danio rerio (zebrafish)
• 559292 - Saccharomyces cerevisiae (yeast)

Confidence Score Guidelines

| Score | Level | Description | Use Case | |-------|-------|-------------|----------| | 0.15 | Very low | All evidence | Exploratory, hypothesis generation | | 0.4 | Low | Medium evidence | Default STRING threshold | | 0.7 | High | Strong evidence | Recommended - reliable interactions | | 0.9 | Very high | Strongest evidence | Core interactions only |

Results Structure

ProteinNetworkResult Object

python
@dataclass
class ProteinNetworkResult:
    # Phase 1: Identifier mapping
    mapped_proteins: List[Dict[str, Any]]
    mapping_success_rate: float

    # Phase 2: Network retrieval
    network_edges: List[Dict[str, Any]]
    total_interactions: int

    # Phase 3: Enrichment analysis
    enriched_terms: List[Dict[str, Any]]
    ppi_enrichment: Dict[str, Any]

    # Phase 4: Structural data (optional)
    structural_data: Optional[List[Dict[str, Any]]]

    # Metadata
    primary_source: str  # "STRING" or "BioGRID"
    warnings: List[str]

Network Edge Format (STRING)

python
{
    "stringId_A": "9606.ENSP00000269305",  # Protein A STRING ID
    "stringId_B": "9606.ENSP00000258149",  # Protein B STRING ID
    "preferredName_A": "TP53",             # Protein A name
    "preferredName_B": "MDM2",             # Protein B name
    "ncbiTaxonId": 9606,                   # Species
    "score": 0.999,                        # Combined confidence (0-1)
    "nscore": 0.0,                         # Neighborhood score
    "fscore": 0.0,                         # Gene fusion score
    "pscore": 0.0,                         # Phylogenetic profile score
    "ascore": 0.947,                       # Coexpression score
    "escore": 0.951,                       # Experimental score
    "dscore": 0.9,                         # Database score
    "tscore": 0.994                        # Text mining score
}

Enrichment Term Format

python
{
    "category": "Process",                  # GO category
    "term": "GO:0006915",                   # GO term ID
    "description": "apoptotic process",     # Term description
    "number_of_genes": 4,                   # Genes in your set
    "number_of_genes_in_background": 1234, # Genes in genome
    "p_value": 1.23e-05,                    # Enrichment p-value
    "fdr": 0.0012,                          # FDR correction
    "inputGenes": "TP53,MDM2,BAX,CASP3"    # Matching genes
}

Workflow Details

4-Phase Analysis Pipeline

┌─────────────────────────────────────────────────────────────┐
│ Phase 1: Identifier Mapping                                 │
│ ─────────────────────────────────────────────────────────── │
│ STRING_map_identifiers()                                    │
│   • Validates protein names exist in database              │
│   • Converts to STRING IDs for consistency                 │
│   • Returns mapping success rate                           │
└─────────────────────────────────────────────────────────────┘
                            ↓
┌─────────────────────────────────────────────────────────────┐
│ Phase 2: Network Retrieval                                  │
│ ─────────────────────────────────────────────────────────── │
│ PRIMARY: STRING_get_network() (no API key needed)          │
│   • Retrieves all pairwise interactions                    │
│   • Returns confidence scores by evidence type             │
│                                                             │
│ FALLBACK: BioGRID_get_interactions() (if enabled)          │
│   • Used if STRING fails or for validation                 │
│   • Requires BIOGRID_API_KEY                               │
└─────────────────────────────────────────────────────────────┘
                            ↓
┌─────────────────────────────────────────────────────────────┐
│ Phase 3: Enrichment Analysis                                │
│ ─────────────────────────────────────────────────────────── │
│ STRING_functional_enrichment()                              │
│   • GO terms (Process, Component, Function)                │
│   • KEGG pathways                                           │
│   • Reactome pathways                                       │
│   • FDR-corrected p-values                                  │
│                                                             │
│ STRING_ppi_enrichment()                                     │
│   • Tests if proteins interact more than random            │
│   • Returns p-value for functional coherence               │
└─────────────────────────────────────────────────────────────┘
                            ↓
┌─────────────────────────────────────────────────────────────┐
│ Phase 4: Structural Data (Optional)                         │
│ ─────────────────────────────────────────────────────────── │
│ SASBDB_search_entries()                                     │
│   • SAXS/SANS solution structures                           │
│   • Protein flexibility and conformations                   │
│   • Complements crystal/cryo-EM data                       │
└─────────────────────────────────────────────────────────────┘

Installation & Setup

Prerequisites

bash
# Install ToolUniverse (if not already installed)
pip install tooluniverse

# Or with extras
pip install tooluniverse[all]

Optional: BioGRID API Key

For BioGRID fallback functionality:

1. Register for free API key: https://webservice.thebiogrid.org/
2. Add to .env file:

bash BIOGRID_API_KEY=your_key_here

Skill Files

tooluniverse-protein-interactions/
├── SKILL.md                    # This file
├── python_implementation.py    # Main implementation
├── QUICK_START.md             # Quick reference
├── DOMAIN_ANALYSIS.md         # Design rationale
├── PHASE2_COMPLETE.md         # Tool testing results
├── PHASE4_IMPLEMENTATION_COMPLETE.md
└── KNOWN_ISSUES.md            # ToolUniverse limitations

Known Limitations

1. ToolUniverse Verbose Output

Issue: ToolUniverse prints 40+ warning messages during analysis.

Workaround: Filter output when running:

bash
python your_script.py 2>&1 | grep -v "Error loading tools"

See KNOWN_ISSUES.md for details.

2. BioGRID Requires API Key

BioGRID fallback requires free API key. STRING works without any API key.

3. SASBDB May Have API Issues

SASBDB endpoints occasionally return errors. Structural data is optional.

Performance

Typical Execution Times

| Operation | Time | Notes | |-----------|------|-------| | Identifier mapping | 1-2 sec | For 5 proteins | | Network retrieval | 2-3 sec | Depends on network size | | Enrichment analysis | 3-5 sec | For 374 terms | | Full 4-phase analysis | 6-10 sec | Excluding ToolUniverse overhead |

Note: Add 4-8 seconds per tool call for ToolUniverse loading (framework limitation).

Optimization Tips

1. Disable structural data if not needed: include_structure=False
2. Use higher confidence scores to reduce network size: confidence_score=0.9
3. Filter output to avoid processing warning messages
4. Reuse ToolUniverse instance across multiple analyses

Troubleshooting

"Error: 'protein_ids' is a required property"

✅ Fixed in this skill - All parameter names verified in Phase 2 testing.

No interactions found

• Check protein names are correct (case-sensitive)
• Try lower confidence score: confidence_score=0.4
• Verify species ID is correct
• Check if proteins actually interact (not all proteins have known interactions)

BioGRID not working

• Ensure BIOGRID_API_KEY is set in environment
• Check API key is valid at https://webservice.thebiogrid.org/
• BioGRID is optional - STRING works without it

Slow performance

• This is expected (see KNOWN_ISSUES.md)
• ToolUniverse framework reloads tools on every call
• Use output filtering to reduce processing time

Examples

See python_implementation.py for:

• example_tp53_analysis() - Complete TP53 network analysis
• analyze_protein_network() - Main function with all options
• ProteinNetworkResult - Result data structure

References

• STRING: https://string-db.org/ (14M+ proteins, 5,000+ organisms)
• BioGRID: https://thebiogrid.org/ (2.3M+ interactions, experimentally validated)
• SASBDB: https://www.sasbdb.org/ (2,000+ SAXS/SANS entries)
• ToolUniverse: https://github.com/mims-harvard/ToolUniverse

Support

For issues with:

• This skill: Check KNOWN_ISSUES.md and troubleshooting section
• ToolUniverse framework: See TOOLUNIVERSE_BUG_REPORT.md
• API errors: Check database status pages (STRING, BioGRID, SASBDB)

License

Same as ToolUniverse framework license.