Boltz-2 Analysis & Reporting

Advanced analysis and visualization tools for Boltz-2 structure predictions, providing comprehensive quality assessment and confidence metrics for predicted protein structures. The Boltz-2 Report automatically evaluates multiple prediction models using combined pLDDT and pTM scoring, classifying structures from excellent (above 0.9) to poor (below 0.5) confidence levels. This interactive dashboard enables researchers to compare model variants, identify high-confidence regions for downstream applications, and select optimal structures for docking studies or experimental validation, ensuring reliable structural foundations for drug discovery workflows.

Key Features

  • Confidence Assessment: Overall prediction quality using pLDDT and pTM scores
  • Model Comparison: Multiple prediction models ranked by quality
  • Interactive Dashboard: AI-powered visualization with quality analysis
  • Structure Export: High-quality models with confidence annotations

Analysis Capabilities

Structure Quality Assessment

Confidence Metrics

  • Overall Confidence: Combined score (0.8 × pLDDT + 0.2 × pTM)
  • Best pLDDT: Highest local confidence score
  • Best Confidence: Top-ranked prediction model
  • Multiple Models: Up to 11 prediction variants

Quality Classification

  • Excellent (>0.9): Very high confidence, reliable for analysis
  • Good (0.7-0.9): Confident regions, suitable for most applications
  • Moderate (0.5-0.7): Use with caution, focus on high-confidence regions
  • Poor (<0.5): Low confidence, experimental validation recommended

Example Results: FMC63 antibody prediction - Confidence: 0.640, pLDDT: 0.667, classified as Moderate Confidence with 11 model variants generated.

Interactive Dashboard Features

Summary Overview

  • Protein Information: Target name and prediction statistics
  • Quality Metrics: Best confidence and pLDDT scores displayed
  • Model Count: Number of prediction variants available
  • Confidence Classification: Overall prediction reliability assessment

Model Analysis

  • Model Gallery: All prediction variants with quality scores
  • PDB Downloads: Direct access to structure files
  • Quality Comparison: Side-by-side model evaluation
  • Confidence Visualization: Color-coded quality assessment

Integration Examples

Structure-Based Design

Sequence → Boltz-2 → Boltz Report
- Protein structure prediction from sequence
- Quality assessment with confidence metrics
- Best model selection for downstream analysis

Output Formats

Report Components

  • HTML Dashboard: Interactive confidence analysis and model comparison
  • Quality Assessment: Detailed prediction reliability evaluation
  • Summary Statistics: Key metrics and recommendations

Citations and References

  • Boltz-2 - Structure prediction engine

Tutorials