Metrics

This section documents the metric families currently intended for user-facing evaluation.

• Metrics Overview
  • Available Metric Families
  • Cross-cutting conventions
  • Recommended Online Subset
• Conventions
  • Coordinate semantics
  • Strand awareness
  • Aggregation across sequences
  • Sentinel values
  • Annotation modes and scopes
  • Background and intron labels
  • Plot conventions
• Region Discovery
  • How Sections Are Matched
  • Output Structure
  • False positives — the same rule for every tier
  • The Four P/R Tiers
  • Double-Penalty Behavior
  • Interpretation
  • Caveats
• Boundary Exactness
  • What Is Computed
  • IoU
  • Boundary Bias and Cumulative Recall
  • One-Sidedness Decomposition
  • Terminal Boundary Flags
  • Caveats
• Nucleotide Classification
  • Example Plot
  • What It Computes
  • Interpretation
  • Caveats
• Structural Coherence
  • Intron/Exon Chain Match Rate (Multi-Exon Only)
  • Single-Exon Gene Match Rate
  • Transcript Classifications
  • Boundary Shift Distribution
  • Segment Count Delta
  • Exon Recovery
  • Splice-Site Junctions
  • Caveats
• Diagnostic Depth
  • Example Plots
  • What Is Computed
  • Interpretation
  • Aggregation
  • Caveats
• Transition Analysis
  • GT Transition Confusion Matrices
  • False Transitions
  • When To Look At These Plots
  • Caveats
• INDEL
  • Example Plots
  • Categories
  • How Classification Works
  • Exon-Position Typing
  • Output Structure
  • Rates and opportunities
  • Interpretation
  • Caveats
• Phase Drift
  • What Is Computed
  • Skipping (not raising) on degenerate inputs
  • Interpretation
  • Caveats
• Global (Annotation-Level) Metrics
  • The eight groups
  • Scoped vs unscoped