SNV Benchmarking Database - Comprehensive Variant Calling Evaluation
Overview: This dashboard provides systematic evaluation and comparison of variant calling performance across multiple sequencing technologies and computational pipelines. Each experiment represents a controlled benchmarking analysis using hap.py (a standard benchmarking tool) to compare variant calls against established genomic reference standards including GIAB (Genome in a Bottle) , CMRG , and T2T truth sets.
Technologies Covered: Compare performance across major sequencing platforms including Short-read sequencing ( Illumina , MGI ) and Long-read sequencing ( PacBio , Oxford Nanopore (ONT) ) using variant callers including DeepVariant (ML-based), GATK (Traditional), DRAGEN (Hardware-accelerated), and Clair3 (Long-read optimized).
How to Navigate:
• Use
sidebar filters
to narrow by technology or caller
•
Advanced Comparison
— filter by technology/caller combinations
•
Manual Selection
— click rows to pick specific experiments
• Expand
▶ arrows
for detailed experiment metadata
Variant Calling Performance Metrics
Performance Overview: Quantitative hap.py benchmarking results showing precision, recall, and F1-score for each technology-caller combination against validated truth sets.
Key Metrics: Precision measures accuracy of called variants (% true positives), Recall measures completeness (% of true variants detected), and F1-score provides balanced performance assessment. Higher percentages indicate better performance across all metrics.
Exploring Results:
• Results reflect experiments selected in
Tab 1
(filters or comparison modes)
• Use
truth set filter
above to focus on specific reference standards
• Click
column headers
to sort by any metric
• Each experiment shows two rows:
SNP
and
INDEL
variants
Performance Characterization Plots
Performance Plots: Precision vs. recall scatter plots with F1-score contour lines for visual comparison of variant calling performance. Curved lines represent constant F1-scores, helping identify optimal precision-recall balance. Each point represents one experiment, colored by sequencing technology ( Illumina , PacBio , ONT , MGI ) and shaped by variant caller ( DeepVariant , GATK , Clair3 , DRAGEN ).
How to Interact:
•
Click and drag
to zoom in |
Double-click
to reset view
•
Hover
for quick performance metrics
•
Click
points and
scroll down
to view detailed experiment metadata
Chart Reference
Stratified Performance Analysis
Regional Breakdown: Performance metrics displayed across genomic regions with different sequence characteristics. Available stratifications include complexity-based regions (easy/difficult), GC content ranges, functional annotations (coding/non-coding), repetitive sequences (homopolymers, tandem repeats, segmental duplications, satellites), and specialized regions (MHC, low mappability areas). These stratifications follow GIAB genome stratification standards .
How to Use:
• Results reflect experiments selected in
Tab 1
• Choose
genomic regions
below (expand ▶ sections for more options)
• Click
Update Analysis
to generate stratified results
Select Regions to Analyze
Primary Stratifications
▶ Repetitive DNA Regions
Simple Repeats:
Tandem Repeats:
Non-Repetitive:
▶ GC Composition
Low GC:
Normal GC:
High GC:
Extreme GC:
SNP Performance by Region
Loading plots...
INDEL Performance by Region
SNP Performance by Region
INDEL Performance by Region
No Data to Display
Please select some regions and experiments from previous tabs, then click 'Update Analysis'.