TT-Mars

TT-Mars evaluates structural variant (SV) callsets by comparing them to haplotype-resolved genome assemblies to benchmark and classify SVs, including in repetitive regions where alignment-scoring metrics may conflict.

Key Features:

• Haplotype-resolved assembly comparison: Compares SV callsets against high-quality haplotype-resolved genome assemblies rather than direct call-to-call comparisons.
• Benchmark concordance with GIAB/truvari: Recapitulates Genome in a Bottle validations with 96.0%–99.6% concordance compared to truvari evaluations across multiple human genomes.
• Repetitive-region assessment: Assesses SV calls in repetitive genomic regions where traditional alignment-scoring metrics may yield conflicting results.
• Extended variant coverage: Evaluates additional variants beyond traditional benchmarks, ranging from 121 to 10,966 extra variants across different callsets.
• Performance distribution analysis: Quantifies SV calling performance over a distribution rather than reporting a single point estimate.
• Comparison to VaPoR and dipcall: Analyzes a greater number of calls on long-read callsets, including short variant calls (< 100 bases), requires smaller input data, and yields favorable results when candidate calls are fragmented into multiple alignments.

Scientific Applications:

• SV benchmarking: Evaluating the accuracy of structural variant discovery methods across diverse human genomic samples.
• Complex-region analysis: Improving reliability of variant calls in repetitive and otherwise complex genomic regions.
• Genomic research support: Supporting studies that require precise structural variation data, including research in genetic disease, population genetics, and evolutionary biology.

Methodology:

Compares SV callsets to haplotype-resolved genome assemblies to assess whether callset variants reflect assembly content, classifies true and false positive SVs across samples, and evaluates performance as a distribution while avoiding reliance on traditional alignment-scoring metrics.