ASVBM: Structural variant benchmarking with local joint analysis for multiple callsets.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Accurate structural variant (SV) detection is crucial for human whole-genome sequencing (WGS) applications.
• Current SV benchmarking methods struggle with variant representation differences and capturing relationships between adjacent variants.

Purpose of the Study:

• To introduce ASVBM, an enhanced benchmarking framework for SV detection.
• To improve the accuracy and reduce false mismatches in SV benchmarking.

Main Methods:

• Developed ASVBM, incorporating latent positives and a joint analysis strategy based on local variants.
• Evaluated six state-of-the-art variant calling pipelines using real WGS datasets.
• Leveraged the equivalence of multiple smaller variants to a larger variant for improved matching.

Main Results:

• ASVBM reduces false mismatches by uncovering potential equivalences between callsets and benchmark sets.
• The joint analysis strategy improves SV detection benchmarking performance across multiple matching criteria.
• Demonstrated improved performance of SV detection pipelines using the ASVBM framework.

Conclusions:

• ASVBM offers a more robust approach to benchmarking SV detection in WGS data.
• The framework enhances the reliability of SV detection evaluation by addressing variant representation challenges.
• ASVBM facilitates advancements in human WGS analysis and SV detection tool development.