SV-JIM, detailed pairwise structural variant calling using long-reads and genome assemblies.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Structural variants (SVs) are crucial genomic alterations.
• Accurate SV detection across diverse callers remains challenging.
• Existing methods lack standardized data-driven evaluation frameworks.

Purpose of the Study:

• To develop and validate a novel software pipeline, SV-JIM, for assessing structural variant (SV) caller performance.
• To enable a data-driven comparison of multiple SV callers using genome assemblies and long-read sequencing data.
• To improve the consistency and reliability of SV detection through caller aggregation.

Main Methods:

• Implementation of the Structural Variant - Jaccard Index Measure (SV-JIM) pipeline using Snakemake.
• Utilizing genome assemblies and long-reads for SV calling.
• Employing the Jaccard index to measure consistency among SV caller outputs.
• Aggregating SV sets based on caller support for enhanced data interpretation.

Main Results:

• SV-JIM identified significant inter-caller variance in SV detection across human and plant genomes.
• Aggregated SV sets improved the retention of rare SV types by implementing minimum caller support thresholds.
• Case studies revealed potential inflation in precision reporting during SV caller evaluations.

Conclusions:

• SV-JIM provides a robust framework for data-driven SV caller assessment.
• The pipeline enhances the reliability of SV detection by consolidating results from multiple callers.
• Findings highlight the importance of standardized evaluation metrics and aggregation strategies for accurate genomic variant analysis.