Sawfish: improving long-read structural variant discovery and genotyping with local haplotype modeling

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Structural variants (SVs) are crucial in genomics but difficult to characterize.
• Long-read sequencing enhances SV analysis, yet caller accuracy can be improved.
• Systematic local haplotype modeling offers a path to higher SV calling accuracy.

Purpose of the Study:

• To introduce sawfish, a novel structural variant caller for long-read sequencing data.
• To enhance the accuracy and resolution of SV discovery and genotyping.
• To improve SV characterization through systematic SV haplotype modeling.

Main Methods:

• Developed sawfish, a structural variant caller for mapped high-quality long reads.
• Incorporated systematic SV haplotype modeling into the calling algorithm.
• Evaluated sawfish performance against benchmarks like Genome in a Bottle (GIAB).

Main Results:

• Sawfish demonstrated the highest accuracy among state-of-the-art long-read SV callers across all tested SV sizes.
• Maintained superior accuracy across various sequencing depths (10- to 32-fold coverage).
• Achieved higher accuracy in medically relevant gene regions and improved pedigree-concordant calls in joint-genotyping analyses.

Conclusions:

• Sawfish significantly improves the accuracy and resolution of structural variant calling from long-read sequencing data.
• The method offers enhanced performance for both individual and joint-sample analyses.
• Systematic SV haplotype modeling represents a key advancement in SV characterization.