Tradeoffs in alignment and assembly-based methods for structural variant detection with long-read sequencing data

landmaba
landmaba

Apr 16, 2024, 2:58 AM

Liu YH, Luo C, Golding SG, Ioffe JB, Zhou XM. Tradeoffs in alignment and assembly-based methods for structural variant detection with long-read sequencing data. Nat Commun. 2024 Mar 19;15(1):2447. doi: 10.1038/s41467-024-46614-z. PMID: 38503752; PMCID: PMC10951360.

Researchers have systematically evaluated a range of tools designed to detect structural variants (SVs) in genomes using long-read sequencing, a method that provides more comprehensive genomic insights. The study compares 14 alignment-based methods, including advanced deep learning options, with four assembly-based methods, revealing that while assembly-based tools more effectively detect larger SVs and are robust against various testing conditions, alignment-based methods offer greater accuracy at lower sequencing coverages and excel at identifying complex SVs. This benchmarking effort helps users select the most suitable tools for different research scenarios and lays a foundation for future improvements in genomic analysis tools.

Complex SV detection in simulated and real cancer datasets. a Heatmap shows overall and genotyping (gt) F1 scores of translocation (TRA), inversion (INV), and duplication (DUP) detection for 10 SV calling methods on 9 simulated PacBio Hifi, CLR, and ONT datasets. b Heatmap shows recall and precision scores of somatic deletion (DEL), insertion (INS), translocation (TRA), inversion (INV), and duplication (DUP) detection for 9 SV calling methods on two publicly available sets of Tumor-Normal paired Pacbio CLR and ONT libraries. Empty cells represent analysis that could not be performed (or finished within 14 days of runtime) for the tool in the corresponding row. Source data are provided as a Source Data file.

Explore Story Topics