SVsearcher: A more accurate structural variation detection method in long read data

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Structural variations (SVs) are crucial genomic alterations involved in diseases and evolution.
• Long-read sequencing technologies like Oxford Nanopore (ONT) enable comprehensive SV detection.
• Existing SV callers struggle with ONT data accuracy, especially in repetitive and multi-allelic regions.

Purpose of the Study:

• To develop a novel method, SVsearcher, for accurate structural variation detection from Oxford Nanopore long-read sequencing data.
• To address limitations of current SV callers in handling high error rates and complex genomic structures in ONT reads.
• To improve the identification of multi-allelic structural variations.

Main Methods:

• Development and implementation of the SVsearcher algorithm.
• Comparative analysis of SVsearcher against existing SV callers using three real-world datasets.
• Evaluation of SV detection accuracy, focusing on F1 scores and multi-allelic SV identification rates.

Main Results:

• SVsearcher significantly enhances SV detection accuracy, improving F1 scores by ~10% (50× coverage) and >25% (10× coverage) compared to existing methods.
• SVsearcher demonstrates superior performance in identifying multi-allelic SVs, detecting 81.7%-91.8% compared to 13.2%-54.0% by other callers.
• The method effectively mitigates errors caused by ONT read alignments in repetitive and multi-allelic regions.

Conclusions:

• SVsearcher represents a substantial advancement in structural variation detection using Oxford Nanopore long-read sequencing.
• The tool offers improved accuracy and comprehensive identification of complex genomic rearrangements, particularly multi-allelic SVs.
• SVsearcher is a valuable resource for genomic research, disease studies, and evolutionary mechanism investigations.