A graph clustering algorithm for detection and genotyping of structural variants from long reads

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Structural variants (SVs) are genomic alterations (>50 bp) including deletions, insertions, and translocations.
• SVs play crucial roles in phenotypic variation and evolution.
• Accurate SV detection is vital for genomic analysis, necessitating advanced computational tools.

Purpose of the Study:

• To develop an accurate and efficient algorithm for germline structural variant (SV) prediction from long-read sequencing data.
• To improve SV calling and genotyping, particularly in challenging genomic contexts.
• To leverage long-read sequencing technologies for comprehensive genomic analysis.

Main Methods:

• An algorithm was developed to collect SV evidence (signatures) from read alignments.
• Signatures were clustered using a Euclidean graph and the DBSCAN algorithm for high-resolution identification.
• A Bayesian model was employed for precise SV genotyping based on supporting evidence.

Main Results:

• The developed algorithm demonstrated superior performance compared to state-of-the-art tools in germline SV calling and genotyping.
• Outperformance was particularly notable at low sequencing depths and in error-prone repetitive genomic regions.
• The approach effectively integrates into existing genomics analysis platforms.

Conclusions:

• The study presents a significant advancement in bioinformatic strategies for SV detection using long-read sequencing.
• The algorithm enhances the utility of long-read sequencing for understanding genomic variation.
• This work contributes to more robust and accurate germline SV analysis.