A multilocus approach for accurate variant calling in low-copy repeats using whole-genome sequencing

Area of Science:

• Genomics
• Bioinformatics
• Human Genetics

Background:

• Low-copy repeats (LCRs), or segmental duplications, comprise over 5% of the human genome.
• Existing short-read variant calling tools struggle with accuracy in LCRs due to mapping challenges and copy number variations.
• Variants within LCRs are linked to over 150 human diseases, highlighting the need for improved detection methods.

Purpose of the Study:

• To develop a novel short-read variant calling method specifically designed for accurate variant detection within LCRs.
• To address the limitations of current tools in handling the complexities of duplicated DNA segments.

Main Methods:

• Introduced ParascopyVC, a method that performs joint variant calling across all repeat copies.
• Utilizes reads independent of mapping quality and aggregates mapped reads from different repeat copies for polyploid variant calling.
• Employs paralogous sequence variants to differentiate repeat copies and estimate genotypes for each copy.

Main Results:

• ParascopyVC demonstrated superior precision and recall on simulated data compared to state-of-the-art callers like DeepVariant and GATK.
• Benchmarking on the HG002 genome showed ParascopyVC achieved high precision (0.991) and recall (0.909) in LCR regions, outperforming FreeBayes, GATK, and DeepVariant.
• Across seven human genomes, ParascopyVC consistently achieved higher accuracy (mean F1 = 0.947) than competing methods.

Conclusions:

• ParascopyVC significantly enhances variant calling accuracy in low-copy repeat regions.
• The method provides a more reliable approach for identifying disease-associated variants within complex genomic structures.
• ParascopyVC is freely available, facilitating its adoption in genomic research and clinical applications.