Samovar: Single-Sample Mosaic Single-Nucleotide Variant Calling with Linked Reads.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Haplotype assembly is crucial for accurate variant detection.
• Linked-read sequencing offers advantages over standard paired-end sequencing for genomic analysis.
• Detecting mosaic single-nucleotide variants (SNVs) in a single sample remains challenging.

Purpose of the Study:

• To develop and evaluate a novel computational tool, Samovar, for enhanced haplotype assembly and mosaic SNV detection.
• To leverage linked-read sequencing data for improved variant phasing and identification within a single sample.
• To enable accurate detection of somatic variants in cancer genomics.

Main Methods:

• Samovar utilizes linked-read sequencing data to evaluate haplotype-discordant reads.
• A random forest model is trained using read quality, phasing information, and linked-read characteristics.
• The tool performs genome-wide phasing and mosaic SNV calling.

Main Results:

• Samovar achieves accuracy comparable to trio or matched tumor/normal analyses for mosaic SNV detection in single samples.
• The tool outperforms existing single-sample callers for minor allele frequencies between 5%-50% at >=30X coverage.
• Somatic variants were identified in 13 pediatric cancer whole-genome sequencing cases, with high recall corroborated by whole exome sequencing.

Conclusions:

• Samovar provides a powerful, open-source solution for accurate mosaic SNV detection and phasing using linked-read sequencing.
• The method significantly advances single-sample variant calling capabilities, particularly in cancer genomics.
• Samovar facilitates robust somatic variant discovery in pediatric cancer cases.