Scalable and comprehensive mosaic variant calling using DRAGEN.

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Detecting low variant allele fraction (VAF) mosaic variants without controls is challenging due to technical noise and scalability issues.
• Existing methods lack robust benchmarks and struggle with computational efficiency for low-VAF detection.

Purpose of the Study:

• To develop a hardware-accelerated method for accurate and scalable detection of low-VAF mosaic variants.
• To establish a genome-wide benchmark for evaluating low-VAF variant callers.
• To investigate mosaic mutation patterns in human tissues and assess the origin of previously identified mosaic variants.

Main Methods:

• Development of the DRAGEN mosaic caller, a hardware-accelerated tool for single nucleotide variant (SNV) and indel detection.
• Creation of a genome-wide benchmark dataset for variants in the 1-10% VAF range.
• Application of the caller to bulk sequencing data from blood, sperm, and brain tissues.
• Utilizing personalized assembly pangenome references for improved alignment in complex genomic regions.

Main Results:

• The DRAGEN mosaic caller identifies variants down to 1-2% VAF with low false-positive rates and hour-scale runtimes.
• A new genome-wide benchmark for low-VAF variants (1-10%) was introduced.
• Analysis of HG002 blood revealed that many previously identified mosaic variants in cell lines may be culture-derived.
• Mosaic hotspots and mutational signatures were identified across different tissues.
• Personalized pangenome references enhanced variant detection in complex genomic areas.

Conclusions:

• The DRAGEN mosaic caller makes routine low-VAF mosaic variant discovery feasible.
• This advancement opens new avenues for studying mosaic mutations in both healthy and diseased individuals.
• The findings highlight the importance of distinguishing in vivo mutations from culture artifacts.