PanVariants

PanVariants integrates k-mer-based and mapping-based methods to detect known and novel single nucleotide variants (SNVs), indels, and structural variants (SVs) across diverse populations using pangenome references.

Key Features:

• Pangenome-based variant detection: Leverages pangenome references to enable high-precision detection of SNVs, INDELs, and SVs beyond variants stored in the graph.
• Hybrid calling approach: Combines k-mer-based and mapping-based methods to reduce variant errors, achieving a 73% reduction compared to BWA+GATK and a 45% reduction compared to DRAGEN for SNVs.
• DeepVariant retraining: Retraining the DeepVariant model with high-quality DNBSEQ data decreases errors by an additional 15%.
• Structural variant detection: Attains an F1-score of 89.39% for SVs, markedly outperforming DRAGEN (68.18%) and BWA+Manta (58.33%) and approaching long-read sequencing performance (95.22%).
• Cohort joint calling: Detects more variants, makes fewer Mendelian inheritance errors, and provides better per-sample accuracy than GATK in population-based analyses.
• Cluster execution support: Supports execution on both local machines and Sun Grid Engine (SGE) clusters.

Scientific Applications:

• Genome in a Bottle validation: Evaluates small variant (SNVs/indels) and structural variant accuracy in benchmark samples.
• Clinical variant detection: Successfully detects all expected pathogenic variants in clinical positive samples where PanGenie fails.
• Population genomics: Enables cohort-based joint calling with improved variant discovery and reduced Mendelian inheritance errors for diverse population studies.
• Diagnostic genomics: Validated for diagnostic applications requiring sensitive novel variant discovery and high accuracy for SNVs, indels, and SVs.

Methodology:

PanVariants integrates k-mer-based and mapping-based methods for variant detection, and retraining of the DeepVariant model with high-quality DNBSEQ data is performed to further decrease errors.