VG-Pedigree

VG-Pedigree integrates pangenome graphs and pedigree-aware analysis to reduce reference bias and improve genotyping and variant detection for clinical genetics and rare-disease family studies.

Key Features:

• Pedigree-Aware Analysis: Tailors analysis to familial relationships, including support for family quartets and quintets to leverage inheritance information.
• Pangenome Mapping with Giraffe: Uses Giraffe (Sirén et al., 2021) to map reads to pangenome graphs that embed population variation and reduce linear-reference mapping bias.
• Variant Calling with DeepTrio: Integrates DeepTrio (Kolesnikov et al., 2021) with a specially trained model optimized for Giraffe-based alignments to improve detection of SNVs and INDELs.
• Improved Detection of Deleterious Variants: Incorporates adapted and upgraded methods from Gu et al. (2019) to streamline detection of deleterious variants (DVs), detecting a slightly greater number of DVs in probands versus their unaffected siblings.
• Benchmarking Against Linear Methods: Demonstrates mapping and variant-calling improvements over BWA-MEM with a linear reference and over Giraffe mapping alone.

Scientific Applications:

• Clinical genetics for rare disease: Enhances genotyping and variant detection in family-based rare-disease cases to aid interpretation of probands and relatives.
• Family-based variant analysis: Enables pedigree-aware identification of SNVs, INDELs, and deleterious variants in family quartets and quintets.
• Comparative variant assessment: Facilitates comparison of variant calls between probands and unaffected siblings to improve detection sensitivity for pathogenic variants.

Methodology:

Maps reads to pangenome graphs using Giraffe (Sirén et al., 2021), performs variant calling with DeepTrio (Kolesnikov et al., 2021) using a model trained for Giraffe alignments, applies adapted methods from Gu et al. (2019) for deleterious-variant detection, and evaluates performance relative to BWA-MEM with a linear reference and Giraffe mapping alone within a pedigree-aware framework.