Leveraging Human Pangenome for Improved Somatic Variant Detection.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Somatic variant detection is hindered by low variant allele fractions, germline variation, and reference bias.
• Linear references (e.g., GRCh38) fail to capture sample-specific genomic variations, leading to alignment and variant calling errors.
• Telomere-to-telomere donor-specific assemblies (DSAs) offer genomic accuracy but are limited by cost and technical feasibility.

Purpose of the Study:

• To benchmark somatic variant detection using GRCh38, graph-based pangenomes, and pangenome-inferred DSAs.
• To evaluate the effectiveness of pangenome strategies in improving read alignment and variant calling accuracy.
• To assess the capability of personalized pangenomes in reconstructing individual genomic content and mitigating germline contamination.

Main Methods:

• Benchmarking somatic variant detection across GRCh38, graph-based pangenomes, and pangenome-inferred DSAs.
• Utilized a HapMap mixture dataset and the COLO829 melanoma cell line for evaluation.
• Employed pangenome-guided alignment to assess read mapping and somatic variant calling performance.

Main Results:

• Pangenome-guided alignment significantly improved read mapping and somatic variant calling accuracy compared to GRCh38.
• Personalized pangenomes partially reconstructed donor-specific genomic content, enhancing accuracy and reducing germline contamination.
• Pangenome approaches enabled detection of variants in genomic loci poorly represented or absent in GRCh38.

Conclusions:

• Graph-based and personalized pangenomes represent effective strategies for enhancing somatic variant detection.
• These pangenome frameworks overcome the limitations of linear references, offering improved accuracy and broader genomic coverage.
• The findings highlight the potential of pangenomes for advancing precision oncology and genomic research.