Personalized Pangenome References

Area of Science:

• Genomics
• Bioinformatics

Background:

• Pangenomes aim to reduce reference bias by incorporating genetic diversity.
• Existing pangenome methods struggle with irrelevant variants, leading to mapping errors and reduced accuracy.
• Allele frequency filters are insufficient for accurately distinguishing relevant from irrelevant variants.

Approach:

• We introduce a novel approach inspired by local ancestry inference to create personalized pangenome subgraphs.
• This method samples local haplotypes based on k-mer counts in sequencing reads.
• The approach is optimized for the Giraffe short read aligner, enabling rapid index construction.

Key Points:

• Personalized pangenome pipelines offer faster read mapping compared to linear reference pipelines.
• Small variant genotyping error rates are reduced by 4x relative to the Genome Analysis Toolkit (GATK) best-practice pipeline.
• Achieves competitive short-read structural variant genotyping accuracy, comparable to long-read discovery methods.

Conclusions:

• This personalized pangenome approach significantly improves the accuracy and efficiency of genomic analysis.
• It addresses limitations of traditional pangenome methods, offering a more precise way to interpret genetic variation.
• Enables high-accuracy variant genotyping for both small and structural variants using short-read sequencing data.